International Market Research

International Market Research Lectures Nikolay V. Pavlov 1. Введение 1.1. Общее описание курса Курс рассчитан на 36 аудиторных часов и самостоятельную работу. Включает теорию и практику. Современные компетенции включают в себя знания и умения. Подробнее можно посмотреть учебную программу. Из опыта: Маркетинговые исследования = 33% можно догадаться + 33% знания + 33% математические методы. Есть много вещей, которые надо просто знать. Теория охватывает все важные вопросы. В общем случае есть субъект (индивидуальный предприниматель, подразделение фирмы, малая или большая фирма, фонд, даже правительство), которому требуется принять определенные маркетинговые решения. То есть вопрос заключается в том, что делать. Здесь будет использовано следующее определение маркетинга From [Kotabe M., Helsen K. GLOBALMARKETING MANAGEMENT JOHNWILEY& SONS, INC., 2010, 751P.]. What is marketing Marketing is essentially the activity, set of institutions, and processes for creating, communicating, delivering, and exchanging offerings that have value for customers, clients, partners, and society at large. Marketing is not only much broader than selling, it also encompasses the entire company’s market orientation toward customer satisfaction in a competitive environment. Есть маркетинговые цели. Есть маркетинговая стратегия (способ достижения целей). Получается, что маркетинговая стратегия требует внимания и к потребителям, и к конкурентам. Если только удовлетворять, конкуренты обойдут более дешевым, более удобным продуктом (что такое продукт). Например, компании США обнаружили потребности, удовлетворили их, но были обойдены европейскими и японскими конкурентами. В этом определении имеется в виду отдельная организация, компания. Это и будет взято за основу. А остальное – как дополнения. В дальнейшем будет использовано более краткое определение: Маркетинг это выгодное для организации удовлетворение потребностей потребителей. Далее. Есть исследователи, которым нужно поставить вопрос: что нужно узнать, чтобы принять решение. Далее данные необходимо собрать и дать рекомендации. Вот суть работы исследователя. В международном аспекте бывает, что нужно узнать либо о какой-то отдельной стране, либо о многих странах. Для такой деятельности созданы специализированные исследовательские фирмы. В данном курсе 1) рассматриваются методы и приемы исследований, то есть деятельности этих фирм; 2) но важно рассмотреть эту деятельность и извне. Чтобы сделать заказ таким фирмам, очень полезно понимать суть их деятельности. В практической части необходимо будет провести исследование уровня выхода отдельно взятой фирмы на рынок другой страны. Будет охвачен цикл от поиска возможностей на рынке другой страны до разработки рекомендаций по собранным данным. Подход к практической работе: ориентация на реальную ситуацию. То есть будут не примеры с известным ответом, а реальные исследования. Если нет особых пожеланий, будет рассмотрен выход на российский рынок как на иностранный. 1.2. Цели курса Сверхзадача: хотелось бы, чтобы были найдены, определены и освоены некоторые принципы, более широкие, чем курс. Это 1. Если что-то изучается, надо понимать место, роль, цели изучения. 2. Осваиваем процесс исследования а примере маркетингового исследования. 3. Уровень магистров: не просто уметь вести интервью, а понимать, для чего это делается, к какому результату должно прийти исследование. 4. Ориентация на результат, а не на процесс. Задачи в реальной жизни слабоструктурированные. Нельзя всегда работать по принципу: вот исходные данные, вот метод решения, найдите ответ. Например, сколько будет 3+4? Можно на пальцах, можно по таблице, можно запомнить. Чаще бывает, что непонятно что делать и тем более непонятно, как. Поэтому в Википедии даже говорится: all … forms of marketing research can be classified as either problem-identification research or as problem-solving research. 5. В рамках курса главная задача: уметь дать обоснованные рекомендации по принятию решений. Тут много имеется в виду: определить, какие рекомендации нужны; определить, какие данные для этого собрать, как и где, как обработать, как сделать вывод. И все это в условиях ограничений по времени (1 семестр) и затратам (только свой труд). Цели курса – см. Слайд. 1.3. Литература Слайды. 1.4. Оценка слушателей Экзамен (классический) 50%; Презентация проведенного исследования 35%; Посещаемость/работа на занятиях 15%. 1.5. Правила посещения и поведения Не опаздывать на занятия. Опоздание свыше 15 минут = допуск, но отмечается отсутствие. Громко не разговаривать. Не носить еду и напитки в компьютерный класс. Верхнюю одежду вешать на вешалку. Если есть вопрос и уточнение – поднимать руку. 1.6. Дополнительные источники информации для слушателей Главный источник – Интернет, где осуществляется самостоятельный поиск. Ниже приводятся некоторые источники, для начала. Ряд книг и статей из открытого доступа в Интернет можно взять у преподавателя. http://cindyking.biz/category/international-marketing/ - надо искать от типов до фильмов http://datalib.chass.utoronto.ca/other/findints.htm - Finding U.S. and International Statistics. http://demoscope.ru/weekly/2011/0471/index.php журнал демоскоп http://ec.europa.eu/internal_market/payments/einvoicing/index_en.htm E-invoicing: (overview) http://europa.eu.iat/ – Европейский союз http://export.gov/ Helping US Companies export есть сведения по странам http://fita.org/ - The Federation of International Trade Associations. http://humanearth.h1.ru/ - Сайт «Проблемы человечества». http://pdfgeni.tradepub.com/ - бесплатные журналы http://wnr.economicus.ru/ – Сравнение уровней развития стран. http://www.alkearney.com/ – Индекс доверия FDI. http://www.allaboutmarketresearch.com/ ресурсы http://www.allworld.wallst.ru/ - Экономическая информация по странам; http://www.aonb.ru/iatp/guide/stat.html - много о статистике разных стран http://www.b2binternational.com/ - много всяких материалов http://www.bmz.de/ - Министерство экономического сотрудничества и развития; http://www.businesspress.org/ http://www.chinatoday.com/ Информационная - Информационная база по КНР; http://www.cisstat.com/ - Международный статистический комитет СНГ; http://www.datacenter.utorauto.chass.ea/ - сайт статистической информации; http://www.destatis.de/ - Федеральная статистическая служба; http://www.ebookee.org – можно скачать много http://www.ec.org/ - сайт ЕС; http://www.ecology.ru/ - Глобальные экологические проблемы. http://www.ein.cm/ – Служба странового риска; http://www.ek.fi/www/en/index.php weekly http://www.emeraldinsight.com/index.htm - немного по менеджменту и МИ http://www.euromoneyplc.com/ – Рейтинги кредитоспособности стран; http://www.firstgov.gov/ - официальный правительственный портал США; http://www.freeonlineresearchpapers.com/ http://www.freetheworld.com/ – Экономическая свобода в мире; http://www.gks.ru/wps/wcm/connect/rosstat/rosstatsite.eng/ - Госкомстат Federal State Statistic Service http://www.gksoft.com/govt/en/ch.html - Официальный ресурс КНР; http://www.gksoft.com/govt/en/cl.html - официальный ресурс Чили; http://www.gksoft.com/govt/en/de.html - официальный ресурс Германии; http://www.gksoft.com/govt/en/jp.html - Ресурсы Японии; http://www.gksoft.com/govt/en/us.html - официальные ресурсы США; http://www.gksoft.com/govt/gb.html - Официальный ресурс Великобритании; http://www.heritage.org/ – Индекс экономической свободы; http://www.humanearth.hl.ru/ - сайт Проблемы человечества; http://www.iccwbo.org/ – Международная торговая палата http://www.iet.ru/ - Институт экономики переходного периода; http://www.imd.ch/ – Ежегодник мировой конкурентоспособности; http://www.imfo.org/external/np/sec/decdo/sela.html - Экономическая система стран Латинсокй Америки; http://www.inc.com/market-research/35 - много статей по исследованиям, но мало международных http://www.internetworldstats.com/ - Internet world stats Usage and population statistics ссылки на данные и немного руководств. http://www.islamic-finance.net/ - Исламский бизнес; http://www.ita.doc.gov/ – Департамент торговли США http://www.kpmg.com/Global/en/IssuesAndInsights/ArticlesPublications/Documents/Corp-and-Indirect-Tax-Oct12-2010.pdf Taxation in different countries: (overview) http://www.kpmg.com/Global/en/WhatWeDo/Special-Interests/Business-outlook-survey/Winter-2010/Pages/default.aspx Global business outlook: (overview) http://www.marketologi.ru/publ.html – Сайт гильдии маркетологов. Публикации о маркетинге, рынках и исследованиях http://www.marketresearchworld.net много, но бестолково http://www.markinvest.ru/rus/links.html - ссылки для международных компаний по России (от фирмы маркинвест Спб) http://www.mendeley.com/research-papers/ - очень много статей, Ми найти трудно, сотни Для скачивания нужна регистрация http://www.nielsen.com/us/en/insights/reports-downloads.html - отчеты и статьи, но надо регистрироваться http://www.oecd.org/ – организация экономического сотрудничества и развития http://www.omirussia.ru/en/analytics/publications/ - онлайновые исследования. Есть несколько статей на рус и англ, надо регистрироваться http://www.piter-consult.ru/services/How-many-marketing-research-cost/Marketing-researches.html- примеры исследований на русском (демо) http://www.sba.gov/ - US Small Business Administration http://www.service-public.fr/accueil/english.html - официальные ресурсы Франции, ЕС и международных организаций; http://www.statinfo.biz/ - международная экономическая статистика http://www.statistics.gov.uk/ - Официальная статистика; http://www.suomenpankki.fi/bofit_en/Pages/default.aspx Economic situation and outlook in Finland and Russia: (overview) http://www.un.org/ - United Nations (International Trade Statistics); http://www.unep.org/ - GEO Data Portal – The online Environmental Database. http://www.univer.kz/econom/ - Классификация стран мира по экономическим признакам. Под ред. Р.Е. Елемесова. http://www.warc.com/ - для консалтинга. Мало без подписки, нужна регистрация и какой-то trial period http://www.wdi.ru/ - Сайт “World development”. http://www.wecon.ru/www.worlbank.org - World Development Report http://www.weforum.org/ – Всемирный экономический форум; http://www.worldeconomy.ru/ – Сайт «Мировая экономика» http://www.worldmarketsanalysis.com/ – Центр изучения мировых рынков; http://www.wto.org/ – Всемирная торговая организация https://www.cia.gov/library/publications/the-world-factbook/rankorder/rankorderguide.html - CIA Factbook www.economicus.ru/glossary.shtml - краткий словарь терминов, сравнение уровней развития стран; О России http://www.barentsinfo.org/ - Международный информационный портал "Информационная служба Баренц-региона" http://www.cbr.ru/ - Центральны банк Российской Федерации http://www.ebrd.org/ - Европейский банк реконструкции и развития http://www.eeg.ru/ - Экономическая экспертная группа (ЭЭГ) - независимый аналитический центр http://www.fom.ru/ - Фонд Общественное Мнение http://www.garant.ru/ - ГАРАНТ: компания "Гарант", система ГАРАНТ, информационно-правовое обслуживание ГАРАНТ http://www.gost.ru/ - федеральное агентство по техническому урегулированию и метрологии http://www.gov.spb.ru/ - официальный портал Администрации Санкт-Петербурга http://www.iet.ru/ - Институт экономики переходного периода, независимая некоммерческая научно-исследовательская организация http://www.lenobl.ru/ - официальное представительство Ленинградской области http://www.leontief.ru/ - Международный центр социально-экономических исследований Леонтьевский центр http://www.markinvest.ru/ - ГОСТ Р сертификаты http://www.northerndimension.net/ - The Advisory Network for the Northern Dimension http://www.resep.ru/ - Российско-Европейский Центр Экономической Политики (РЕЦЭП) http://www.rus-forum.ru/ - информационный портал для международных компаний, которые планируют заниматься экспортом и/или импортом в России http://www.rusin.fi/info/infoRUS.htmL - Институт России и Восточной Европы http://www.russiainfo.org/ - информационный навигатор о России на английском и финском языках http://www.vniiki.ru/ - информационный сайт о ГОСТ, ГОСТ Р, ПР, Р, РД, стандартам международных организаций(ИСО, МЭК) и национальным стандартам зарубежных стран. http://www.wciom.ru/ - Всероссийский центр изучения общественного мнения (ВЦИОМ) http://www.ey.com/cis - компания "Ернст энд Янг", Юридическая информация 1.7. Содержание курса 1.7.1. Что такое МИ Подход: исследования  маркетинговое исследование  международное маркетинговое исследование. European Market research = American Marketing Research. Research for Marketing. Research of markets is a part of Marketing Research. Мы будем изучать то, что помогает компании принимать решения в международной деятельности. В общем случае маркетинговые исследования бывают различных типов [2005_525P_Craig_International Marketing Research] (Обсудить). За основу возьмем часть только для коммерческих целей. From Churchill. Структура маркетинга по [31] представлена на рис.Рис. 1. В центре маркетинговой активности находится потребитель. Его окружает комплекс маркетинга: продукт, цена, продажа, продвижение1. Это объект деятельности маркетологов фирмы. Рис. 1. Структура маркетинга Факторы внешней среды – законы и политическое окружение, технология отрасли, конкуренция, экономическое окружение, культурная и социальная среда – исследуются и учитываются специалистами по маркетингу, но управление ими они обычно не производят. Осуществление маркетинговой деятельности показано на рис.Рис. 2. На рисунке стрелками показано движение информации, с чем и связаны маркетинговые исследования. Рис. 2. Маркетинговая деятельность Маркетинговые исследования – это деятельность, которая связывает потребителя, покупателя, внешнюю среду фирмы с маркетинговыми структурами через информацию, предназначенную для: обеспечения достижения целей фирмы; мониторинга внешней среды; разработки, оценки и корректировки маркетинговых действий на основе изучения потребителей; определения возможностей и проблем маркетинга. Традиционно исследования в строгом смысле слова связывались с естественными науками, прежде всего с физикой. Достижения в области теории и практики маркетинга позволяют с полным основанием использовать термин исследования и применительно к маркетингу. Однако маркетинговые исследования имеют свои особенности. Маркетинговые исследования носят прежде всего прикладной характер. Цели маркетинговых исследований и решаемые при исследованиях вопросы практически всегда оказываются связанными с деятельностью фирмы (Табл. В.2)2. Сущность маркетинговых исследований. Таблица В.2 1. Связь маркетинговых исследований с деятельностью фирмы Деятельность фирмы Цели маркетинговых исследований Основные вопросы, решаемые при исследованиях Планирование Определить возможности рынка Кто покупает товар, где производятся покупки? Сколько покупателей у товара? Каковы доходы покупателей? Расширяется или уменьшается рынок? Существуют ли другие перспективные рынки? Каково состояние каналов распределения? Можно ли улучшить каналы распределения? Решение проблемы Дать информацию для управления комплексом маркетинга Продукт Номенклатура выпускаемых продуктов Какую выбрать упаковку? Каким должно быть качество продукта? Цена Какую цену установить на товар? Как изменить цену? Продажа Где и кем продаются товары? Как поощрить торговлю? Продви-жение Размер затрат на продвижение Как распределить эти средства между товарами и регионами? Какие средства выбрать для рекламы (газеты, журналы, телевидение, радио)? Управле- ние Выявить проблемы, спланировать действия Доля рынка фирмы (в целом; по районам; по сегментам) Каковы степень удовлетворенности покупателей, состояние сервиса, процент возврата товаров? Каков имидж фирмы, какова ее репутация? 1.7.2. Особенности МИ Если в физике полезным является и отрицательный результат, то маркетинговые исследования призваны прояснить и решить конкретную проблему, дать вполне определенные рекомендации по исправлению или улучшению ситуации в фирме. Поэтому в данной области наряду с чисто формальными и точными, используются неформализуемые методы, а результат не всегда безупречно точен. В некоторых случаях получается только качественный результат типа скорее всего, данная рекламная кампания принесет прибыль, а не убытки. Практически всегда исследователи работают в условиях жестких ограничений как по времени, так и по выделяемым средствам. Особую важность приобретает получение наиболее полной и достоверной информации исходя из имеющихся ресурсов. В отличие от физических объектов и явлений, маркетинговые исследования имеют дело с крайне изменчивыми и неповторяющимися социальными и психологическими процессами. Если физический эксперимент можно повторить в любое время, точно воспроизведя условия его проведения, то любое исследование вкусов, пристрастий и мнений потребителей является уникальным. Если его провести даже через неделю, результат будет уже другим, так как измеряемая величина изменилась. Поэтому особую важность приобретает корректность используемых методов и тщательность его проведения. Профессионализм исследователя рынка проявляется в том, что он не только правильно интерпретирует полученные данные, но и подробно описывает, как именно проводилось исследование: каковы были задаваемые вопросы, как именно они задавались. Именно по этому описанию специалисты могут оценить достоверность полученных результатов и применимость их для своих целей. Если в физике имеется большое число различных измерительных приборов, от весов и термометров до электронного микроскопа, то в области маркетинговых исследований приходится практически каждый раз разрабатывать новые инструменты исследования. Речь идет прежде всего о главном инструменте – анкете. Вот почему важно освоить как теоретические основы, так и наиболее эффективные приемы проведения маркетинговых исследований. 1.7.3. Суть МИ Тут тоже есть неоднозначность. Ряд авторов считают, что это систематическая деятельность. Вот что предлагает Черчилль. Для получения информации, необходимой для принятия управленческих решений, сегодня используют ряд средств. Маркетинговая информационная система (MIS). Согласно [17], представляет собой «индивидов, оборудование и процедуры сбора, сортировки, анализа, оценки и распределения используемой при принятии маркетинговых решений своевременной и достоверной информации». Это может быть как информация о внутреннем состоянии фирмы, так и данные о конкурентах, спросе, поставщиках. Информация собирается в базах или банках данных, нужные сведения представляются в форме таблиц (отчетов). Особенности MIS: • формат входной информации и способ хранения четко определены; • данные в базах данных постоянно обновляются; • задачи по обработке данных запрограммированы и не требуют от пользователя квалификации программиста; • данные представляются в заранее заданном виде всем, кому они нужны; • параметры отчетов фиксированы. Наиболее существенна последняя особенность. В любой момент можно, нажав одну-две кнопки, получить новейшие сведения, которые будут представлены в привычном виде3. Недостатками MIS является то, что их разработка дорога и сложна. Она выполняется высококвалифицированными программистами. В то же время требования к информации для принятия решений постоянно меняются. Менеджеры (особенно высшего звена) часто заранее не знают, какая именно информация им потребуется, а быстрый доступ к свежей информации при принятии решений в нестандартных, критических ситуациях имеет решающее значение. Многие проблемы принятия управленческих решений относятся к плохо структурированным4, они связаны с персональным выбором и ответственностью за него. Принятие управленческого решения, особенно стратегического – творческий процесс. Система стандартизированных форм представления данных недостаточно гибка для решения таких задач. Информационно-поисковые системы являются разновидностью MIS и служат для быстрого поиска информации, содержащейся в основном в текстовых документах. Для этого в них применяются специальные средства. Примерами таких систем являются Rambler (http://www.rambler.ru/) и Yandex (http://www.yandex.com/). К этому же типу можно отнести и информационно-правовые системы, например, Кодекс и Консультант плюс, содержащие постоянно обновляемую базу данных, хранящую законы, указы и нормативные документы. Системы поддержки принятия решений (DSS5). DSS, которые все чаще используются в настоящее время, – это скоординированный набор данных, систем, инструментов и технологий, программного и аппаратного обеспечения, с помощью которого в организации под управлением пользователя6 собирается и обрабатывается информация о бизнесе и окружающей среде с целью обоснования маркетинговых действий. DSS состоит из трех основных частей. 1. Система данных для сбора и хранения информации о маркетинге, финансах и производстве, получаемой из внутренних и внешних источников. Обычно это база или банк данных, как и в MIS. 2. Система диалога, позволяющая пользователю задавать, какие данные следует выбирать и как их обрабатывать. 3. Система моделей – идеи, алгоритмы и процедуры, которые позволяют обрабатывать данные и проводить их анализ. Пользователь имеет опыт, знает ситуацию и руководствуется определенными соображениями при выборке данных. В обработке данных используются различные процедуры, от простого суммирования до статистического анализа и нелинейной оптимизации. Типовыми процедурами являются: • объединение в группы; • получение сводных показателей; • ранжирование; • выделение особых случаев; • графическое представление данных. Несмотря на кажущуюся простоту, важность процедур последнего типа трудно переоценить. Иногда достаточно только взглянуть на графическое представление данных, чтобы понять, даст ли хороший результат кластерный анализ, какой вид регрессионной функции выбрать и т.д. Модели принятия решений служат для обработки данных, нужных для решения, и по способам представления результатов своей работы подразделяются на информационные (что есть и что будет, если...), советующие (в меру своего «разумения») и (редко) управляющие. Типы моделей представлены в табл. В.1. В настоящее время идеи DSS получили свое дальнейшее развитие. Прогресс в области вычислительной техники сделал возмжным новые подходы к анализу данных. Одним из новых направлений является онлайновая аналитическая обработка данных (OLAP7). Данные обычно берутся из уже существующих баз данных и подвергаются быстрому, но достаточно поверхностному предварительному разведочному анализу [13]. В OLAP обычно используется многомерная модель данных. Это позволяет гибко манипулировать информацией, но требует довольно серьезной специальной подготовки. Таблица В.1 Типы моделей в DSS и их особенности Тип модели Особенности Примеры применения Уровень менеджмента Используемая информация Период моделирования Рабочие Низший Внутренняя День, неделя Обоснования скидок Рекомендации для продавцов Тактические Средний Внешняя Месяц, год Ценообразование Субъективные данные Выбор средств рекламы Стратегические Высший Внешняя Обычно годы Оценка объема продаж нового продукта Субъективные данные Решение о снятии продукта с производства Кроме ставших уже традиционными реляционных баз данных, в последнее время развиваются более гибкие и универсальные средства, получившие название хранилища данных8. В них могут находиться не только табличные данные, но и данные других типов (текстовые, графические, звуковые и т.д.), чем достигается полное представление информации об организации. Работы в этой области еще весьма далеки от завершения. Для обработки данных, в том числе и находящихся в хранилищах, предложена концепция Data Mining – «добычи данных»9. Это, согласно [13], «процесс обнаружения в сырых данных ранее неизвестных; нетривиальных; практически полезных; доступных интерпретации знаний … для принятия решений». Новизна подхода заключается в том, что современные мощные компьютеры в состоянии переработать огромные массивы данных и найти в них что-то полезное. Однако не следует считать, что компьютер полностью заменяет исследователя-человека. Наоборот, применение методов Data Mining – процесс, требующий от исследователя глубоких знаний. Система Data Mining требует четко согласованной работы всех своих компонентов. Пользователь должен быть квалифицированным специалистом в таких областях, как работа с базами данных, анализ данных как традиционными математическими методами, так и с использованием средств искусственного интеллекта. Наконец, интерпретация полученных данных и использование полученных результатов также остаются прерогативой человека. Необходимо отметить, что в настоящее время лишь небольшое число российских фирм имеет хорошо организованные, достаточно полные и длительно ведущиеся базы данных, что затрудняет применение этих средств. Значительно помогает в работе маркетолога использование сканеров штрих-кодов, которые, помимо ускорения оплаты товаров в кассе, позволяют автоматизировать сбор больших объемов полезной информации об объемах продаж, ее динамике, совместно покупаемых товарах и т.д. Автоматизированные средства помогают в принятии стратегических решений, позволяют получить информацию о текущем состоянии фирмы, весьма хороши для раннего предупреждения о возникающих проблемах. К сожалению, они не дают подсказки в специальных, «нестандартных» случаях (что делать с новым товаром, как оптимизировать каналы товародвижения и т. д.). Для сбора информации по отдельным конкретным проблемам, для поиска новых идей и гипотез используются маркетинговые исследования. Правильно проведенные, они дают исчерпывающую информацию о ситуации на рынке или о проблеме, возникшей в фирме. Итак, Черчилль говорит, что если MIS можно уподобить свече, которая горит долго, но довольно тускло, то маркетинговые исследования уподобляются фотовспышке, которая освещает всё очень ярко, но очень ненадолго. Обычно рассмотренные средства используются совместно. 1.7.4. Задачи и этапы МИ Основными задачами маркетинговых исследований являются: определение вида требуемой информации; осуществление ее сбора; анализ результатов; выдача рекомендаций по применению полученных результатов. В управлении экономическими системами обычно используются два уровня планирования. Программа маркетинговых исследований определяет тип исследования и его цели. В некоторых случаях требуется длительное наблюдение за объемами продаж, долей рынка. В других – проводится единовременное исследование для анализа причин возникновения проблемы или перед принятием важного решения. Проект исследований относится к конкретному исследованию и задает способ сбора и обработки данных10. Для удобства рассмотрения выделяются следующие этапы исследований, которые чисто условно формируют магистральную последовательность действий. 1. Формулировка проблемы. Определяются цели исследования. 2. Проектирование исследований. Выбирается тип исследований: если недостаточно концептуальной информации, то выполняется поисковое11 исследование. Оно отличается гибкостью: в ходе проведения исследования может измениться его область или тематика. Если же проблема четко обрисована, то проводится описательное12 исследование или исследование причинности13. Исходя из типа определяется метод проведения исследования (от обзора литературных источников до проведения глубинного интервью с покупателями). Здесь же уточняются способы статистической обработки собранных данных. 3. Проектирование процедуры сбора данных. Если требуется опрос респондентов14, то нужно спроектировать анкету, определить, как будет проводиться анкетирование, задать способ регистрации полученных данных. Если проводится эксперимент, то его необходимо тщательно спланировать, уточнить, какие данные и каким образом будут собираться. 4. Проектирование выборки и сбор данных. Как правило, исследуются не все объекты, представляющие интерес (например, не все покупатели в магазине), а только их часть. Эта часть должна быть тщательно отобрана. Сам процесс сбора данных часто занимает достаточно большое время. Он нуждается в контроле, особенно если проводится силами привлекаемых со стороны интервьюеров. Ошибки в проведении исследований лучше обнаруживать и устранять как можно раньше, пока их еще можно исправить с малыми потерями. 5. Анализ и интерпретация данных. В этот этап входят: просмотр и редактирование (данные должны быть полными, правильными, собранными согласно инструкциям); кодирование (каждому ответу должно быть присвоено определенное обозначение); табуляция (данные представляются в табличном виде, группируются); статистическая обработка (ее вид должен быть задан до начала исследования). Важно отметить, что анализ с применением статистических методов обработки данных выполняется практически в самом конце исследования, и его результаты зависят от правильности выполненных ранее шагов. Данные следует собирать, только если уже известно, какие параметры исследуемого объекта требуется получить, и как будет обрабатываться собранная информация. 6. Подготовка отчета. Отчет пишется для заказчика исследования. В нем обобщаются и интерпретируются полученные результаты, делаются общие выводы и даются рекомендации. Большинство людей узнает об исследовании только из отчета, поэтому он должен составляться тщательно и точно. Маркетинговые исследования – сложный процесс, все его этапы взаимосвязаны. Часто в реальных исследованиях приходится возвращаться к более ранним этапам в свете вновь полученной информации. Например, если выяснилось, что для проведения сбора данных недостаточно денежных средств или времени, может оказаться необходимым скорректировать цели исследования. 1.7.5. Международная специфика В международном контексте подход дал Малхотра [Malhotra, N.K. (1992a), “Designing an international marketing research course: framework and content”, Journal of Teaching in International Business, Vol. 3 No. 3, pp. 1-27.]. В принципе это развитие общего определения. Видно, что особенности в окружении. Это окружение и само является объектом исследования, и влияет на процесс исследования. Роль маркетинга в экономическом развитии страны может меняться. Например, развивающиеся страны ориентированы на производство, а не на маркетинг (российские автомобили. Что еще?). Это идет традиционно от закрытого рынка, от планирования, от ориентацию на какие-то цели, например, титаны в вагонах, места дач, метро. Это осталось еще в сознании. В таких странах спрос превышает предложение и мало внимания уделяется удовлетворению потребности. (у нас до кризиса средняя цена купленной машины уже была 25 тыс. долл. А до этого покупали старые машины). Мало конкурирующих фирм в отраслях (авиапром – 1 предприятие). Надо смотреть: разнообразие [variety] продуктов, ценовая политика (дома), контроль правительства над СМИ, мнение общества относительно рекламы, эффективность системы распределения на массовом рынке, поведение потребителей (у нас только начинаются проблемы широкого выбора). В развитых странах можно задавать вопросы о разнообразии производителей и брендов. В некоторых странах это некорректно. Малхотра пишет, что в восточной Европе, но это меняется, у нас уже большое разнообразие. Но где-то бывает и дефицит [shortage] economy. В некоторых странах принято торговаться (Египет). Реклама на ТВ может контролироваться правительством, на правительственных каналах (у нас заработок – главное. Даже в газете комсомольская правда рекламировали фальшивую лотерею!!). Кое-тчо может быть табу. Но не у нас, (песни для взрослых, Pussy riot ['raɪət]). Правительство. Демократия, диктатура, монархия, социализм, коммунизм, или что-то переходное. Общественная политика, регулирующие органы, помощь или наказание [incentives and penalties], инвестиции в государственные предприятия. Часто ограничивается иностранная конкуренция (В старой России, теперь). Барьеры [tariff barriers]. У нас: поощряют производство, не поощряют ввоз готовых продуктов. Автомобили. Правительство обеспечивает инфраструтуру, управляет рынком. Само является предпринимателем. В развитых странах роль государства велика. Оно разрабатывает и реализует промышленную политику (Рассказать о своей стране). На тактическом уровен государство определяет tax structures, tariffs, product safety rules and regulations, promotion, and often imposes special rules and regulations on foreign multinationals and their marketing practices. (О высоких налогах в России). Государство много закупает. Потом перепродает (распределяет). Политика – концентрированное выражение экономики, согласно марксизму). Так что надо вместе рассматривать. Законодательство [legal environment encompassing, codes (civil, labor, land etc.), international law, antimonopoly, antibribery, and taxes]. Законы о продукте: качество, упаковка, гарантии, послепродажный сервис, патенты, торговые марки, авторское право. Законы о цене: фиксация цен, ценовая дискриминация (ужасно для иностранцев), и т.д. Дистрибуция: территориальное деление, типы каналов, дистрибуторство, оптовые соглашения [agreements]. Продвижение. Есть ограничения (сигареты, пиво). Экономическое окружение. Размер экономики (GNP), уровень, источники и распределение доходов, тенденции роста, структурные тенденции [sectoral trends]. У нас очень много добывающей промышленности. И еще больше торговли. В перестройку было 95% торговли. Смотрим на стадию экономического развития, стандартизацию рынков. Например, рынки типизируются работа, отдых делаются однородными [homogenized] блгодаря экономическому развитию и технологиям. Структурные различия: транспорт, связь. (о России). Телефонизация, почта. Информационная и технологическая среда. Информационные и коммуникационные системы, компьютеризация, использование электронного оборудования, энергии, наука и изобретательство. Есть развитые (США). Есть менее развитые. Иногда есть изобретения и наука, но мало влияет это на жизнь (Россия). Финляндия – высокое использование технологий. В России за 2011-12 произошло резкое изменение. Интернет, компьютерные технологии. До этого всё вручную, например, счета в Сбербанке. Социокультурные факторы. Грамотность [literacy], ценнгости, язык, религия, правила общения [communication patterns], семейные и общественные институты. Отношение к работе, ко времени, достижениям, богатству, риску, изменениям, нововведениям, к Западному миру – всё надо учитывать. Исследование на должно конфликтовать с культурой. Для малограмотных стран нужны другие шкалы. Языки и диалекты разнятся на малой территории. Пример. Бургерленд добился успеха в Саудовской Аравии. Многие фастфудные сети провалились. Поняли социокультурную среду. Арабы любят новое. В их жизни много ограничений, они хотят нового. Арабы любят детей, сделали места для семей. Свидания [dating] не приняты, но встречаются братья и сестры. Арабы не смотря на калории. Чем тяжелее, тем лучше. Каждое окружение уникально. Просто перенести метды в другую страну нельзя. 1.7.6. Порядок работы Начнем от задачи. 1. Кому это интересно. 2. В чем проблема организации. 3. Что надо найти. 4. Как. 5. Как провести исследование, 6. Как получить смысл. Значит, формулируем задачу и решаем ее. 2. Определение цели исследования 2.1. Деятельность организации на междурнародной арене 2.1.1. Процесс интернационализации From [Kotabe M., Helsen K. GLOBALMARKETING MANAGEMENT JOHNWILEY& SONS, INC., 2010, 751P.]. From [Kotabe M., Helsen K. GLOBALMARKETING MANAGEMENT JOHNWILEY& SONS, INC., 2010, 751P.]. From [Kotabe M., Helsen K. Global Marketing Management John Wiley& Sons, Inc., 2010, 751P.]. Сейчас слышно global markets, global competition, global technology, and global competitiveness. Раньше, в 1980-е, было слышно international or multinational instead of global. Что произошло? Насыщение национальный рынков Saturation of Domestic Markets. В развитых странах происходит насыщение местного рынка. Компании ищут рынки вовне. В развивающихся странах к тому же растет население. У нас китайские, корейские, шведские, немецкие и др. компании. Из развивающихся стран приходят в развитые (Китай в Кентербери). Из отдаленных регионов (Австралия и Новая Зеландия) строят, например, сети кофеен (D^ome Coffees Australia) Развивающиеся рынки Emerging Markets. В 20 веке большие экономики и большие торговые партнеры были расположены в основном ы торговых регионах мира: Северная Америка, Западная Европа, Юго-Восточная Азия. Там делали 80% ВНП. (GDP). И было там 20% населения. Потом появился ten Big Emerging Markets (BEMs)—the Chinese Economic Area, India, Commonwealth of Independent States (Russia, Central Asia, and Caucasus states), South Korea, Mexico, Brazil, Argentina, South Africa, Central European countries, Turkey, and the Association of Southeast Asian Nations (Indonesia, Brunei, Malaysia, Singapore, Thailand, the Philippines, and Vietnam). В последнее время говорят о БРИК или БРИКС. Оттуда появляются конкуренты. Но там и возможности для торговли. Глобальная конкуренция Global Competition. Сейчас что-то совершенно новое появляется в конкуренции. Были мировые лидеры автопроизводители General Motors, Ford, and Chrysler (явно пишет американец!). Сейчас Тойота, Honda, BMW, Renault, and Hyundai, among others, stand out as competitive nameplates in the global automobile market. В 2008 доля рынка Тойоты в мире стала больше, чем дженерал моторс. И в США очень большая доля. Персональные компьютеры. В 10990-е была белая, желтая и красная сборка. Персональный компьютер был = IBM. Сейчас Dell and Hewlett-Packard (HP) from the United States, Sony and Toshiba from Japan, Samsung from Korea, Acer from Taiwan. Indeed, Lenovo, a personal computer company from China, acquired the IBM PC division in 2005. This challenges the world top players, Dell and HP/Compaq, respectively. Nike is a U.S. company with a truly all-American shoe brand, but all its shoes are made in foreign countries and exported to many countries. Food: McDonalds, 7 eleven (Japan). Media: MTV, targeting teenage audiences, has 35 channels worldwide, 15 of them in Europe, produces a large part of its channel contents locally. CNN has 22 different versions. The video game industry is truly global from day one; Nintendo’s Wii, Sony’s Playstation 3, and Microsoft’s Xbox now vie for customers in the Triad regions simultaneously. Глобальная кооперация Global Cooperation. Конкуренция приводит к кооперации. IBM and Japan’s Fujitsu used to be archrivals. Beginning in 1982, they battled each other for fifteen years in such areas as software copyright. But in October 2001, they developed a comprehensive tie-up involving the joint development of software and the mutual use of computer technology. IBM would share its PC server technology with Fujitsu and the Japanese company would supply routers to IBM. Japan’s Sony, Toshiba, and U.S. computer maker IBM are jointly developing advanced semiconductor processing technologies for next-generation chips. As part of the project, IBM transfers its latest technologies to Sony and Toshiba. In 1999 Renault took a 36.8 percent stake in Japanese carmaker Nissan Motor Corp. The two companies began producing cars on joint platforms in 2005. Информацию довольно сложно добыть. Интернет революция Internet Revolution. Уже давно больше миллиарда пользователей Интернет, с 2008. с 2000 выросло в 3 раза (см. динамику мировую и по странам). Очень быстро растет оборот электронной коммерции. Триллионы долларов. B2B растет быстрее, чем B2C. Это объясняется не толкьо разхвитием техники, но и растущей свободой движения товаров, услуг, капитала, технологий, людей. Конечно, уровень зависит от страны: дохода, населения, доступности кредита, инфраструктуры, налогов, вложений в R&D, языка и проч. Who could have anticipated the expansion of today’s e-commerce companies, including Amazon, eBay, and Yahoo in the United States; QXL Ricardo and Kelkoo in Europe; Rakuten and 7dream in Japan, and Baidu in China? Открылась возможность продавать прямо по всему миру. Конечно, не так intimately, чем лично, но более целенаправленно по демографическим и психографическим характеристикам. Можно CRM строить, двунаправленное общение. Сейчас очень много нововведений. Например. Догадываются, кто сидит за компом. Что спрашивал – долго дают информацию. Пользователи знают о мировых брендах, это увеличивает их продажи. Можно прямо по телефону или по почте заказать комп у Dell. Автопроизводители, билеты и многое другое. Но надо знать языки. Хотя есть переводчики. 2.1.2. Этапы деятельности Деятельность организации на международном рынке происходит непрерывно. Постоянно собираются данные, разрабатываются и принимаются решения, они реализуются. Однако для удобства изучения можно ввести следующую идеализированную последовательность. • Определение целей организации. • Разработка стратегии достижений этих целей. • Разработка и реализация тактических решений. • Оперативное управление деятельностью. На каждом этапе используется определенная информация. Особенно важно иметь адекватную задаче информацию при работе на международных рынках. Activity in the company may be done in the following ways [Котлер, маркетинг-менеджмент]: Текущая работа Current work. “The work is too complex to plan”. Indeed, almost any plan is being changed right from the beginning. Engineers are inventing, marketing specialists working out strategies… It can be done only inside a company and tends to slow down. In Russia about one half of the companies, especially small, are working in such a way. Запланированные работы (проекты) Planned work (projects). “The work is too complex to work without a plan”. This is done when work is performed within frames of agreement. One side pays definite sum money to get a definite result in definite time. It is also good to work in such a way within one single company. Project is a complex of activities aimed to obtain definite results during given time within given financing. As this way is most formalized and spread, it is reasonable to rely on this method. Пробные начинания. Sample initiatives. This may be done in technical projects (for example, alcohol as car fuel was suggested and work started in 80-ies, but then the idea failed. Now it is another attempt, in 2007 Sweden declared not using gasoline in 2010, but now they use 10% or 5% alcohol in gasoline and it is not popular. The same way is often used to get to foreign market. If the initiative fails, it is temporarily stopped. Task Find sample initiatives of foreign companies in Russia. (Joint production of computers. IKEA, ?, ?, ?). Вообще: топливные элементы, спирт вместо бензина, лазерный термояд. It is obvious that such a way is used by large companies or governments. To some extent it is similar to venture business. 2.2. Бизнес-план Business plans are most usually made for projects. It is convenient to demonstrate what is necessary to know when making such a plan. Now the well-known abroad idea of business plan spread in Russia. Usual structure of business plan is as follows [Клоков И. В. Бизнес-план на компьютере: быстро и просто – СПб.: Питер, 2007. – 176 с]: 1. Резюме. Приводится краткое описание всего бизнес-плана. 2. Цели и задачи. 3. Продукт (услуга). 4. Анализ рыночной ситуации. 5. Производственный план. 6. Маркетинговый план. 7. Организационный план. 8. Финансовый план (бюджет). 2.2.1. Структура целей Работа организации должна быть подчинена определенной цели. На высшем уровне рассматривается следующий треугольник целей (рис. Рис. 3) [Завгородняя, Ямпольская]. Рис. 3. Треугольник целей организации Для коммерческих организаций приоритет обычно находится в целях предпринимательской деятельности. Предпринимательство согласно ГК России: деятельность по систематическому получению прибыли на собственный риск. Прибыль. Упрощенно, это средства, которые остаются после получения всех доходов и осуществления всех расходов. Они используются в качестве дивидендов или для развития организации. Рентабельность. Сколько денежных единиц получено от каждой вложенной единицы. Немного по-другому. Прибыль в 1 млн евро можно получить и вложив 1 млн, и вложив 100 млн. Первое лучше. Но сравнивать можно только сопоставимые проекты. Можно получить 100 евро, вложив 1 или получить 1 млн, вложив 1 млн. Разная рентабельность, но и разные масштабы. Варианты несравнимы. Имеются и другие показатели (какие?) Корпоративные цели. Миссия часто рассматривается как цель существования организации. Если это не извлечение прибыли, то что? Обеспечение своих работников. Развитие самой фирмы. Часто говорят высокие слова об обеспечении какой-либо потребности (Wolkswagen). Более правильным следует считать определение миссии как области, в которой работает фирма. Для формулировки миссии, в соответствии с советами П.Друкера [Друкер П.Ф. Классические работы по менеджменту. М.: Вильямс, 2011, 704 с.] надо ответить на следующие вопросы: • Что есть Ваш бизнес? • Кто Ваш клиент? • Что ценно для ваших потребителей? • Чем будет Ваш бизнес? • Чем должен быть Ваш бизнес? Совокупность ответов на эти вопросы позволяет более глубоко понять миссию, получить ориентиры в практической деятельности, не распылять силы и средства на ненужную работу. В миссии должно быть также определено: 1. В какой отрасли работаем 2. Каковы направления деятельности и виды выпускаемого продукта 3. Каковы компетенции работников 4. Целевые сегменты 5. Географические границы деятельности. 6. Вертикальная интеграция – количество звеньев цепочки создания ценности. Лукойл в своей рекламе утверждал, что занимается всем, от научных исследований в области геологии до розничной торговли топливом. Другая крайность – так называемые "оболочечные" организации, лишь координирующие заключение и исполнение договоров. Считая миссией либо высокую идею, либо перечень упомянутых признаков, здесь рассматривается обеспечение этой миссии. Прежде всего, это захват, удержание и расширение выбранных целевых сегментов, пусть даже с низкой прибылью. Функциональные цели обычно формируются на основе двух вышеперечисленных типов целей. Задачи – это цели более низкого уровня, достижение которых необходимо для достижений главных целей. Из рисунка видно также, что для формирования этих целей требуется изучение как внутренней, так и внешней среды организации. 2.2.2. Описание продукта Для того, чтобы проект был успешным, необходимо ответить на ряд вопросов. Причем это не только уверенность в том, что делают работники компании. Это нужно и для поиска финансирования, и для разработки рекламных обращений. Какие потребности должен удовлетворить продукт (услуга)? Пример: «мы печатаем фотографии, но продаем воспоминания». Это будет и содержанием рекламных обращений. Именно четкое описание этого аспекта помогает раскрыть бизнес-возможности. Что особенного в товаре (услуге) и почему потребители будут отдавать ему предпочтение? «Дифференцируйся или умирай» (Друкер). Надо рассмотреть все аспекты. И технические характеристики продукта и дизайн, и особенности поставки, и упаковку, и сервис. Не обязательно это должно быть что-то уникальное. В 2010 году более успешными были традиционные проекты, например, сетевые розничные магазины. Но почему потребители пойдут именно в них? Они могут быть ближе к дому, дешевле, с более широким ассортиментом. Один наш студент подавал собственноручно написанную книгу по саморазвитию. Вначале не хватало именно дифференциации, какого-то особого образа. Каков жизненный цикл товара, как скоро он устареет? Каковы перспективы совершенствования продукта? Какими патентами защищен или может быть защищен продукт? Какова предполагаема цена продукта? Какова себестоимость продукта? 2.2.3. Анализ рынка Прибыль можно получить, только сбыв товар. Более надежны фирмы, выпускающие традиционный товар и имеющие налаженный сбыт, чем новые, предлагающие пусть и суперновинки. Главная задача здесь – определение спроса. 1. Вначале определяется целевой сегмент, то есть приводится описание потенциальных покупателей и потребителей. (Большинство покупателей электробритв – женщины). Обычно это пол, возраст, образование, профессия, доход и другие социально-демографические и экономические характеристики. 2. Далее нужно спрогнозировать, кто, почему и сколько будет покупать вашей продукции в ближайшем и более отдаленном будущем. Такой расчет производят в три этапа. 2.1. Первый этап — оценка общей потенциальной емкости рынка. Это значит, что нужно рассчитать суммарную стоимость товаров, которые покупатели данного региона могут купить за расчетный период. Эту величину определяет множество факторов, в том числе следующие: • социальные условия; • национальная специфика; • культурные традиции; • климатические и географические условия. Однако главное, конечно, — это экономические факторы: • уровень доходов; • структура расходов; • наличие аналогичных товаров; • темпы инфляции. 2.2. На втором этапе рассчитывается та доля рынка, которую вы намереваетесь заполнить своим продуктом. Таким образом вы определите максимальную сумму реализации, на которую можете претендовать. Чтобы правильно рассчитать долю рынка, следует учитывать возможность эволюции конкурентов, но поскольку спрогнозировать ее часто нет никакой возможности, все допуски следует делать в сторону уменьшения оценочных данных. 2.3. Третий этап — на базе максимальной суммы реализации произвести реальную оценку уровня продаж вашей продукции в конкретных условиях деятельности при определенных затратах на рекламу и определенном уровне цен. Важно также определить динамику реальных продаж, то есть как они будут изменяться от месяца к месяцу. Такой прогноз чаще всего делают методом экспертной оценки, то есть на основании собственного опыта работы в этой сфере или с привлечением консультантов. В ходе расчета можно предусмотреть несколько вариантов, различающихся по уровню цен на вашу продукцию и по затратам на ее продвижение. Косвенным результатом этой работы может стать изучение конкурентов: цен на аналогичную продукцию, различий в качественных характеристиках, ценовом диапазоне и условиях, на которых распространяется продукт. Такая информация тоже может быть отражена в бизнес-плане для потенциальных инвесторов. Сведения о конкурентах лучше всего подавать в следующем виде. 1. Информация о крупнейшем производителе аналогичных товаров. 2. Характеристики товара и способов его продвижения, отзывы покупателей. 3. Ценовая политика конкурентов. Сильные и слабые стороны конкурентов. Основываясь на замеченных ошибках соперников, можно создать собственную успешную бизнес-стратегию. [Клоков] приводит следующие данные о причинах неудач реализации бизнес-планов: 1. Ошибочное определение объемов спроса — 45%. 2. Завышенная цена — 18 %. 3. Нерешенные производственные проблемы — 12%. 4. Дефекты товара — 9 %. 5. Ответные действия конкурентов — 7 %. 6. Недостаточная реклама и усилия по продвижению — 5 %. 7. Неверно выбранное время выхода на рынок — 4%. Итог: производственные проблемы: 21%, рыночные – 79%. Такова роль информации в успехе или провале предприятия. 2.2.4. Производственный план Производственный план содержит характеристики расположения производственных площадей, оборудования и описания процессов, сопровождающих производство. Кроме того, здесь оценивают необходимость привлечения субподрядчиков и условия, на которых их нанимают. Важно не упускать из виду вопросы контроля качества производимой продукции и регулирования соответствующих производственных процессов. Важный момент — регуляция основных составляющих стоимости, таких как, например, заработная плата работников или стоимость используемых материалов. Важнейшие моменты этого раздела: возможности по увеличению или уменьшению производства, взаимодействие с поставщиками и сроки поставок. Для международной организации важно определить, где будет производится продукт. От этого зависит и взаимодействие с партнерами, заработная плата, стоимость материалов. Task Evaluate the “screw driver” production of cars in Russia, its pluses and minuses 2.2.5. План маркетинга [Клоков] Маркетинговый план должен определить, почему вашу продукцию будут покупать, что для этого нужно сделать и чего избежать. Также требуется максимально точно рассчитать объем продаж. Но это – функции многих переменных. Маркетинговый план состоит из следующих компонентов. 1. Ценообразование. Ценообразование подчиняется политике предприятия и соответствует возможностям рынка. Критическим моментом при расчете цен является так называемая точка безубыточности. Что надо знать при разных методах? Вот их описание. Метод издержки плюс прибыль возможен только при отсутствии конкурентной среды. В противном случае конкурент, снижая собственные издержки, легко вытеснит ваш товар с рынка. При расчете издержек необходимо помнить, что их можно разделить на постоянные и переменные. Постоянные не зависят от объемов производства. Это арендная плата, административные и накладные расходы. Переменные издержки непосредственно связаны с производством; это расходы на сырье, материалы, упаковку, транспорт и заработную плату. По мере роста объемов производства они также растут. В точке безубыточности суммарный объем издержек равен суммарному объему реализации. Если объем реализации превышает издержки, производство начинает приносить прибыль. Метод следования за конкурентом. Может применяться небольшим предприятием при наличии на рынке крупного конкурента, который ведет основную ценовую политику. Если у конкурента большие обороты, очевидно, он провел серьезное исследование рынка и определил оптимальную цену. В принципе, ваше небольшое предприятие может, пристроившись «в хвост», завоевать свою долю рынка за счет каких-нибудь дополнительных преимуществ, например послепродажного сопровождения. Серьезным недостатком метода является отсутствие контроля над ситуацией с вашей стороны. Лидер может подготовить основательную перестройку своей ценовой политики, к которой вы будете не готовы и потеряете свои позиции на рынке. Затратно-маркетинговый метод. Это самый сложный метод ценообразования; он сочетает анализ затратного механизма производства и реализации с маркетинговой тактикой. Это творческий метод, который трудно формализовать, но который сулит в случае правильного применения большие успехи. Реальное ценообразование учитывает также и политику скидок, и механизм корректировки цен с учетом жизненного цикла товара. Все эти методы преследуют одну цель — привлечение клиента. 2. Схема распространения товара от производства до конечного потребителя Обсудить варианты для международного маркетинга. Что для этого надо знать. 3. Продвижение продукции. (обсудить, что надо знать) 3.1. Реклама 3.2. Стимулирование сбыта 3.3. Личные продажи 3.4. PR Какие еще составляющие маркетинга бывают? Для услуг; 3.5. Physical evidence 3.6. Personnel. 4. Послепродажное сопровождение. 2.2.6. Организационный план Организационный план описывает распределение обязанностей, квалификацию специалистов и управленцев. Особенно важно охарактеризовать роль каждого члена руководящей группы, его полномочия и ответственность. Оптимальным решением будет привести здесь штатное расписание. Неплохо предусмотреть также методы стимулирования труда персонала, в том числе управленческого. Что для этого требуется узнать? 2.2.7. Финансовый план Финансовый план раскрывает как возможности получения прибыли, так и предполагаемую степень риска. Это заключительный раздел бизнес-плана, здесь указывают важнейшие данные, влияющие на финансовую состоятельность проекта. В том числе приводят все основные исходные данные, которые были использованы для расчетов финансовых прогнозов, инвестиционных затрат, налогов и сборов. Приводят также данные о потребности в привлеченном капитале и планируют мероприятия по обеспечению максимальной доходности инвестиционного капитала, тем самым делая бизнес привлекательным с точки зрения инвестора. Кроме того, нужно отобразить гибкость системы финансирования и возможные варианты выхода инвесторов из бизнеса. Форма: обычные финансовые документы, но с плановыми числами. 2.3. Потребность в информации Зависит от опыта фирмы, степень проникновения на международные рынки. На начальной стадии выхода на международный рынок нужно знать возможности проникновения [assess opportunities], риски в разных странах, чтобы спланировать international market entry and mode of operation. Если решения по выходу на рынок уже приняты, внимание переходит на решения по маркетинг-миксу: как разработать новый продукт, как его протестировать, (авто для России) как рекламировать, надо изучить медиа и ценовую эластичность (отношение к ценам в США, Финляндии, России). Потом, когда будет опыт, надо информационные системы строить для улучшения распределения ресурсов по рынкам и странам, чтобы использовать возможную синергию от лучшей интеграции и координации международных стратегий. 2.3.1. Phase 1 – Информация для выхода на международный рынок Нужны данные двух уровней. 1. Об общей ситуации в бизнес-окружении в стране или регионе. Политика, финансовая стабильность, regulatory environment, размер рынка и его рост, инфраструктуру. Эта информация считается известной для своего домашнего рынка, так как менеджеры знают, контачат с местным бизнес-окружением. Для других стран это крайне важная информация для определения возможностей и способа выхода на рынок 2. Информация о специфическом рынке определенного продукта или услуги, на который компания собирается выйти. Это рыночный потенциал, уровень роста рынка, структура рынка, источники прямой и непрямой (борьба за семейные бюджеты: производителей мебели беспокоят расходы на туризм; говорить о субститутах) конкуренции, общая конкурентная ситуация. Надо определить метод оценки спроса, привлекательности рынка. Методы от качественной оценки, ранжирования, до построения имитационных моделей. Зависит от необходимой точности, имеющихся денег и времени 2.3.2. Phase 2 – Информация для планирования рынка Обычно о других рынках знают мало, поэтому возникает предварительная стадия сбора информации. Она нужна, чтобы определить проект исследования и требования к исследованию. Часто полезны качественные исследования, чтобы дать информацию для обзорных исследований [market survey]. Это помогает определить что изучать на следующих стадиях исследования: какие свойства товара, отношение [attitudes], поведение покупателей. Предварительные исследования также используются для сбора информации по рынку продукта, товарам-комплементам и субститутам, имеющимся исследования по [attitudes], конкурентам и т.д.. Исследования на этой стадии в основном касаются маркетинг-микса. Несколько глубоко надо изменять имеющиеся продуктовые и продвигательные стратегии со своего домашнего рынка. Продукты могут преназначаться для разных сегментов, выгоды и предпочтения потребителей различаются от страны к стране (эскимосы покупали холодильники, чтобы обогреть иглу). Для автомобилей разная важность у экономичности, [gas mileage], road handling, безопасности различаются. В России популярны внедорожники (как меня везли). Для пищи тоже есть различия: вкусы, предпочтения, сценарии потребления очень различны. Надо определить, как модификация продукта и его позиционирования поможет увеличить продажи. Либо расширив число клиентов, либо поглубже охватив рынок. Темы продвижения, текст рекламы, упаковка также нуждаются в проверке их эффективности для местного рынка. Тут опять возникают вопросы грамотности, [literacy], культурных норм относительно пола и юмора, эстетических вкусов, ассоциаций цветов (отношение зеков к красному цвету, цвет траура), интерпретация символов влияют на интерпретацию и реакцию на визуальные стимулы, emotional appeals and promotional arguments. Чувствительность к цене тоже надо исследовать, она разная в разных странах, зависит от доходов, сегмента, субститутов, восприятия цены и проч. и проч. Исследование каналов. Заинтересованность в сервисе и доставке, удобстве, лояльности к бренду или магазину, время на покупку, предпочтения различных видов дистрибуции различаются. В России сейчас только начали доверять Интернету. Яндекс-деньги. Веб-деньги, электронные магазины. Возможности новых продуктов и услуг. Исследование от мониторинга трендов в окружающей среде и технологии (сканеры штрих-кодов [bar codes] внедрились. Тут может быть обман в рознице?), стиля жизни (дачи были источником еды, сажали картошку в полях, теперь больше отдых), удовлетворенности потребителей (я рад тому, что нет очередей). Делают глубинное интервью, брейнсторминг, фокус-группы. Концепции новых продуктов тоже надо проверять многошаговым тестовым маркетингом (хотя его у нас мало) 2.3.3. Phase 3 – Информация для глобальной рационализации Новые требования к информации. И более эффективное использвоаине собранных данных. Вторичные данные помогут отслеживать изменения в окружении фирмы, степень взаимосвязи и интеграции рынков. Была страна стабильная, привлекающая иностранных инвесторов. Стала нестабильная, против иностранцев. Рост может ускориться или замедлиться. Инфляция. Надо следить за торговыми потоками, коммуникационными связями. Увидим, как меняются границы рынков. Может понадобиться изменгить стратегию. Тут надо собирать информацию глобально. Надо расматировать пространственную [spatial] конфигурацию of its assets and resources чтобы построить сильную конкурентную позицию со стратегической гибкостью в условиях изменяющейся рыночной динамики и ресурсных условий. Надо также развивать средства передачи информации, опыта и идей с одного рынка на другой, чтобы использовать различный опыт [diversity of experience] при разработка стратегии на международном рынке. Надо также объединить данные, собранные по странам, о потребительских вкусах и предпочтениях. Будут найдены общие черты, возщникающие тренды. Это поможет при разработке глобальной стратегии: когда и где вводить новые глобальные бренды, и как координировать стратегии на разных рынках. Надо найти баланс важности глобальных или местных брендов (учитывающих национальные особенности). В результате получится структура брендов, чтобы фирма получила свое место и узнаваемость [identity] на международных рынках. Надо и собственную деятельность на разных рынках мониторить. И сравнивать со вторичными данными о внешней среде, о рыночных условиях. Получается, что надо обобщать и сравнивать данные из различных источников. Надо строить информационную систему. В результате отследиваем деятельность и смори, куда направить ресурсы. В инфосистеме информация собирается из отдаленных углов, изучается и анализируется при разных условиях деятельности, чтобы быть полезной дял маркетинговых решений. Технические вопросы сейчас решены, внимание уделяется полезности и применимости информации. Есть еще и проблема перегрузки [overload ]информацией. Тут надо собирать систематически. Конечно, бывает дорого. Но надо. Иначе будет плохо. Какой рынок какой страны наиболее привлевателен для выхода, насколько надо адаптировать позиционирование и тактику маркетнга на локальном рынке, что происходит, как передать хорошие идеи и практики на друге рынки. 2.4. Информация для стратегических решений Глобальные стратегические решения принимаются на корпоративном и региональном уровне. Это долгосрочные направления и цели фирмы, например, выход на международный рынок, развитие рынка, развитие фирмы. Рынки каких стран и каких продуктов завоевывать, какой портфель стран даст хороший баланс для будущего роста. Как сегментировать рынки: по странам или по группам стран, через страны. Еще стратегии брендов, стратеги позиционирования на каждом целевом сегменте. Надо знать распределение ресурсов в странах. Получается, что это не только маркетинговые функции, но и финансовый и производственный менеджмент, бюджетирование, бухучет, планирование производства. Хотя бы стоимость выхода на рынок разных стран различна. Ресурсы ограничены. Но рынки взаимосвязаны. Решения о выходе получаются сложными. Надо координировать продуктовые линии и географию. Надо учитывать уже то, что сделано в других странах и на других рынках. Нужна разная информация для разных уровней решений. Для всей корпорации, для Strategic Business Units, для продуктовой линии, географических единиц, рыночных сегментов. О связях между элементами, людьми, идеями. В таблице примеры. Видно, что пока это все просто набор. Тут есть типы индикаторов (факторы) и сами индикаторы (переменные). Это примерно (как во многих учебниках пишут), надо для каждой задачи искать, что собирать. Ясно, что полезно, не неясно, как использовать. На уровне страны: возможности и риски. Режим работы [modes of operation] и маркетинговые стратегии требуют различных затрат (электричество и проч). На уровне рынка продуктов требуются свои показатели, например, рентабельность Return Of Investments. Для новых рынков можно взять косвенные показатели (в таблице). Включаем в БД и обновляем. Есть проблема сравнимости данных. 2.5. Информация для тактических решений Сегментация прежде всего. Какую рекламу давать, насколько стандартной может быть тема для имеющегося бренда (ma girl for Mars), как менять стандартную рекламную кампанию (пример продажи арбузов, билетов на автобус). Решения по продукту: изучение выгод и свойств [benefit and attribute], проверка концепции, тестовый маркетинг во всех странах! Адаптировать ли продукт (двойная оцинковка и высокая подвеска машин для России). Новые продукты могут быть для разных стран или для группы стран. Каковы приемы рекламы в разных странах. И т.д. Про это уже говорилось. Изучается маркетинг микс. Это похоже на исследование домашнего рынка, методы почти те же. (табл.). Побольше качественных исследований, особенно на начальных стадиях. Посложнее, так как в разных условиях. Все, что точно известно, на работает. Надо сомневаться во всем. 2.6. Информационная единица – объект исследования Знаем, каковы вопросы исследований и какую информацию собрать. Теперь надо определить единицу исследования. Географически: город, район, страна, группа стран, регион, мир. О ком данные? О типах и группах потребителей (организаций), обо всех потребителях (организациях) в географической единице, или просто обо всей единице (продажи, потребление, patterns of expenditure (расходы) and pricing patterns. Это зависит от проблемы. 2.6.1. Мир Это самая большая единица. Но вторичных данных мало. Немного о глобальные индустриях: аэрокосмос, фармацевтика, данные об объема операция по глобальной торговле. Сейчас растет сбор первичной информации об отрасли промышленности, специфических сегментах, типах потребителей. 2.6.2. Регион или группа стран Регион и группа стран. Есть вторичные данные, например. О евросоюзе, об ASEAN (Association of South East Asian Nations). Но обычно это разбивается по странам. 2.6.3. Страна Это чаще всего. Вторичные данные из страны. Gross National Product (GNP), население, потребление энергии, стали, ценовые тенденции, расходы на потребление [consumption expenditures]. Много исследований по рынкам и индустриям. Первичные исследования – специализированными фирмами. 2.6.4. Города Иногда надо и так. Если например, особый сегмент, например молодые горожане-профессионалы. Или рестораны и универмаги (хотя в Финляндии они располагаются везде). В развивающихся странах в городах более привлекательно, сельские жители небогаты и географически разбросаны. В России вообще вымирают деревни. Часто уровни бывают разные. Например, выбрали группу стран и там изучили пожилых горожан. 3. Оценка глобальных маркетинговых возможностей This task • первая задача is the first in the process of international market research; • важнейшая is of vital importance • нужна для начала практических работ is necessary to start practical work; • до некоторой степени предпринимательская to some extent it is entrepreneurial task. Thus, it is considered separately from other tasks. [Conducting Market Research for International Business / S. T. Cavusgil, G. Knight, J. Riesenberger, Attila Yaprak. NY, Business Expert Press, 2009, 129P]. Хороший выбор – залог успеха. Что и когда предлагать. Надо знать возможности. Особенно для международных рынков, где неопределенность и риск. Итак, надо найти бизнес возможности за границей. Возможность это благоприятное сочетание обстоятельств, расположения, времени или партнеров на иностранном рынке. Компании все время изучают возможности продать что-то, открыть завод, обеспечить доставку [procure goods] или войти в сотрудничество. Результат бывает обычно гораздо лучше, чем на местном рынке. Вот хотя бы только экспорт и маркетинг за границей (остальное похоже). 1. Global Market Opportunity Assessment (GMOA) это процесс анализа готовности к интернализации, определения соответствия продукта или услуги для иностранного рынка, оценка стран для проникновения: рыночного потенциала и потенциала продаж, выбор партнеров. Вот активности (табл). Ta+ble … Global Market Opportunity Assessment: Key Activities ACTIVITY RATIONALE TYPICAL TASKS 1. Анализ готовности организации к интернационализации Provides an objective assessment of the firm’s preparedness to undertake an international business activity. List the firm’s strengths and weaknesses, as well as recommendations for addressing resource deficiencies and other shortcomings that can hinder achieving company goals abroad. 2. Проверка готовности продукта к выходу на международный рынок Provides a systematic assessment of the suitability of the firm’s products and services for international customers. Evaluate the degree of fit between the given product or service and customer needs and characteristics in the target market. 3. Изучение стран для выявления целевых рынков Reduces the number of countries that warrant in-depth investigation to a manageable few. This helps ensure that organizational resources are used efficiently and lessens the complexity of the assessment task. Identify five or six high-potential countries that are most promising for the firm. Consider market size, market-growth rate, market intensity (i.e., buying power of the residents in terms of income level), consumption capacity (i.e., size and growth rate of the country’s middle class), infrastructure appropriate for doing business, degree of economic freedom and political risk, and other appropriate variables. 4. Определение потенциала рынка по отрасли Allows the manager to gain an understanding of the total potential sales of a product or service in a given foreign market. Estimate the most likely share of industry sales in each target country. The unit of analysis is the firm’s specific industry. Accordingly, investigate and evaluate industry-level barriers to market entry. Develop a 3- to 5-year forecast of industry sales in the market. Identify market-entry barriers. Examine key variables such as market size, market growth rate, and trends in the industry. Assess the nature of competitors in the market. Investigate the degree of industry-specific protectionism. Analyze standards and regulations that apply to the firm’s products. Evaluate the availability and sophistication of distribution infrastructure in the marketplace, appropriate for the firm’s industry. 5. Выбор партнеров Collaborating with suitable partners helps the firm achieve its goals in foreign markets. This stage helps ensure that the manager identifies and decides on the most appropriate partners. Prepare a “wish list” of ideal partner qualifications, such as the value-adding activities required of foreign business partners, desirable attributes in foreign business partners, and the nature of activities that the partners will perform. 6. Оценка потенциала продаж компании Allows the manager to develop a reliable forecast of the most likely share of sales that the firm can achieve during a given period in the particular market of interest. Develop a 3- to 5-year forecast of company sales in the target market. The unit of analysis is the specific market(s) that the firm is targeting. Acquire an understanding of factors that influence company sales potential and estimate the ability of the firm to sell its products in the market. Examine the capabilities of partners, available distribution channels, the level of competition, appropriate pricing schemes, and the risk tolerance of upper management for foreign market entry. Это, конечно, framework. Надо найти бизнес-идею, бизнес-модель и бизнес-планThe kernel of this is to find business opportunity to start international business? To develop business idea, business model and business plan. Бизнес-идея часть формируется неформализуемыми методами. US scientists think this informal part is most important. One should see opportunities. Here one may remember an experiment with losers. On one of so-called losers could find opportunity: they were asked to count pictures in newspaper. But among other titles it was: “Go to the experimentator and he/she gives you 200 pounds. No one found this text. Now activities in details. 3.1. Определение готовности организации С этого надо начать. Надо формально оценить. И для опытных, и для неопытных фирм. Анализ похож на СВОТ. В фирме смотрим мотивацию (цели [goals and objectives], ресурсы (какие есть, что надо развить – это главное), компетенции [skills and capabilities] для успешного иностранного бизнеса. Не хватает ресурсов – надо их обеспечить. Во внешней среде тоже требуется, но отдельно: нужды потребителей, природу конкурирующих продуктов, риски на предполагаемом рынке. Вопросы. • Что фирма надеется получить от интернационального проекта [venture]. Может быть, увеличить продажи и прибыли. Может быть, сократить стоимость, перенеся производство на новое место. • Согласуется ли проект выхода с другими целями фирмы, сейчас и в будущем. С миссией и стратегическими планами. Все возможности, которые появляются, на охватить. Надо выбрать такие, чтобы наилучшим способом использовать ресурсу. • Какие будут требования к ресурсам, если начнем проект? Время руководителей, персонала, производственные мощности [capacity], маркетинговые мощности. Должны быть достаточными. • В чем источник конкурентных преимуществ фирмы. Делаем что-то лучше. Тут обычно нужно R&D, [sourcing] подбор, выбор источников (финансирования, поставок и т. п.), конкурентное производство, skillful маркетинг, эффективная дистрибуция или другая деятельность в цепочке создания ценности. И как эти преимущества можно использовать. Это надо делать все время. Готовность – переменная величина. Есть инструменты для самоаудита, например, экспертная система CORE™ (Company Readiness to Export). Задает детальные вопросы, надо отвечать. Размещена globalEDGE™ (www.globalEDGE.msu.edu). 3.2. Оценка пригодности продукта к продажам за рубежом Далее надо понять степень, [degree], с которой товары и услуги фирмы соответствуют международным рынкам. Такие продукты обычно должны иметь одну или несколько из следующих характеристик. Хорошо продавиться на домашнем рынке. Удовлетворять универсальную потребность. Например, везде заботятся о здоровье, так что везде нужны personal care products. Удовлетворять (тут слово addressed; у нас направлено на) потребность, которая не удовлетворяется на иностранном рынке. Это обычно в развивающихся рынках и развивающихся экономиках. Например, в Африке мало чистой питьевой воды. В России примерно то же. Водопроводная вода недостаточно чистая, труды старые, текут. Много ржавчины и хлора. Но у нас об этом не говорили, пили воду, без фильтров. Денег на чистую воду жалко. Тем более в бутылке. Фильтр. Тут есть резерв. Хотя уже продаются разными фирмами. Address новую развивающуюся потребность за границей. Цунами в Индонеии, (мб в Японии) = потребность в портативных домах. Развитие мобльников = потребность в услугах через мобильники. Еще что надо знать. Кто инициирует покупку (B2B, В2С), кто ее реально использует, почему они обычно покупают и где они обычно покупают. Какие факторы (экономические, культурные, географические, и другие) ограничивают продажи. Тут проще всего спросить потенциальных посредников [intermediaries] и потенциале продаж. Полезно узнать, как похожий товар продавался и какова была динамика. Сколько такого продукта делают на месте, сколько импортируется и сколько экспортируется. Чтобы ознакомится с использованием и популярностью продукта, полезно посетить выставки, industry trade show на рынке, опросить дистрибуторов и покупателей. Trade shows часть охватывают целые регионы, например, Европу. Можно быстро получить информацию по многим рынкам Посмотреть, какие выставки, выбрать одну и посмотреть, что там и как. 3.3. Просмотр стран для определения потенциальных рынков Есть более 200 стран. Все не охватить [target]. Надо ограничиться более перспективными. Экспортирующие фирмы предпочитают страны с низкими торговыми барьерами (Россия хочет, чтобы производили внутри), хорошими квалифицированными посредниками (с этим сложно) и надежной маркетинговой инфраструктурой (мешает коррупция). Для тех, кто outsource value-chain деятельность, надо выбрать страны, где хорошие поставки [supplier countries]. Если фирма хочет заняться прямыми инвестициями за границу, надо искать низкорисковые страны, где ожидается высокая прибыльность и долгосрочный рост. Коммент: низкий риск – не бывает высокой прибыльности. В России стало стабильнее. Было всё краткосрочно. Перед кризисом 2008 стали планировать на 3 (больше половины) и даже 5 лет (20%). Были сверхвыгдные отрасли (экспорт леса), Но они зависели от таможенных пошлин, то есть от мановения руки правительства. Лоббирование. Это рискованно. Теперь все это прекратили. Вообще практически любую выгодную деятельность прекращают. Игры. Извоз. Но тут сложно. Вкладывают в добычу полезных ископаемых. Но тут тоже ограничения. Можно курить рудники, но не для нефти. Это контролируется государством, государство имеет акции. Хорошо посмотреть статистику экспорта и импорта товарной категории (такой же или похожей). Видно, где большой рынок и который растет. Есть в Международном банке статистика. Есть правительственные отчеты по рынкам отраслей (рынок строительных материалов Индии). Их можно найти. Не найдем походящий рынок – понесем большие потери. Надо найти наиболее перспективные относительно нужд и целей компании. Для этого есть способы. 1. Постепенное удаление gradual elimination and (лексикографическое упорядочение). Доводим до 5…6 стран. Стараемся неглубоко изучать, чтобы было быстро и дешево. Для начала собираем информацию по макроэкономическим индикаторам: население, показатели, связанны с доходом. Сокращаем по ним список потенциальных стран. Потом поглубже, более точные и специфичные индикаторы. Например, статистику импорта и экспорта по рынкам каждой из отобранных стран. Есть такая статистика, например, UN (www.comtrade.un.org/db/) and the Organization for Economic Cooperation and Development (OECD; www.oecd.org). Есть в странах. Еще сокращается список. (когда корректно применять это способ???) 2 этап indexing and ranking. Рассчитываем оценки [scores] общей привлекательности. Для начала соответствующие задаче переменные, присваиваем им значения – оценки. Часто относительные (КАК?) Можно уаазать веса (важность). Считаем взвешенную сумму. Это будет ранг. (когда корректно применять это способ???) Пример. Взяты переменные market size, market growth rate, and political stability. Для начала размер рынка можно оценить размером населения. Но этого недостаточно. Надо еще, чтобы быстро рос рынок, с точки зрения наседения или доходов. Значит, надо смотреть на население, его доходы и рост доходов. За 4..5 лет. Многие продукты для среднего класса. Он обычно самый большой и в сумме самый богатый (у него больше всего денег). В Индии и Китае по 200 млн чел. Перспективно даже по сравнению с Европой. A recent ranking (globalEDGE™, 2005) of attractive markets на основе нескольких критериев поставило на 1 место Китай по размеру, но на низкие места по “economic freedom” and “infrastructure.”. (найти надо данные поновее, посмотреть динамику!). Сингапур – наоборот. Индия – растет, но мало маркетинговой инфраструктуры. Соответственно, почти всегда есть компромисс [trade-offs]. Ранжирование не постоянно. Экономические события, технологии, отдельные страны развиваются. Надо смотреть. Было плохо – стало хорошо. Пример: рубли. Были деревянные. Теперь есть обмен рублей. У нас все-таки непонятна политика. Налоговая, госзакупок, контроля, Но хотя бы без резких изменений. Выбрали набор стран. Их мало. Теперь анализируем рынки. Там свои переменные. Если это soft drinks, то надо смотреть, сколько молодежи в каждой стране и качество и развитие инфраструктуры для продаж. Если это медицинское оборудование, то смотрим затраты на медицину, количество врачей per capita, количество коек [hospital beds] per capita. Банковские услуги – смотрим коммерческий риск, interest fluctuations. У нас не очень видны данные о том, кто сколько отдает, что с ними делают. Кредиты очень развиваются. Но это непривычно. В зависимости от industry, нужно ставить разные веса. Например, если речь о leisure boating industry, размер населения не очень важен. Для footwear industry размер населения очень важен. 3.3.1. Просмотр стран для глобального сорсинга и прямых инвестиций Сорсинг - ряд действий, нацеленных на нахождение, испытание и вовлечение поставщиков товаров и услуг. То есть по сути, сорсинг, это западное понятие, которое можно расшифровать как смещение центра компетенции по технологии или услуге с плеч не заинтересованных, по каким-либо причинам, в этом специалистов/руководителей обладателей этих услуг на плечи работников, заинтересованных в результате данного процесса, по заранее оговоренным условиям. Может быть не только экспорт. Бывает sourcing. Надо найти inputs от иностранных поставщиков. Или инвестиции, когда надо построить production and assembly facilities за границей. Идея та же, но другие критерии. Инвестирование. Долгосрочно. Строим или приобретаем acquire physical assets, например, фабрику или маркетинговую дочернюю компанию [subsidiary]. Тут смотрим, например, доступность квалифицированной рабочей силы или менеджерского состава. Смотрим • country risk, including regulatory (медленно ухудшается ситуация), political (стабильно), economic (теперь стали зависеть от мировых кризисов), and cultural barriers (снесло крышу), and the legal environment (развивается, но строгость законов компенсируется необязательностью их применения. Скорость на дорогах, езда по обочине, извоз, ГМО) for protecting company assets; • government incentives (стимулы), such as the availability of low-interest loans, tax holidays, subsidized training, or direct grants; Это есть, но часто меняется; • long-term growth prospects in the target country; • cost of doing business, including rates for wages and taxes (низко стало, но меняется), the cost and availability of commercial infrastructure (хотят урвать, но по сравнению с заграницей не дороже, and access to skilled workers (очень сложно!), as well as capital markets (теперь много иностранных банков); and • competitive environment, including the intensity of competition from local and foreign firms (свои дешево, но очень плохо. Машина Патриот. Есть еще патриоты, покупающие свое. Динамика была, растет производство продовольствия. Можно найти информацию. Например, методология provided by the United Nations Conference on Trade and Development (UNCTAD, available at www.unctad.org) benchmarks работу в этой области, определяет потенциал и риски для разных стран, как приемников, так и источников [recipients or originators] инвестиций. Есть Foreign Direct Investment Confidence Index, provided annually by the consulting firm A. T. Kearney (www.atkearney.com), отслеживает намерения [intentions] фирм, основанные на показателях многих стран [numerous countries]. Если ищем место для global sourcing, надо смотреть, где как покупаются готовые продукты, полуфабрикаты [intermediate goods], услуги от поставщиков за границей. Sourcing важно для фирм, особенно если они предлагают услуги. Смотрим: стоимость и качество of inputs, стабильность обменного курса; наличие рабсилы с определенной квалификацией. The consultancy A. T. Kearney (www.atkearney.com) prepares an annual Offshore Location Attractiveness Index that supports managers in understanding and comparing the factors that make countries attractive as potential locations for offshoring of service activities such as information technology, business processes, and call centers. В этом индексе учитываются следующие измереиня: • финансовая структура; величину компенсационных выплат [compensation costs] (e.g., average wages), цену инфраструктурных элементов [infrastructure costs] (e.g., electricity and telecommunications systems), затраты на налоги и регулирующие выплаты [tax and regulatory costs] (e.g., tax burden, corruption, and fluctuating exchange rates); • квалификация людей и доступность рабочей силы, включая, например, образование и язык; • бизнес-окружение – эокномика и политика. 3.4. Оценка потенциала отраслевого рынка Итак, есть 5..6 стран. Теперь глубоко смотрим. Смотрим на уровне промышленности – какие есть индексы привлекательности рынка. Рыночный потенциал. Сколько всего можно продать всем фирмам отрасли за определенный период. Потенциал продаж компании – то же, но с учетом доли рынка. Смотрим на 3 года вперед. Иногда последовательно: сначала отбираем по потенциалу рынка, потом для компании. Одновременно смотрим, как и насколько надо переделать продукт. Переменные: • размер рынка, • степень роста, • тарифные и нетарифные барьеры входа, • стандарты и regulations в промышленности, • доступность и сложность [sophistication] сети дистрибуции (у нас было очень много посредников), • характеристики и предпочтение покупателей. Есть еще и специфика по отраслям. Если фирма продает skiwear, то надо смотреть на климат, например, количество снежных дней в году. Фармацевтика. Надо смотреть сколько больных определенной болезнью, governmental expenditures on health care. 3.4.1. Методы оценки потенциала рынка Есть такие: • Посетить шоу и выставки international trade shows. Industry trade fairs and exhibitions are excellent venues to obtain information on potential markets. Посетить a trade fair in the target country. Узнаем характеристики рынка, узнаем дистрибуторов и партнеров; • ask supplier networks. Многие поставщики обслуживают нескольких клиентов и могут стать источником информации о конкурентах и их деятельности.; • monitor key industry-specific indicators. Изучаем специфические industry drivers of market demand. Комацу, manufacturer of earthmoving equipment, изучает количество объявленных проектов строительства, количество разрешений на строительство, количество домохозяйств в динамике, развитее инфраструктуры, и др.; • monitor key competitors. Смотрим на их активность. Комацу смотрит в России на Катерпиллер; • perform simple trend analysis. Производство плюс импорт минус экспорт. 3.4.2. Источник и данных Отрасль, экспорт, импорт. Полезно National Trade Data Bank (NTDB), available from the U.S. Department of Commerce’s STAT-USA and www.export.gov databases. Посмотреть, что там есть. Specific reports available from the NTDB include: • Best Market Reports, which identify the top 10 country markets for specific industry sectors; • Country Commercial Guides, which analyze countries’ economic and commercial environments; • Industry Sector Analysis Reports, which analyze market potential for individual industrial sectors; and • International Market Insight Reports, which cover country- and product-specific topics, providing various ideas for approaching markets of interest. Если информации нет то непрямо. Надо изобретать и консультироваться. Национальная статистика часто неточная, неполная, старая, и в ненужных единицах. В России. Кто производит конфеты? В статистике есть мало. Но там есть «прочие» или «смешанные». Допустим, решили продавать запчасти к мобильникам в Японию. Смотрим отчеты International Telecommunications Union, UN publications, investigate the size of the Japanese middle class and its average income, the nature of support infrastructure for cellular systems in Japan, and the nature and number of retail stores that handle cell phones. The researcher also uncovers statistics from the National Telecommunications Trade Association on the number of competitors already active in Japan and their approximate sales volumes. From these sources, the firm can arrive at a rough estimate of market size for cellular telephones and prevailing prices in Japan. 3.4.3. Выбор зарубежных бизнес-партнеров Это следующая задача. Партнеры тоже жизненно важны. Это посредники [intermediaries] в распределении, поставщики, collaborative venture partners, как например, партнеры по совместному предприятию и franchisees. Цели сотрудничества: pooling resources, sharing costs, or pursuing goals that one firm on its own cannot achieve. Надо определить qualifications of potential foreign partners. Например, если ищем партнеров для to represent продукт на иностранном рынке, ищем опытных в маркетинге, имеющих каналы. Надо, чтобы было fit между партнерами: по общим целям, по complementarity ресурсов и компетенций В динамической среде! Что смотреть еще: • financial resources and stability, возможность вложиться в совместное предприятие и обеспечить будущий рост; • competent managers, technical staff, and sales personnel; • глубокое знание рынка и отрасли; • хорошая репутация на рынке, связи с местным правительством; • готовые каналы до конечного потребителя; • sense of commitment (обязательность) и лояльность к экспортеру. Если не дотягивает, надо помочь партнеру. Передать ноу-хау и ресурсы. Как найти? Есть много предложений. Надо искать в разных местах. И консультации, и полевые исследования. Есть списки организаций, торговые и промышленные журналы, консалтинговые фирмы. У нас предлагали услуги фирмы по безопасности. Есть матметоды, например, имя нереальных организаций. Правительства предлагают недорого найти партнеров. globalEDGE™ дает инструмент. Это все надо обновлять. Приходится и индивидуально работать. Особенно на конечном этапе. Тут и сайты, и выставки. Полезно попросить составить бизнес-план. Тут выяснится, что партнер может предложить, и его квалификация. 3.4.4. Оценить объем продаж компании Это следующий шаг. Доля рынка и объемы продаж. Это очень сложно [challenging]. Информация нужна точная. Приходится делать допущения о рынке, firm’s expenses and revenues на 3–5 лет вперед. Что смотреть: • Устойчивость к риску Risk tolerance топ менеджеров. Сколько они готовы вложить ресурсов. • Финансовые и денежные ресурсы. От этого зависит скорость выхода. • Цены и затраты на продажи. Они должны быть привлекательны для покупателей и для участников каналов. • Сетевые отношения. Каковы отношения с покупателями, участниками каналов, поставщиками, консультантами, финансовыми институтами. • Возможности [capabilities] партнеров. • Доступ к каналам распределения. Establish and make best use of intermediaries and distribution infrastructure. • Intensity of competition. Местные или из третьих стран конкуренты могут интенсифицировать усилия, если кто-то еще влез. • Расписание проникновения на рынок. Постепенно или сразу. Медленно – можно исправить стратегию и получить уже прибыть для дальнейшего развития. Но может дать преимущества конкурентам. Итак, изучается много переменных. Нужны специальные знания, Важны правильные суждения. Это больше искусство, чем наука. Часто готовят пессимистическую, оптимистическую и наиболее вероятную оценку. На некоторые переменные (цены) можно влиять. На другие (реакцию конкурентов) – нет. В таблице – переменные. Table … Key Variables to Investigate in Opportunity Assessment VARIABLE TYPICAL DIMENSIONS Восприимчивость потребителя Customer receptivity Выгоды Perceived benefits of the product or service Коммуникации Nature of marketing communications directed at customer Brand positioning УТП Unique selling proposition of product Superior features of the product, compared to competitors’ offerings Каналы Channel effort and productivity Предложения для посредников Margins and incentives offered to intermediaries Конкуренты Competitors Интенсивность конкуренции Competitive intensity Наши и их относительные силы Relative strength Возможная реакция Potential reactions to market entrants Покупатели Customers Размер сегмента Size of customer segment Покупательная сила Purchasing power Рост потребности Demand growth Демография Demographic characteristics Цены Pricing Себестоимость Cost of product or service “landed” in the foreign market Маржа для посредников Usual margins for intermediaries Стратегия ценообразования Basic pricing strategy (e.g., penetration, skimming, life-cycle pricing, cost-based pricing, differentiated pricing) 3.4.4.1. Практические методы оценки потенциала продаж компании Вот они: • Conducting trade audits. Посещаем розничные outlets, интервьюируем channel intermediaries о потенциале продаж. Это помогает также определить характеристики покупателей, правильные уровни цен, может быть, различные подходы к маркетингу, природу конкурентов. Смотрим с точки зрения intermediaries, которые много знают. • Obtaining estimates from local partners. Интервью с suppliers, franchisees, and banks. Они уже опытные. • Surveying end users. Проводим (develops and administers) опрос выборки покупателей. Можно и фокус-группы. • Engaging in test marketing. Прямо можно измерить потенциал продаж на малой части рынка. Можно хорошо долгосрочно предсказать. Хорошо в развивающихся рынках, где мало вторичной информации, Но дорого. • Using analogy. Берем статистику из одной страны и по ней узнаем [gain insights] того же явления в другой стране. Если хотим продать candy bars в Пакистане, смотрим, как было в Индии. Допустим, потребление не сильно разнится. Корректируем по численности населения. • Using proxy (замещающие) indicators. Оцениваем продажи одного продукта по известным продажам аналогичного. Например, надо определит продажи компьютерных клавиатур. А есть только исторические данные о продажах мониторов. Обычно же вместе, так что можно оценить. Хорошо, если это комплементы. В России печатали объявления в газете. Клавиатура выдерживала месяц, мышь – неделю. • Assessing competitors. Можно benchmark себя против главных [principal] конкурентов и оценить, сколько можно перетащить [attract away] к себе. Последнее плохо, так как конкуренты могут иметь большие ресурсы для борьбы. Детально – потом. 4. Практическая задача The Idea of this first serious task is to focus on real problem on which it will be possible to see specifics in practice. Maybe the chosen task will not include all methods and objects of research, but practical approach is of great use: one can see what to find in the research. The task is to find perspective business possibility. Найти перспективную маркетинговую возможность. Здесь рассмотрим возможные ситуации в рамках курса. Важно: шаг за шагом! Не торопясь, не забегая вперед. 4.1. Определение потребностей в исследованиях для типовых ситуаций Before any other steps it is necessary to learn the following. Do not think it is easy to define what is necessary to organization. At first when the need for internationalization is just revealed as possibility, no one knows what exactly it should be. In Russia it is very often that top management does not see the variants of future development. 4.1.1. Выход организации на внешний рынок Возможные цели этой организации Получить не менее заданного объема прибыли в заданный срок. Например, 20 млн. евро за 3 года. Что такое прибыль и зачем она нужна? Занять определенную долю рынка к заданному сроку. Какие возможны другие цели? Что требуется узнать Определить маркетинговые возможности Оценить их Дать рекомендации: какой товар на какой рынок когда и как выводить15, чтобы достичь поставленной цели. Какие данные требуется собрать О внутренней среде • Какие продукты уже выпускаются • Какие могут выпускаться или какие возможны направления развития • Какова себестоимость выпускаемых продуктов или тех, которые можно будет выпускать О внешней среде • Каков прогноз спроса на выпускаемые продукты на каждом рынке • Каков прогноз конкуренции для каждого вида продукта и каждого рынка • Какую цену возможно установить на продукт Это исходный перечень. Вроде, все важное перечислено. Всё ли здесь, как дополнить список? Исходное положение было такое: надо, чтобы продукт продавался в заданном объеме по цене, на заданную величину выше себестоимости. Объем продаж зависит от размера рынка, нашей доли на нем. Цена, которую можно установить, зависит от потребности, цен конкурентов. Стоимость зависит от нашего производства. До некоторой степени на параметры можно влиять с помощью маркетинга (у нас в аптеках есть разные марки лекарства, например, Мезим. Мезим разрекламирован, цена на него высокая. Есть такое же лекарство, дешевле. И есть такое же, дороже). Надо еще источники финансирования Надо бы построить модель факторов и переменных Может быть, рыбий скелет. Возникает ряд вопросов: а знаем ли мы все данные, которые требуется собрать? Знаем ли, как собирать? Не сложновато ли будет: для каждого рынка и для каждого товара? Как отсеять лишнее? Может быть, что-то пропустили? И общий вопрос отсюда: чего мы не знаем? Пока даже не КАК будем собирать, а ЧТО собирать? Каков порядок: как отсеять лишнее сразу? Может быть, уже есть какие-то результаты, известно, что не надо делать определенные вещи (исследовать определенные рынки?). Получается, что исследования бывают разные, цель достигается не сразу. Не просто задать параметр (рыночный потенциал) и его определить каким-либо способом. Бывает, что еще и неясно, какие данные надо собирать. 4.1.2. Создание нового бизнеса Возможные цели предпринимателя Аналогично предыдущему случаю Что требуется узнать Определить маркетинговые возможности Оценить их Дать рекомендации: какой товар разработать, какой рынок когда и как выводить, чтобы достичь поставленной цели. Какие данные требуется собрать Домашнее задание для бригады 1, с презентацией. 4.1.3. Разработка маркетингового исследования на продажу Возможные цели исполнителя Аналогично предыдущему случаю Можно ли сказать, что область возможных задач охвачена вся? Как это проверяется? Намек: аналогично сегментированию рынка: пол М,Ж, возраст: до 16, 16-21, 21-55, свыше 55. и т.д. Тут ясно, что охватывается весь рынок. Что требуется узнать Определить возможные тематики исследования Оценить их Задать тему исследования, определить способ сбора, обработки и представления данных. Какие данные требуется собрать Домашнее задание для бригады 1, с презентацией. *** На основе проделанной работы возникает ряд вопросов: а знаем ли мы все данные, которые требуется собрать? Знаем ли, как собирать? Не сложновато ли будет: для каждого рынка и для каждого товара? Как отсеять лишнее? Может быть, что-то пропустили? И общий вопрос отсюда: чего мы не знаем? Пока даже не КАК будем собирать, а ЧТО собирать? Каков порядок: как отсеять лишнее сразу? Может быть, уже есть какие-то результаты, известно, что не надо делать определенные вещи (исследовать определенные рынки?). Получается, что исследования бывают разные, цель достигается не сразу. Не просто задать параметр (рыночный потенциал) и его определить каким-либо способом. Бывает, что еще и неясно, какие данные надо собирать. 4.2. Шаги 4.2.1. Найти предпринимательские возможности Это наименее формализуемая задача. Способы очень разные. Это и наблюдение за жизнью за границей, и анализ литературы и т.д. Пример. Гемодиализ. Фирма занимается производством материала для фильтров, используемых при гемодиализе. Шаг 1 обнаружение отклонений. Исходная информация: отчет международных медицинских организаций по количеству пациентов на 1000 человек. Видно, что в некоторых странах 1000 больных на 100 тыс. чел., чаще от 500 до 900, в России – меньше 100! То есть страна сильно выделяется. Шаг 2. Анализ причин. То ли все здоровые, то ли плохое лечение. Метод: опрос информированных лиц: медицинских работников, предпринимателей из отрасли медоборудования. Результат: лечение очень дорогое (10 тыс. долл в месяц), осуществляется практически по дотациям государства. Денег мало выделяется, многие остаются без помощи. Но лечение этой болезни еще неплохо датируется. Можно ли улучшить ситуацию? Если сделать продукт значительно дешевле. Но это высокотехнологический процесс. Пока не получается. Вторая сложность: есть материал, но нет оборудования. Его надо производить. В России дешево, но мало кто хочет: требуется сертификация, иностранный материал – сложно. Проще одноразовые шприцы делать. Из-за границы – дорого. Оказалось, что проблема есть, но реализовать не получается. Необходимо найти несколько возможностей. Или обосновать нецелесообразность некоторых предложений. 4.2.2. Предварительная оценка целесообразности работы с конкретным продуктом 1. Повторное использование одноразовых продуктов: заправка одноразовых картриджей. Огромный бизнес. Большой спрос на расходные материалы. Отечественные производители делают оборудование. Например, сверло и пробку для заправки одноразовых картриджей. Пытаются делать и чернила. 2. Госзаказы (саженцы деревьев). Источник информации? Периодика, ТВ новости. Были питомники, Саженцы липы, елки и проч. Растут лет 10. Оказалось, что всё заброшено, закупать приходится за границей. Это дорого. Какие тут есть возможности? Перечислить, проанализировать. 3. Продажа ювелирных изделий. 4. Продажа программного обеспечения 5. Разлагающиеся упаковочные материалы 6. Переработка использованных батареек. 7. Сбор бутылок автоматами, как за границей 8. Продажа саженцев 9. Офисное корпоративное программное обеспечение 10. Мочалки-полотенца 11. Феномен ИКЕА 12. Другое (что?). Требуется: определить тип исследования, метод, собираемые данные. способ обработки для оценки перспективности. Затем провести исследование по данным из Интернет. Выбрать более перспективный вариант. 5. исследовательский проект Этот раздел – краткая описание процесса маркетинговых исследований вообще. В нем приводятся лишь самые необходимые сведения, предназначенные для напоминания об этапах маркетинговых исследований. Подробные сведения о маркетинговых исследованиях можно почерпнуть из книг Черчилля и Малхотры. *** Единого алгоритма проектирования исследования не существует. Особенности проекта, в том числе и методы исследования, определяются конкретной проблемой. Все же можно выделить определенные типы исследований и наиболее часто используемые для них методы. Пока здесь методы даются обзорно, чтобы начать процесс маркетингового исследования на практике. Следует отметить, что, пожалуй, главное влияние на методы исследований оказывает та информация, которой уже обладает исследователь. Чем больше априорных знаний. Тем легче проводить исследование. Так, анкета, в которой респондент легко и просто отмечает галочкой вариант ответа и которую потом легко обрабатывать, основывается на довольно большом объеме уже имеющейся к началу исследования информации. Ведь нужно так сформировать вопрос, чтобы дать перечень практически всех возможных ответов! А знание всех возможных вариантов означает, что исследователь уже довольно глубоко проработал проблему. 5.1. Определение информации, которую требуется получить Должна быть четкая связь с проблемой управления организацией. 5.2. Определение типа и метода исследований 5.2.1. Поисковое исследование Sometimes it is called Marketing Intelligence. В этих исследованиях осуществляется поиск идей, возможных объяснений происходящего, производится разбиение общей проблемы на подзадачи, выдвижение гипотез. Пример вопроса, на который отвечает поисковое исследование: «какие переменные и как влияют на долю рынка компании». Для ответа на него прежде всего следует узнать, связано ли изменение доли рынка с деятельностью компании или с ситуацией в промышленности и экономике. Вначале надо обратиться к вторичным данным. Обычно доля рынка растет, если растет объем продаж компании. Но это происходит в условиях, если отрасль в целом стабильна. Если же отрасль претерпевает подъем или спад, то доля рынка будет изменяться и при постоянном объеме продаж. Естественно, исследования будут сильно различаться для приведенных случаев. Поэтому начальный шаг исследования – прояснение проблемы – очень важен, так как он задает направление дальнейших усилий. Часто проблема имеет простое решение, которое становится очевидным после анализа ситуации в собственной фирме.. Основная цель поисковых исследований – формирование исходной гипотезы (например, о наличии зависимости количества потребляемых конфет от семейного положения). Гипотеза это утверждение, определяющее, как две или более измеряемые величины относятся между собой. Хорошая гипотеза должна также давать четкий способ проверки данного утверждения. 5.2.1.1. Цели поискового исследования Целями поисковых исследований являются: • формулировка проблем для дальнейшего исследования; • улучшение понимания проблемы; • выдвижение новых гипотез; • уточнение концепций, гипотез; • определение приоритетов в исследованиях. 5.2.1.2. Основные методы Анализ литературы Он занимает сравнительно мало времени, дешев, позволяет обнаружить гипотезы других исследователей по данной проблеме. Литература, которой обычно пользуются при маркетинговых исследованиях, может быть разделена на • теоретическую (по психологии, социологии, кадровым проблемам); • содержащую практический опыт (результаты внедрения прогрессивных методов); • статистические данные. Поиск проводится в библиотеках (просматриваются книги, периодика, материалы конференций, сборники трудов, статистические обзоры и т. д.) или среди материалов фирмы (изучаются отчеты о предыдущих маркетинговых исследованиях, финансовые документы, отчеты и предложения сотрудников фирмы и др.). Для ориентации в потоке публикуемой информации библиотеки составляют тематические каталоги, каталоги по авторам, по названиям. Ежемесячно издаются реферативные журналы по разделам науки и техники. В них имеются резюме на большинство вновь вышедших книг и статей. В последние годы инструментом распространения информации стала компьютерная сеть Интернет. Поисковые системы предоставляют возможность быстро найти источники, в которых имеется заданное при поиске слово. Многие библиотеки также перешли на электронные каталоги. Опрос информированных лиц Key informatant survey Это опрос тех, кто хорошо знаком с проблемой. Полезно опросить всех, кто потенциально может дать полезную информацию, прежде всего: • высшее руководство; • менеджеров по продажам; • менеджера по товару; • торговых представителей; • продавцов; • покупателей. Иногда вопрос, к кому обратиться для поиска решения проблемы, не столь очевиден, например, ответ на вопрос о спаде спроса на детские книги был найден в библиотеке [Chirchill] Главная цель опроса – получить ответ на то, как взаимосвязаны переменные. Это не сбор статистических сведений и не поиск рецептов того, что следует делать. Выбор опрашиваемых следует производить очень тщательно. В данном случае он никоим образом не является вероятностным. Исключаются некомпетентные лица, а также те, кто не может рассказать о своих знаниях. Полезно выслушать различные точки зрения. Анализ избранных случаев Для выяснения того, как добиться успеха на поприще продавца, естественно исследовать несколько худших и нескольких лучших продавцов. Скорее всего, их сравнение и даст ответ. Аналогично рассматриваются характеристики фирм, территорий. Целью применения метода является поиск возможных объяснений, а не их проверка. Особенности метода: выбор случаев для анализа зависит от отношения исследователя; направление исследования может изменяться в зависимости от полученной информации; результат определяется способностью исследователя к обобщению, умением отличать уникальное от закономерного. Фокус-группы Наибольшее распространение метод получил в 1990-е годы. Суть его состоит в том, что коллективное обсуждение более эффективно чем индивидуальна работа. Несколько людей (обычно это покупатели или потребители) собираются вместе и беседуют на интересующую исследователей тему. Ведет обсуждение moderator. Он старается направить дискуссию по нужному руслу и регистрирует ценные мысли по ходу обсуждения. Такой подход хорошо оправдал себя в случаях, когда требовалось: • сформулировать новую гипотезу, которую можно было бы далее разрабатывать; • собрать мнения об общей концепции нового товара; • определить восприятие сети магазинов или имидж торговой марки. Главная роль принадлежит ведущему, который должен вести дискуссию и фиксировать ее результаты. Аудио и видеозапись полезна, но пользоваться ей надо осторожно, получив согласие участников. Она может немного сковывать их. Главное в поведении фокус-группы – не опрос по анкете и не обобщение данных в виде подсчета и группировки мнений, а именно фиксация отдельных высказываний. Важна каждая фраза! Именно она может оказаться ключевой, даже стать лозунгом, обеспечивающим успех! При обсуждении продукта для промышленного применения задача усложняется. Если все мы в той или иной степени являемся покупателями потребительских товаров, то в данном случае • от участников требуется большой опыт работы; • как правило, все участники обсуждения знают гораздо больше, чем ведущий. Положительные стороны метода: • идеи рождаются как бы «из воздуха»; • происходит «жонглирование идеями». Один бросает мысль, другой ее развивает и т. д.; • многие чувствуют себя более свободно в группе, чем при интервью. Здесь можно бросить идею и не заботиться о ее обосновании. Отрицательные стороны метода: • обсуждение трудно вести и трудно интерпретировать его результаты; • основная проблема при использовании данного метода – выбор и подготовка ведущего. Проблемные группы Исследования показывают, что групповое мышление производит на 70% больше ценных идей, чем сумма индивидуальных мышлений [8]. Поэтому групповое обсуждение часто используется и специалистами-маркетологами. При разработке анкеты такое обсуждение позволяет значительно повысить ее полезность для исследований и понятность для респондентов. Коллективное решение проблемы поможет выяснить, что и в какой последовательности надо делать в сложных ситуациях. Работникам промышленности знакомы «летучки» и «пятиминутки». В начале рабочего дня собираются начальники групп отдела. Они обсуждают текущие проблемы и намечают пути их решения. Аналогично действует и рабочая группа, выполняющая определенные исследования. Общие собрания должны проводиться не реже одного раза в неделю. Обычно размер такой группы невелик (до 20 человек), а состав длительное время остается постоянным. В таких группах, как показывают исследования16, происходит долговременное распределение ролей. В группе появляются такие «персонажи», как «генератор идей», который выдает широкий спектр гипотез, «эрудит» приводящий соответствующие случаю факты, «критик», который отвергает большинство идей как нереальные. «Руководитель» наблюдает за ходом дискуссии и выносит окончательный приговор. Примером эффективности группового обсуждения является игра Что? Где? Когда? Результат видели миллионы телезрителей: часто никто из команды не знает правильного ответа на поставленный вопрос, но через одну минуту этот ответ появляется. Аналогичная команда может решать и более серьезные проблемы. Экспертные оценки Получить результаты формальными методами удается далеко не всегда. Случается так, что единственным способом решения проблемы является использование интуиции и неформализуемого опыта. В маркетинговых исследованиях такой подход используется достаточно часто. Эксперт – лицо, имеющее высокую квалификацию, опыт и интуицию в определенной области. На первый взгляд кажется, что метод достаточно прост: надо узнать мнение экспертов. Однако требуется владеть методическими основами проведения экспертных опросов. Применяются как индивидуальные, так и групповые экспертные оценки. Отдельный эксперт может быть полезным советником для лица, принимающего решения. Он может проанализировать и обобщить результаты, представленные другими экспертами. Отдельному эксперту лучше, чем группе, поручать разработку сценариев развития событий и прогнозов. Групповые оценки экспертов одной или нескольких специальностей проходят путем анкетирования или обсуждения. При этом высказываются различные мнения по определенным вопросам. Группа экспертов хорошо решает проблемы поиска всех альтернатив решений, всех возможных вариантов ответа на вопрос анкеты. Есть также некоторые основания надеяться, что усредненная оценка мнений нескольких экспертов будет точнее индивидуальной. Эксперты могут давать оценки различных типов: • качественные (каковы возможные пути развития фирмы и ее маркетинговой деятельности, какой из них наиболее предпочтителен, какие мероприятия требуются для продвижения нового товара, какими средствами рекламы лучше воспользоваться и т. д.); • сравнительные (новые потенциальные рынки просят расположить от наиболее перспективного до наименее перспективного); • балльные (дать общую оценку плана маркетинга по шкале отличный, хороший, удовлетворительный, плохой); • количественные (ожидаемый уровень инфляции, прогнозируемая доля рынка, важность определенной цели). Точность этих оценок практически не поддается контролю. • В зависимости от типа даваемых ответов происходит их обобщение. Среднее значение можно определить для количественных оценок, для качественных ответов возможно определение моды (какой ответ встречался чаще всего). Индивидуальные эксперты Однократные опросы обычно заключаются в том, что эксперты заполняют специально составленные анкеты. Далее полученные оценки обобщаются (в простейшем случае для количественных оценок берется среднее значение, а для качественных – медиана или мода). Отличие от потребительских анкет состоит в том, что даются справочные материалы, а ответ обычно просят сопроводить кратким обоснованием. Как правило, опросы проводятся анонимно из этических соображений: эксперт может дать и ошибочную оценку, а отказаться от своего мнения – значит рискнуть репутацией. Метод Дельфи Метод «Дельфи», первоначально примененный в 1940-х годах, получил свое название от древнегреческого города Дельфы, знаменитого своим оракулом – советом мудрецов, на котором тщательно обсуждались предсказания, даваемые пифией, перед тем, как их обнародовать. Современный метод «Дельфи» предусматривает проведение экспертных опросов в несколько туров. Процесс выработки суждений экспертами управляется исследователями с помощью обратной связи. Прямое обсуждение заменяется обменом информацией и мнениями с помощью анкет. Участники не только высказывают свое мнение, но и обосновывают его. Собранная информация анонимно передается для повторного осуждения. Если ответы сильно разнятся, определяют верхний и нижний квартиль оценок, а затем экспертам, давшим низкие оценки, мнения, дают прочитать мнении экспертов, давших высокие оценки. Так может повторяться до трех раз, однако часто уже во втором туре оценки получаются довольно единообразными. Метод успешно применяется, например. При определении стоимости сложных маркетинговых проектов. Мозговой штурм Brainstorming В этом методе происходит коллективное обсуждение идей группой экспертов. Поэтому метод напоминает фокусную группу. Перед началом обсуждения следует точно сформулировать проблему и основной решаемый вопрос. В процессе обсуждения следует обеспечить свободное высказывание идей. Для этого • принимаются меры к запрещению любого рода критики высказываний; • сообщается, что результаты обсуждения считаются плодом коллективного труда всей группы. С помощью этого метода успешно решаются следующие задачи: • формулировка полного набора альтернатив для принятия решения (обычно это касается стратегических вопросов, возможных направлений развития); • выявление полного набора методов, применяемых для решения имеющейся проблемы; • определение полного набора важных факторов, учитываемых при выборе оптимального решения. Оценка точности экспертных опросов Важно отметить, что достоверность результатов экспертного опроса не поддается измерению. Например, стоимость проекта может быть оценена только после его реализации или хотя бы тщательного планирования. Часто есть соблазн подсчитать ширину доверительного интервала, как это делается для выборочных исследований. Однако это неверно. Если в выборочных исследованиях ищется значение для всей совокупности, то при экспертных опросах получаются просто некоторые числа, а истинное значение может лежать далеко от полученного среднего. 5.2.2. Описательные исследования Целью исследований этого типа является определение частоты какого-либо события или определение количественной зависимости между величинами. Обычно задается требуемая точность получаемых результатов. 5.2.2.1. Цели Описательные исследования могут проводиться с целью: • описать характеристики определенной группы, например, «среднего пользователя» – его доход, возраст17; • оценить долю людей, которые ведут себя определенным образом (например, делают покупки преимущественно в супермаркете), среди всей совокупности; • сделать специальный прогноз, например, каким образом лучше продавать конкретный товар. Важно отметить, что задача состоит не просто в сборе фактов, а в их объяснении, осмыслении, увязывании с теорией. Поэтому до начала исследования необходимо некоторое априорное знание, гипотеза, которая и должна определять направление исследования. Пусть открылся новый магазин и требуется узнать отношение к нему среди посетителей. Исследование должно быть спланировано так, чтобы были четко определены ответы на следующие вопросы. Зачем нужно исследование? Для организации рекламной кампании? Тогда важно, как именно покупатели узнали о магазине. Для поиска места нового магазина? Тогда важен район, обслуживаемый магазином. Что требуется узнать? Пол? Возраст? Место жительства? Способ, которым узнали о магазине? Кого спрашивать? Того, кто вошел? Но он, может быть, просто зашел в надежде получить бесплатный образец товара на открытии. Того, кто купил? Но покупка могла быть сделана только для себя лично или на всю семью. Когда спрашивать? В момент покупки? При выходе из магазина? А может быть, положить анкету в упаковку, чтобы на нее ответили дома? Проводить ли опрос сразу на открытии магазина или через несколько недель, когда ситуация стабилизируется18? Где проводить опрос? Внутри магазина, у входа, у кассы, при выходе, на улице, дома? Как собирать информацию? Анкетированием (какой из многочисленных типов анкетирования выбрать? Каким способом распространять анкету: давать в руки, посылать по почте, проводить интервью или телефонный опрос?), наблюдением? Шаблоны таблиц результатов исследования помогут определить, какие вопросы задавать, на что обращать внимание при наблюдениях. Например, для определения предпочтительной марки кофе, заготавливается таблица, аналогичная табл. 2.1. Такая таблица определяет содержание и ход опросов, а также кодирование ответов. Таблица 2.1 Шаблон результатов исследования Количество потребителей, предпочитающих различные марки кофе, в зависимости от возраста19 Возраст Предпочитаемая марка Jacobs Tchibo Nescafe20 до 30 30 – 39 40 и более 5.2.2.2. Descriptive Research Methods Описательные исследования классифицируются согласно рис.Рис. 4. Каждый из типов будет подробно рассмотрен ниже. Рис. 4. Классификация описательных исследований Панельные исследования Panel Research Список постоянных респондентов помогает получить важную информацию о переходе от одной марке к другой. Список – набор респондентов, согласившихся участвовать в исследовании и давших о себе подробную информацию. При опросе известны данные каждого респондента, причем их опрашивают повторно. В результате определяется, как изменились предпочтения каждого респондента в результате определенных маркетинговых действий, например, введении новой упаковки товара. Основным недостатком метода является его нерепрезентативность21: далеко не все хотят тратить свое время, заполняя дневник. Поэтому фирмы всегда ищут хороших участников для списка. Часто при составлении списка используется принцип квот, чтобы охватить различные слои населения. Список «омнибус» отличается тем, что один список используется для исследований различной тематики. Разовые исследования Cross-sectional research Эти исследования, как следует из их названия, проводятся единовременно. Полное обследование имеет место в случаях, когда исследователей интересует небольшая группа людей (студенты определенной кафедры), фирм (потенциальные потребители крупного промышленного оборудования), магазинов (сеть фирменных магазинов района). Чаще исследуется только определенная часть этой группы (выборка), по которой делаются выводы о характеристиках всей группы. Характерные черты метда. • Он дает «фотографию» одного момента. • Выборка является репрезентативной, обычно – случайной. • Широта опроса заменяет глубину. 5.2.3. Исследования причинности В них определяется или проверяется причинная взаимосвязь событий и явлений. Если наличие причинно-следственной связи не вызывает сомнений (например, известно, что при увеличении температуры воздуха увеличивается потребление безалкогольных напитков), и требуется определить силу этой связи (некоторые торговые автоматы использую датчик температуры воздуха для изменения цены), то исследования причинности становятся похожими на описательные. Если же причинно-следственная связь не ясна с самого начала, то ход исследований будет определяться последовательной детализацией вопроса и вскрытием новых проблем. В этом случае исследования приближаются к поисковым. Например, если уменьшаются объем продаж и падает доля рынка, то причиной этого может быть неудовлетворенность покупателей качеством продукции или просто неблагоприятное мнение о магазине. Причинность – сложная философская категория, которая даже до конца не определена, хотя ею широко пользуются на практике. Идея данного типа исследований – проверить гипотезу о том, что причина Х вызывает следствие Y. Примеры гипотез: • пятипроцентное увеличение цены не скажется на объеме продаж; • введение упаковки с более легким способом открывания повысит спрос. Если в обычной жизни, как правило, говорится или подразумевается, что именно Х вызывает Y, то в научной постановке Х признается как одно из возможных условий. Если в обыденном смысле Х всегда ведет к Y, то ученые чаще говорят о том, что событие Х повышает вероятность события Y. Полной уверенности в правильности гипотез о причинности быть не может. Существуют только принципы, позволяющие повысить уверенность в их справедливости. Их достаточно много. Здесь будут приведены только те из них, которые наиболее часто используются в маркетинговых исследованиях. 5.2.3.1. Принципы Анализ совместного изменения Concomitant Variation Analysis Анализируется вероятность, с которой Х и Y происходят совместно или совместно изменяются (количественно и качественно) согласно гипотезе. Пусть Х – оценка качества дилера (как его оценивать – отдельная проблема. Пусть для определенности, это будет оценка дилера его непосредственным начальником), а Y – доля рынка в регионе, который обслуживает дилер. Естественным было бы утверждение: доля рынка зависит от качеств дилера (там, где дилер хорош, доля рынка велика и наоборот). Пусть собраны следующие данные (табл. 2.7): Таблица 2.7 Количество и процент регионов с различной долей рынка в зависимости от оценки деловых качеств дилера Оценка дилеров Доля рынка в регионе Большая Небольшая «Хороший» 50 (67%) 25 (33%) «Плохой» 15 (25%) 45 (75%) Видно, что хорошие дилеры обеспечивают большой процент благополучных регионов, в то время как у плохих – высок процент неблагополучных. Табл. 2.7 называется таблицей сопряженности22. Такие таблицы часто используются и в описательных исследованиях, однако в данном случае на их основе происходит проверка гипотезы о взаимосвязи переменных. Важно отметить, что для гипотезы требуются два утверждения. Так, высказывание у хороших дилеров продажи успешны, основанное на первой строчке таблицы, еще не есть гипотеза, поскольку неясно, как обстоит дело у плохих дилеров (может быть, так же или даже лучше). Поэтому обязательно продолжение: …а у плохих – нет. В более правильной формулировке данная гипотеза может звучать так: у дилеров с хорошими деловыми качествами процент регионов с большим объемом продаж на (67-25)=42 процентных пункта23 больше, чем у дилеров с плохими деловыми качествами. В результате анализа таблицы гипотеза делается более вероятной, но не доказанной на 100%. Отсутствие связи и даже обратная зависимость также ничего не доказали бы, поскольку на результат могут влиять и другие факторы (например, хороших дилеров, может быть, специально послали в плохие регионы для налаживания продаж). Анализ временной последовательности Time Sequence Analysis Предполагается, что причина всегда происходит раньше, чем следствие или одновременно с ним24. Например, если каждый раз после того, как в магазине появляется товар в новой упаковке, в этом магазине падают продажи данного товара, то это – серьезный аргумент в пользу того, что упаковка оказалась неудачной. Сложность применения метода состоит в том, что причину и следствие не всегда можно четко разграничить. Например, имеется гипотеза: повышение затрат на рекламу повышает объем продаж. Но многие фирмы тратят на рекламу определенный процент от доходов прошлых периодов, которые, естественно, зависят от объемов продаж. Здесь причина и следствие находятся в тесной взаимосвязи и их трудно разделить. Исключение других возможных объяснений Elimination of Other Possible Explanations Этот метод широко освещен в детективной литературе. Основной подход – попытаться зафиксировать все возможные причины изучаемого явления и исключить не связанные с проблемой. Например, причиной большого числа зрителей телепередачи может оказаться просто плохая погода, что никак не говорит об истинных достоинствах этой передачи. Главная проблема состоит в том, чтобы выявить все возможные объяснения. Это не просто трудоемко. Это требует творческих усилий, практического опыта работы. Интересные исследования проводились для определения влияния буквенных обозначений на восприятие марок кофе. Оказалось, что если при дегустации пометить банки с кофе буквами, то сами буквы могут определить отношение к сорту кофе. Исследовалось восприятие букв. Буква А была для большинства наиболее приятной, а вот буквы Q, X, Z вызывали отрицательную реакцию. На основе этих исследований для маркировки подбирались буквы, воспринимаемые одинаково. Это исследование проводилось в США. Очевидно, для каждой страны (а может быть, и для определенных групп людей) восприятие букв может оказаться различным25. Уровень добросовестности исследователей в любой области науки определяется тщательностью устранения или хотя бы учета посторонних факторов26. Видно, что, хотя по собранным в результате исследования данным можно выдвинуть гипотезу и получить подтверждения ее правильности, полной уверенности в ее справедливости добиться не удается. 5.2.3.2. Экспериментальные и неэкспериментальные методы В разделе «Анализ совместного изменения Concomitant Variation Analysis» было показано, как причинно-следственная связь определяется по перекрестной таблице, полученной в результате исследований. Применявшийся при этом метод сводится к анализу уже собранных данных о значениях переменной Y и переменных Хi , которые могут оказывать на нее влияние. При этом достаточно высока вероятность наличия других причин изменений. Их, по мере возможности, выявляют и учитывают. Это достаточно распространенный, и зачастую единственно возможный способ анализа реальной рыночной среды. Уверенность в правильности заключений о причинно-следственной связи стала бы значительно выше, если бы удалось, произвольно изменяя значения причинной переменной, наблюдать за изменениями переменной-следствия. При многократном повторении таких действий с аналогичным результатом уверенность в наличии и силе причинно-следственной связи могла бы стать почти полной. Эксперимент – это научное исследование, в котором исследователь сознательно манипулирует одной или несколькими независимыми переменными и наблюдает за возникающими в результате вариациями зависимых переменных. Хотя пока в российских условиях маркетинговые эксперименты не имеют большого распространения ввиду сложности организации и достаточно больших затрат, этот мощный инструмент может принести большую пользу и имеет хорошие перспективы. В полевых экспериментах, которые проводятся в реальных условиях, нельзя полностью освободиться от влияния посторонних факторов. В лабораторных экспериментах создаются особые условия, позволяющие минимизировать все посторонние влияния. Результаты лабораторного и полевого экспериментов могут отличаться по различным причинам. Валидность результатов эксперимента – степень уверенности в их правильности. Существует множество видов валидности. В лабораторных экспериментах исследовалась зависимость цена – спрос. Испытуемым (добровольцам, собравшимся в лабораторию) давались карточки с названиями товаров и ценами. Эти цены обсуждались, составлялся список «псевдопокупок». Эта же зависимость исследовалась в полевых условиях, в супермаркетах. Записывалась недельная продажа перед изменением цен и через две недели после изменения. Никакой рекламы об изменениях цен не было. Результаты двух исследований показали тенденцию к преувеличению роли цены в лабораторных экспериментах. Для лабораторных условий, помимо указанного эффекта, следует иметь в виду и то, что участвовать в таких экспериментах соглашается определенная категория людей (прежде всего, имеющих свободное время), что может исказить результаты. Внешняя валидность – степень уверенности в соответствии результатов исследования действительности. Из приведенного выше примера видно, что в лабораторных экспериментах эта валидность довольно низка. В полевых экспериментах наблюдается реальная ситуация, а не ее модель, поэтому внешняя валидность обычно выше. Внутренняя валидность – степень уверенности в том, что наблюдаемый эффект обуславливается именно экспериментальной переменной, а не другими факторами. В рассмотренном полевом эксперименте не было рекламы, специально оформленных витрин и тому подобных посторонних для эксперимента факторов. Но не исключено, что одновременно с ним проводилась реклама другого товара или магазина, что могло повляить на результаты. Так что внутренняя валидность полевых экспериментов обычно низка. Она выше у лабораторных экспериментов. Таким образом, чтобы результаты эксперимента оказались полезными, следует тщательно учитывать возможные источники ошибок. 5.3. Формы сбора данных Первичные данные – данные, собранные специально для проводимого исследования, в первый раз. Вторичные данные – данные, собранные ранее для других целей. 5.3.1. Сбор вторичных данных This is starting point of most researches. Вот как можно провести исследование по вторичным данным. Пусть требуется оценить рынок еды для собак. Из ветеринарного справочника можно узнать потребность различных пород в пище, а из отчетов клубов собаководства – численность собак в районе по породам. Далее следуют несложные вычисления. Вторичные данные целесообразно использовать, когда требуется: • уточнить формулировку проблемы; • определить перечень данных и наилучшие методы их обработки для решения проблемы; • получить сравнительные данные, которые помогут интерпретировать первичные данные. 5.3.1.1. Особенности Достоинства вторичных данных – в простоте их сбора. Но вторичные данные имеют следующие недостатки. 1. Данные не точно соответствуют проблеме по определенным параметрам: • по единице измерения (нужен доход индивидуума, а данные есть только по семьям; размер магазина требуется оценивать по площади торгового зала, а сведения имеются по ежедневной выручке или по количеству работников); • по шагу измерения или границам классов (потребителей требуется разделить по уровням доходов 200 руб/мес, 400 руб/мес, 600 руб/мес, а данные имеются по границам 250, 500, 750); • по свежести (данные переписи населения могут быть собраны несколько лет назад). 2. Данные недостаточно точны. При использовании вторичных данных следует учитывать: 2.1. Источник данных. Первичный источник вторичных данных публикуется непосредственно той фирмой или человеком, которые занимались исследованием, а вторичный источник вторичных данных представляет собой обзор исследований других исполнителей. Хотя последние источники часто дают более широкую картину проблемы, следует по возможности пользоваться первичными источниками вторичных данных, так как • в них как правило описываются методы сбора и анализа данных, что позволяет оценить точность и применимость данных для нашего исследования; • они как правило наиболее точны и полны, т.е. в них есть ссылки, комментарии; • в них меньше ошибок, возникающих при перепечатках. 2.2. Цель публикации. Как можно понять фразу: «Увеличилось число жалоб на авиакомпанию Х»? Полетите ли Вы на самолете этой компании? Если же посмотреть на то, откуда взята эта фраза, то выяснится, что из статьи, изданной фирмой-производителем сигарет, в которой обсуждается запрет курения в самолетах. Теперь смысл фразы для кого-то станет противоположным. Подозрительными являются источники, направленные на защиту интересов промышленников и торговцев; на продвижение товара; документы политических партий; пропагандистские материалы; анонимные; выпущенные организацией, которая хочет оправдаться в чем-либо; полемического характера; чересчур откровенные; опровергающие выводы других источников. Заслуживают доверия сведения фирмы, для которой основной деятельностью является проведение исследований и публикация данных. В то же время бывает, что исследования, опубликованные в маркетинговых журналах, не имеют четко поставленной цели, а служат лишь для того, чтобы показать, какие интересные исследования может проводить фирма. В таких случаях ценность такой информации низка, так как для решения конкретной задачи обычно удается использовать лишь около 10% представленных результатов. 2.3. Возможности авторов собрать достоверные данные. Как правило, данные крупного государственного учреждения заслуживают большего доверия, чем сведения мелкой неизвестной фирмы. Но, с другой стороны, ответ на вопрос о величине доходов может быть разным для государственного налогового инспектора и интервьюера исследовательской фирмы. При анализе достоверности вторичных данных полезно задаваться следующими вопросами. • По каким принципам производился отбор элементов исследования? • Каким методом собирались данные: опросом или наблюдением; как это следовало бы сделать? • Каково качество подготовки полевых работников? • Как проверялась их работа? • Каков был процент отказов и есть ли отдельная статистика по этому вопросу? • Хорошо ли представлена информация? • Не противоречивы ли данные? • Правомерны ли выводы, сделанные по собранным данным? 2.4. Добросовестность и профессионализм исследователей. В настоящее время публикаций об исследованиях достаточно много. Много материалов по маркетинговым исследованиям можно найти в Интернет. К сожалению, некоторые из таких публикаций не отличаются профессионализмом. Если не приведены точные формулировки вопросов анкеты27, не описано, как опрашивались респонденты, как обрабатывались данные, то это вызывает сильные сомнения в состоятельности исследователей. Иногда при внимательном изучении материалов становится видно, что анализ производился некорректно, а полученные результаты почти не имеют практического значения. 5.3.1.2. Форматы Имеются печатные источники, CD-ROM и сайты. В последнее время большую роль играют сайты. Электронный вид более удобный, так как можно проводить поиск информации. Печатный удобнее читается. Источники даны в списке. 1. Список источников – дают сайты университетов и международных организаций. 2. Обзоры по странам – организации помощи бизнесу. Но они слишком общие. В основном основные показатели, их динамика. 3. Платные отчеты (можно по заказу или готовые). По странам, отраслям. 4. Научная литература – довольно долго публикуется и отрывочная. 5. Правительственные данные. Довольно подробные, но для маркетинга недостаточные, это больше экономические данные. 6. Сайты исследовательских компаний. Больше социологические. РБК, Левада-центр. Более полезны для маркетинга. 7. Внутрифирменные данные. 7.1. Накладные, которые включают: вид товара (услуги), сведения о покупателе, объем продажи и условия продажи, цену и многое другое. 7.2. Другие документы: книга жалоб и предложений, кредитные записи, гарантийные карты, отчеты продавцов. 7.3. Отчеты о предыдущих исследованиях). 5.4. Первичные данные 5.4.1. Типы Если требуемую информацию не удалось найти среди вторичных данных и нельзя заказать стандартизированный отчет, то собирают первичные данные. Этот процесс достаточно сложен и требует отдельного рассмотрения. Первичные данные классифицируются по их содержанию. Демографические и социоэкономические данные28. Эти данные являются обычной основой для сегментации. Некоторые сведения точны (возраст, пол), другие – приблизительны (доход), третьи – неточны (социальный класс). Психология/стиль жизни. Среди данных этого типа важнейшим для маркетинга параметром является индивидуальность. Индивидуальность – характеристики поведения человека в нормальной ситуации. Чертами индивидуальности являются агрессивность, дружелюбие, коммуникабельность. В маркетинговых исследованиях эти данные важны, например, для определения типа магазина, который предпочтет покупатель. Покупатели могут быть разбиты по типу и особенностям поведения на следующие типы: • неактивные; • активные; • требовательные; • традиционные; • оригиналы; • ценовые; • транзитные. Требовательные покупатели, к примеру, предпочитают магазины с высоким уровнем сервиса, а для ценовых самым главным является цена. Транзитные покупатели еще не определили своего любимого типа магазина, поэтому представляют собой благодатную почву для рекламных усилий. Индивидуальность продавца (приветливость, открытость) также способствуют успеху как самого продавца, так и товара, который он продает. Отношение/мнение. Отношение выражается как предпочтения, взгляды, чувства. Мнение есть устное выражение отношения. Для маркетинга эти параметры представляют интерес в основном как предшествующие поведению (грубо говоря: нравится – купил, не нравится – не купил). Знакомство/знание. Здесь различаются: самостоятельное воспоминание; несамостоятельное воспоминание; узнавание: товара, черт товара, цены, места, где его можно купить, производителя, способов использования. Эти характеристики используются для измерения отношения к рекламе, продукту, магазину. Намерение. Это запланированное будущее поведение. Например, перед покупкой может существовать • определенное намерение купить; • неопределенное намерение купить; • отсутствие намерения; • определенное намерение не покупать. Этот параметр не может считаться важным, так как существует разница между высказанным желанием купить и реальной покупкой29. Параметр работает несколько лучше для дорогих товаров длительного пользования (например, автомобилей). Подразумевается, что большие траты обычно планируются более тщательно. Мотивация. Исследователи вкладывают в этот термин различный смысл. Один из подходов определения этого понятия – это все то, что отвечает на вопрос «почему?», то есть нужда, желание, побуждение, импульс, и т. д. Зная мотив, можно лучше узнать поведение и успешнее влиять на будущее поведение. В целом мотивы более стабильны, чем поведение, и поэтому лучше предсказывают будущее поведение, чем это делает прошлое поведение. Поведение. Это физическая активность объекта исследования при определенных обстоятельствах, в определенное время, включающая одного или нескольких участников. Типовая запись поведения при покупке включает пункты: что было куплено, сколько, где (тип магазина, его название, расположение, отдел...), когда, в какой ситуации, кто совершил покупку. Этот параметр важен для маркетинга, поскольку к поведению относится совершение покупки. 5.5. Основные методы сбора 5.5.1. Опрос Survey Именно к опросам практически всегда сводится общение респондента и интервьюера. Методы проведения опросов включают персональные интервью, телефонные опросы, почтовые анкеты. Используются и опросы смешанного типа, например, анкета вкладывается в упаковку товара или дается покупателю в магазине, а респонедент, заполнив ее, посылает затем по почте, чтобы участвовать в лотерее. Структурированность опроса – уровень предопределенности задаваемых вопросов и даваемых ответов. Высокоструктурированная анкета содержит строго заданные вопросы и предопределенный набор ответов на них. В неструктурированном интервью задана лишь тематика вопросов, а ответы имеют произвольную форму. Промежуточное положение занимает анкета, содержащая так называемые открытые вопросы: текст вопроса задан жестко, а ответ можно дать своими словами. По наличию маскировки целей опросы подразделяются на незамаскированный, когда респонденту сообщается цель опроса, и замаскированный, в котором цель не сообщается или дается легенда. 5.5.2. Наблюдение Помимо уже рассмотренных характеристик, присущих опросам – структурированности и замаскированности – наблюдения по способу организации подразделяются на натурные наблюдения и наблюдения со специальной подготовкой. Кроме того, бывают прямые и непрямые наблюдения. При прямых наблюдениях наблюдается та переменная, которая представляет интерес для исследователя, а при непрямых – какая-то другая переменная, с помощью которой можно путем вычислений определить интересующую исследователя величину. Например, если требуется узнать количество продаваемых в день в универсаме пакетов майонеза, то их подсчет на кассе будет прямым наблюдением, а определение остатка на утро, количества принесенных в зал пакетов и остатка на вечер с последующими расчетами количества проданных пакетов – непрямым. Наблюдения выполняются человеком-исследователем, иногда с использованием технических средств. 5.5.3. Их сравнение По разносторонности. При общении можно получить самую разнообразную информацию, надо только уметь спрашивать. Правда, точность ответа зависит от многих факторов. Это и содержание вопроса, и умение его задать, и личные качества интервьюера. При наблюдении можно определить только поведение в настоящий момент и некоторые социоэкономические или демографические характеристики (пол, возраст, косвенно – доход). При наблюдениях неизвестно, определяет ли отношение поведение или наоборот (может оказаться, что покупка сделана на пробу, отношение сформируется потом). По времени и стоимости. По этим параметрам предпочтительнее общение, так как при наблюдении приходится ожидать события, затрачивая время впустую. По объективности и точности. У наблюдения эти характеристики лучше, так как оно не зависит от нежелания или невозможности дать информацию. Дело в том, что вопросы бывают унизительные, шокирующие, требующие ответов, создающих неблагоприятное впечатление. Ответы на них вызывают затруднения. Кроме того, наблюдение не зависит от памяти респондентов, так как они склонны забывать неважные или «неудобные» для себя факты. Интервью – это всегда взаимодействие двух сторон, так что и опрашивающий влияет на респондента, и респондент может дать ответ, который, по его мнению, устроит интервьюера. *** Как для проведения опроса, так и для наблюдения необходима тренировка полевых работников, чтобы снизить влияние личного восприятия. 5.5.4. Методы опросов Основными способами проведения опроса являются: • персональные интервью (на улице, в магазине или на дому); • телефонный опрос; • анкетирование по почте (включая электронную почту и SMS); • онлайновые опросы через Интернет Особенности различных методов. • Интервьюеру можно указать последовательность задаваемых вопросов: если ответ на вопрос 4 положительный, то перейти к вопросу 10. Последовательность может быть достаточно сложной, что обеспечивает гибкость проведения интервью. При самостоятельном заполнении анкеты такие ветвления не могут быть слишком сложными. • Если анкета рассылается по почте, то она видна при заполнении вся и ответы влияют друг на друга. С другой стороны, почтовая анкета заполняется в индивидуальном темпе, то есть ответы даются продуманно. Но при этом нельзя уточнить неясные моменты. Если видно, что анкета анонимная, ответы становятся более откровенными. • При телефонных опросах легче устранить ошибку, вносимую интервьюером, так как на опрос влияет меньше параметров. Интервьюер знает, что его может прослушать начальство, поэтому работает более добросовестно. Практика показывает, что при телефонных опросах респонденты часто интересуются целями и организаторами этих опросов. При выборе вида проведения опроса следует учитывать также следующие параметры. Количество отказов. Важнейшим параметром исследования, определяющим его достоверность, является процент неответов. Как отмечалось выше, потери в ходе исследования могут внести ошибку в его результаты. Наименьший процент отказов в личном интервью. При телефонных опросах с первого раза удается связаться лишь с малым числом респондентов. Имеется и еще одна проблема – с кем разговаривать, если к телефону подошел кто-то из членов семьи30? При почтовых опросах процент ответов наименьший и нельзя гарантировать, что анкета будет заполнена именно тем человеком, на имя которого пришло письмо. Обычно такие анкеты заполняются коллективно всей семьей. Получаемая информация. В персональном интервью можно спросить практически о чем угодно. Почти так же обстоит дело и при телефонном опросе, однако в этом случае нельзя предъявить картинки и образцы. Хуже всего в этом отношении почтовые опросы. Они не могут быть слишком длинными, иначе их никто не заполнит до конца. Если уж не удалось избежать длинных опросов, то их лучше всего проводить путем персонального интервью, хуже всего – по телефону. Скорость опроса. Этот параметр значительно различается у разных способов опроса. При телефонных опросах можно выполнить 15…20 звонков за час, опрос ускоряется при увеличении числа интервьюеров. Скоростью почтовых опросов практически нельзя управлять. Время таких опросов составляет до нескольких недель (при охвате нескольких городов) и мало зависит от размеров выборки. Время персонального интервью на дому варьируется в зависимости от местожительства респондентов. Следует отметить, что при большом количестве полевых работников точность опроса оказывается низкой из-за их плохой подготовки и незаинересованности в результатах, так что увеличение числа опрошенных обычно не повышает, а снижает точность исследования. Стоимость. Почтовое интервью достаточно дешево, но при малом проценте ответов данные могут стать недостоверными. Стоимость телефонных опросов также достаточно низка, конечно, если нет междугородных звонков. Персональное интервью на дому – наиболее трудоемкий и дорогой метод исследований. 5.6. Выборки Объектом маркетинговых исследований может быть: Население или сегмент покупателей; Множество магазинов или сервисных центров и т.п. Такой объект состоит из элементов исследования. Это отдельные покупатели, отдельные магазины. Сплошное исследование всех элементов исследования обычно дорого; невозможно; неточно. (обсудить, почему). Поэтому исследуют лишь часть элементов объекта исследования. Такое исследование называется выборочным. Выборки бывают различного типа. Подробно они будут описаны ниже, а здесь приводятся некоторые соображения по различным типам выборки. Пожалуй, самая известная из теории – простая случайная выборка. В ней у каждого элемента имеется равный шанс быть выбранным для исследования. Выбор производится случайно или квазислучайно. Для этой выборки рассчитывается ошибка и достоверность. Существуют и более сложные случайные выборки. Однако у такой выборки имеется ряд недостатков: требуется для начала перечислить все элементы исследования, что вряд ли возможно для населения города, например; выбор производится случайно, поэтому в выборку попадают разные категории респондентов. Однако некоторые могут отказаться от интервью. Тогда возникает ошибка, так как не производится опрос определенных категорий респондентов. Требуется принять все маары, чтобы все-таки опросить тех, кто был выбран случайно. Поэтому используются детерминированные выборки, из которых наиболее простая – по удобству. Такая выборка бывает, например, при опросе в рамках телевизионного шоу: «Если Вы согласны с данным утверждением, позвоните по телефону…, если нет – по телефону …». Чтобы обеспечить репрезентативность детерминированной выборки (то есть соответствие ее параметров параметрам генеральной совокупности), использую квотную выборку, при которой в нее должны попасть разные типы респондентов в определенном количестве. Например, при опросе студентов университета опрашивать можно всех встреченных, но при этом должно быть опрошено, например, 30 девушек 1-го курса, 32 юноши 1 курса, 33 девушки и 33 юноши 2-го курса и т.д. Таким образом, обеспечивается охват всех важных категорий респондентов. 5.7. Сбор данных Исследования можно разбить на кабинетные и полевые. Кабинетные исследования, которые в настоящее время проводятся в Интернет, проводятся маркетологом как правило в одиночку. Для полевых исследований часто требуется привлечение интервьюеров. Это довольно сложно. Во-первых, интервьюеров требуется нанять. В России в начале 2000-х официально практически не было постоянной работы интервьюера, их нанимали только временно. На эту работу охотно идут студенты и пенсионеры. Второй вопрос – обучение интервьюеров. Интервью – процесс общения, при котором возможно возникновение множества ошибок и неточностей. Базовые характеристики респондентов и респондентов (пол, возраст, раса, национальность, религия, и т. д.) более целесообразно сделать как можно ближе, чтобы общение проходило легче. Дополнительные характеристики интервьюеров (стиль опроса, интонации, умение правильно записать ответ и т.д.) должны быть на должном уровне. Процесс сбора надо контролировать, чтобы как можно раньше устранить все недоработки. 5.8. Анализ данных Тут имеется огромное количество методов, они будут описаны применительно к международным исследованиям. Важно отметить здесь предварительные шаги: 1. Редактирование 2. Кодирование 3. Простая табуляция 4. Кросс-табуляция (Обсудить) 5.9. Подготовка отчета Существует ГОСТ на ответ о НИР. Есть методические указания. 5.10. Классификации исследований В конце краткого обзора маркетинговых исследований следует привести различные классификации. Это необходимо потому, что их имеется множество, они несколько противоречивы. Довольно логичным представляется приведенная выше классификация [Черчилль]: 1. Поисковые исследования 2. Описательные исследования 3. Исследования причинности 3.1. Экспериментальные 3.2. Неэксперимнетальные Другая классификация основана на различии количественных и качественных данных. 1. Qualitative Research - is generally undertaken to develop an initial understanding of the problem. It is non statistical in nature. It uses an inductive method, that is, data relevant to some topics are collected and grouped into appropriate meaningful categories. The explanations are emerged from the data itself. It is used in exploratory research design and descriptive research also. Qualitative data comes into a variety of forms like interview transcripts; documents, diaries and notes made while observing. There are two main methods for collecting Qualitative data 1.1. Direct Collection Method-When the data is collected directly, it makes use of disguised method. Purpose of data collection is not known. This method makes use of- 1.1.1. Focus Groups 1.1.2. Depth Interview 1.1.3. Case Study 1.2. b. Indirect Collection-Method 1.2.1. Projective Techniques 2. 2. Quantitative Research - quantifies the data and generalizes the results from the sample to the population. In Quantitative Research, data can be colleсted by two methods. 2.1. Survey Method 2.2. Observation Method Craig’s book of international Market research подразумевает (implies) that research is classified in the following types, which differ by methods of data collection. 1. Secondary data research. 2. Standard reports. 3. Survey data research. Here survey is a non-experimental, descriptive research method Survey is a systematic careful selection from a representative sample that provides valid and reliable information about the larger target population. Once sample data has been collected and analyzed, it is described in terms of the confidence that researcher's have in the validity and reliability of the data. The results are usually qualitative, as research is descriptive, but may also be quantitative? For example, most popular candidate for president or most spread opinion about brand. Another definition: A survey is a research method for collecting information from a selected group of people using standardized questionnaires or interviews. While many people think of a questionnaire as the “survey”, the questionnaire is just one part of the survey process. Surveys also require selecting populations for inclusion, pre-testing instruments, determining delivery methods, ensuring validity, and analyzing results. Types of Survey are on Figure …. 3.1. Cross-sectional 3.2. Longtitudal 3.2.1. Trend studies 3.2.2. Cohort studies 3.2.3. Panel studies 4. Nonsurvey data research. Here, according to previous definition, results are only qualitative. 5. Experimental research. It is now seen that different researchers have different approaches to research classification and if you do not find some information about definite method, it is not easy to find relevant information in another book. Next, it is convenient to consider the latter classification. 6. Использование вторичных данных в международных исследованиях In international market research secondary data is usually used for the following • Selecting different markets to evaluate for initial entry. • Estimating demand for a company’s products or services in international markets. • Assessing market interconnectedness to guide resource deployment across country markets or between and within regions. In the first case, secondary data can be used systematically to assess the market potential, risks and the likely costs of operating in different countries throughout the world. Countries that appear to be the most attractive can then be investigated in greater depth. Additional data will have to be collected to evaluate attractive countries more fully. This step may involve direct market contact. However, before the more expensive step of collecting primary data locally, secondary data can be used to develop more precise estimates of demand. As markets become more interrelated and links are established across national boundaries, firms need to understand the relationships between the different markets in which they conduct business. Secondary data can be used to understand the nature of these relationships. Through an understanding of market linkages, the firm is in a better position to develop a balanced portfolio of businesses and compete more effectively. 6.1.1. Выбор страны Надо определить страну и как выходить. По вторичным данным определяется привлекательность. Data have to be collected to assess the market potential and investment climate in all countries that are being considered, as well as the risks and costs of operating in these different environments. A major problem in the initial stages of international market entry is the bewildering array of countries and markets that can be entered. Since it is too time consuming to evaluate all possible countries in depth, the first step is to establish screening procedures to select those for further investigation. For firms that already have some international operations, screening procedures and criteria are already established and routinized. To begin with, management does not have to overcome the initial uncertainty associated with operations outside of the home market. More importantly, the experience gained by operating in different country environments provides additional information to assess new country environments. Countries that provide operating environments similar to those where the firm has already been successful are likely to be attractive prospects. In either case, secondary data are useful for this evaluation. They can be used to develop general procedures to categorize countries based on overall attractiveness or risk to suggest which countries should be eliminated from further consideration, and which should be investigated in more depth. Alternatively, secondary data can be incorporated into customized screening procedures that are geared to company objectives and specific industries. 6.1.1.1. Общие процедуры Группировка похожих стран по возможностям. Классификация стран Подход [of the Marketing Science Institute (1960-s, 1970-s)] основан на двух измерениях: - уровень демографической и экономической мобильности, измеряемый 21 переменной: индустриализация, маркетинговая ориентация, коммуникации, транспорт, организованность населения, образование, здравоохранение; - стабильность и сплоченность (измеряется 4 индикаторами: смертность от группового насилия, единство культуры, фрагментированность, срок существования национальной идентичности. Другой подход того же института. - делим страны на 5 уровней технологического развития: очень сильно развитые, развитые, «полуразвитые», «недоразвитые», «очень недоразвитые» (most highly developed, developed, semi-developed, underdeveloped, and very underdeveloped). Сходство внутри каждой группы измеряются 12 характеристиками: рост населения, городское население, религиозная и расовая гомогенность. Позже появились другие переменные, связанные с уровнем развития и ростом. Идея: в страны, близкую по культуре и физически легче войти. Отсюда культурное сходство Хофстеде. Метод определения сходства: многомерное шкалирование на основе различий по факторам (до уровня объяснений 95% различий). Еще измерения: стандарты жизни и стоимость жизни. Сейчас обычно: макроэкономический факторы+потреблеине продуктов Многофакторные индексы Economist Intelligence Unit (EIU) регулярно публикует информацию о бизнес-среде 60 стран. Основано на 70 индикаторах по 10 категориям: политика, макроэкономика, рыночные возможности, политика по конкуренции, политика по зарубежным инвестициям, управление внешней торговлей, налоги, рынок труда, инфраструктура. Michigan State University’s CIBER публикует индексы для 24 развивающихся стран. The World Economic Forum публикует свой Global Competitiveness Report. Специализированные модели Альтернатива – строить свои модели по вторичным данным, адаптированные под задачу. Некоторые компании уделяют большое внимание политическим рискам, другие – инфляции. Производителям снэков важно потребление еды людьми от 20 до 30 лет. Как создавать специализированные модели Многошаговый процесс. Сначала по общим показателям типа численности населения и GDP на душу. Климат: не надо продавать сноуборды в жаркие страны. Затем макроэкономика. Политические риски. Затем – по определенному рынку. Спрос, количество потребителей… Можно взвесить индикаторы по важности. Последний шаг – оценка затрат. 6.1.2. Оценка спроса Отобрали страны, теперь глубже изучаем оставшиеся. Надо смотреть то, что есть сейчас для начала, и прогноз – для длительного успеха. Если есть данные по продукту, строим тренд. Для новых продуктов или для фрагментированных рынков – опросы о покупательских намерениях. Если брать по аналогии с другими странами, то делаютcя допущения: о сходстве процессов, структуры рынка, сравнимости единиц измерения (валют). Должны быть сравнимы и категории продуктов. 6.1.2.1. Lead–Lag анализ То, что было раньше в одной стране, повторяется в другой. Продажи кондиционеров, телевизоров. Сейчас – интернетизация, мобильная реклама, мобильный банкинг. Допущение о сходстве процессов диффузии инноваций. Повтор, конечно, с определенным коэффициентом. И есть изменения: более новые процессы идут быстрее. 6.1.2.2. Shift–Share анализ Оцениваются изменения уровня занятости, скорости экономического роста, уровня производства. Собираем данные по многим странам. Рост прогнозируем исходя из среднего роста. А далее вводим поправки исходя из имеющихся отклонений от среднего. Если идет рост, то отклонения прогнозируются уменьшающимися и наоборот. 6.1.2.3. Барометрический анализ Допущение: есть единообразная зависимость в разных странах. Обобщенные барометры Годится для базовых продуктов: бумаги, сахара, цемента. Часто связано с GDP. Строим регрессию. Видно, что не очень хорошо. Делим на группы. 12 стан, где >=$10000 и 53, где меньше. Гораздо лучше! Еще есть показатель промышленного производства: сколько производится упаковки. Другие модели: уровень урбанизации и жд пассажирокилометры – для предсказания уровня телефонизации. Бывает нелинейная зависимость (логарифмическая). Тогда это похоже на эластичность! Годится не для всех стран. Можно делать такой анализ по сегментам. Анализ когорт Когорта: группа, испытавшая некоторое событие в определенный временной интервал. Чаще всего – родившиеся в определенное десятилетие. Классическая ошибка: прогноз потребления бевереджей. 1979 изучали. Видно, что с возрастом уменьшается. Населеине стареет, значит, будет спад. Но если посмотреть на когорты (по диагонали!), видно, что те же группы пьют больше! 6.1.2.4. Аналитические модели Сколько будет студентов? 1. Количество на 10000 чел. Численность населения Прогноз: число на 10000* численность 6.1.2.5. Разные методы В некоторых странах работает одно, в других – другое. Если требуется точность, надо использовать различные источники данных. Но многие данные основываются на оценках UN или World Bank. 6.1.2.6. Оценка взаимозависимости рынков Рынки сейчас связаны, надо строить общие оценки. Связь - географически, макроэкономически (смотрим тесноту связи и схожесть рынков); - по продуктам на одном рынке. Связь, если есть общие конкуренты, общие покупатели, общие дистрибуторы, общие медиа. Действия на одном рынке влияют на другой. Связь рынков a – 11,2% германского импорта и экспорта – с Францией b – 17,9% французского импорта и экспорта – с Германией. Для отдельных продуктов надо еще смотреть рынки по товарам. 6.1.3. Точность и эквивалентность международных данных Особенно важно для развивающихся рынков. Даже для общих показателей. GDP может включать доходы, полученные в других странах. Иногда расчеты производятся редко. Механизм сбора данных бывает неточным (население неграмотно). Бизнес-статистика зависит от структуры налогообложения (скрывают). Разница в данных: тв в Германии = развлечение, в США = мебель. Городское население: Япония : shi (city) где 50 000 чел. В Индии – где 5000, в Ирландии - где 1500, в Канаде – 1000, Исландия – 200. Социальные программы. Есть страны с большими налогами и большой соц программой (Швеция). В результате там более занижается GNP, покупательная способность. 6.1.3.1. Источники информации о конкурентах Помимо упомянутых выше наблюдений, исследования специально купленных товаров конкурентов, пробных заказов услуг, опросов потребителей и продавцов, существует целый ряд легальных источников данных о конкурентах. Они приведены на Интернет-странице http://www. it2b.ru/print2.view3.page207.html. К числу федеральных информационных ресурсов, открытых для свободного доступа, относятся: Единый государственный реестр юридических лиц и Единый государственный реестр индивидуальных предпринимателей, ведущиеся в соответствии с Федеральным законом «О государственной регистрации юридических лиц и индивидуальных предпринимателей» и постановлениями Правительства РФ. Единый государственный реестр налогоплательщиков содержит огромный объем информации, но доступно из него немногое. Государственный земельный кадастр. Единый государственный реестр прав на недвижимое имущество и сделок с ним. Сведения о лицензиях, выданных юридическому лицу или индивидуальному предпринимателю, могут быть получены не только из Единого государственного реестра юридических лиц и Единого государственного реестра индивидуальных предпринимателей, но и из реестров лицензий, которые ведут лицензирующие органы. Немало информации может быть собрано через частных детективов. Но это всегда дорогие услуги. В печатных средствах массовой информации тоже можно найти много ценной информации о конкурентах, включая рекламную, но этот способ весьма трудоемок. Облегчить сбор сведений может обращение к бюро вырезок, профессионально оказывающему услуги по сбору вырезок из газет и журналов. Информация о судебной практике конкурента может найтись в правовых базах данных Кодекс, Гарант, КонсультантПлюс. Но обычно эта информация сводится к постановлениям Высшего арбитражного суда РФ и федеральных арбитражных судов. Возможно, в скором будущем появится еще один важный источник информации – кредитные бюро. За рубежом кредитные бюро существуют уже давно и доступ к информации многих из них не ограничен банками. 6.2. Стандартизированные отчеты Удобным способом заказа маркетинговых исследований является заказ на стандартизированный отчет. Данные таких исследований занимают промежуточное положение между первичными и вторичными данными. Они являются первичными по содержанию, так как делаются на заказ, но вторичными по форме, так как выполняются по стандартизированным формам и методикам. Такие исследования достаточно дешевы. Исследования подобного типа могут включать: 1. Сегментирование рынка с определением демографических характеристик покупателей по данным переписи или результатам выборочного опроса. 2. Определение объема продаж и доли рынка. Исследования проводятся по большим представительным спискам. Для них определяются такие параметры, как размер рынка, объем покупок, приверженность марке, частота покупок, эффект изменения стратегии рекламной кампании. 3. Аудит магазинов с проверкой движения товаров, накладных оптовиков, динамики общего объема продаж, закупок розничными торговцами, периодичности поступления товара, оптовых и розничных цен и пр. 4. Телефонный опрос «омнибус», когда каждый заказчик может включить в опрос выборки респондентов свой вопрос. Заказы на стандартизированные исследования часто выполняются с применением технических средств, специально разработанных для этой цели. Ниже приводятся основные виды стандартизированных исследований. 6.3. Сбор первичных данных 6.3.1. Определение элемента исследования Это задает параметры исследования. Fig. Страна берется часто: есть данные, и собирать легче (язык хотя бы). Сравнение тоже обычное. Если есть специфика, выбираем другой размер. Часто схожесть в регионе (среди студентов). Мир как элемент исследования: поведение молодежи и элиты по миру. Этнография – более узко, чем страна. Другое измерение: принадлежность. (membership). По социодемографическим параметрам, например. Пример: сравнение интересов и покупательских моделей детей от 5 о 11 в городских семьях среднего достатка в Европе и Азии. Везде проблемы. Возраст: разный, в Азии 1 год при рождении. Городское население – см. выше. Средний класс – смутно! Средний доход – сложно! Для фирм: 1) по размеру, 2) по тем, кого опрашивать в организации. Макроуровень – развивающиеся, переходные, развитые экономики. Микроуровень – сельское и городской население, место исследования: на работе, в магазине. 6.3.2. Проекты исследования 6.3.2.1. В одном месте Все-таки, есть неявное сравнение с собственной страной. Часто это этнографические исследования. Делают обычно местные исследователи. Или местные все подготавливают. 6.3.2.2. В разных местах Это – наиболее часто и в коммерческих, и в академических исследованиях. Изучается разница в отношении или поведении. Например, элитрые потребители по миру – изучались в 71 стране. Иногда изучается влияние определенного фактора. Например, роль мам в дарении подарков в англо-кельтской, китайско-вьетнамской и израильской культурах. Самое важное – исключить влияние посторонних факторов. Что относится к макрокультуре, что – к социодемографии, а что – собственно к изучаемому фактору. Например, потребительские паттерны нью-йоркцев еврейского происхождения чем обусловлены? Страной происхождения, проживанием в Нью-Йорке, … На рис. Еврейские потребители похожих социоэкономических параметров и занятости в Париже, Милане и Лондоне (1,2,3) сравниваются с еврейскими потребителями в Нью-Йорке и Чикаго (4 и 5). Западное наследие в сравнении с американским. Можно проследить еще и происхождение: из Польши, из России. Ситуационный факторы: тинэйджеры могут быть глобальными в музыке, досуге и отдыхе, но не в еде. То же и с праздниками. Т.о, может быть принадлежность к разным группам. Католики-швейцарцы-французы-тинэйджеры. Разная сила принадлежности. 6.3.2.3. Изучение внешних влияний Например, эффект от визита или проживания в другой стране. Или непрямое влияние – через масс-медиа, кино, … Для изучения прямого влияния подходят длительные исследования. На рис. Изучается влияние. Серые – съездили, белые – не съездили в другую страну. Проблема!!! Съездили просто более активные!!! Надо исключить это влияние. Бывает и более сложное влияние (следующий рис). 6.3.2.4. Адаптация Более сложный случай – следующий рисунок. С какой скоросью ассимилируются иммигранты из Марокко (macro unit J) и Вьетнама (macro culture L) во французскую макрокультуру (K). Факторов много. + связи с местными группами, изучение языка. Замедляется, если создаются нацгруппы, нацинституты, инфраструктуру. Fig Важно учесть созревание. Старение происходит одновременно с адаптацией. С возрастом люди более консервативны, этноцентричны. Изучать надо когорты в одной социокультурной группе, кросс-секциональным исследованием (поперечное сечение). Если проводить длительные исследования, можно изучить изменения каждого индивида. Например, какие изменения от приезда молодых иммигрантов до образования ими семьи и хозяйства. Результат – получить данные для бенчмаркинга разных групп. Этот тип проектов дает больше всего информации. 6.3.3. Культурное смещение. Проект исследования, коммуникации и интерпретации Еще один фактор – сами исследователи. Бывает, что люди одной культуры проводят исследования в другой. Отсюда 3 типа ошибок. Дизайн, коммуникации и интерпретация с точки зрения своей культуры. См. рис. Полезен опыт работы в аналогичных странах. Опыт из Японии полезен в Корее. 6.3.3.1. Проект исследования Есть нюансы. Что значит определенное поведение? Как выражается отношение? Надо знать ценности и паттерны поведения, которые лежат в основе процесса покупки и использования товара. Тогда можно оценить, как сработают маркетинговые действия. Если этого не учитывать, не будут заданы нужные вопросы. Надо не забывать о чувствительности результатов к социокультурной среде. 6.3.3.2. Коммуникации Между представителями разных культур. Если интервьюируют местные, им трудно объяснить, что требуется. Респонденты тем более могут не понять вопросов от иностранцев. Трудно оценить, правильно ли провели исследование. 6.3.3.3. Интерпретация Можно не понять собранные данные, особенно если речь об отношении или поведении. Пример: респондент низко оценил важность инноваций. Значит ли это, что он не купит новый товар? Как снизить культурную ошибку? 1. Определить проблему в терминах родной культуры. 2. Определить ту же проблему в терминах изучаемой культуры. Пока не делать выводов! 3. Изолировать self-referent cultural (SRC) влияние (влияние своей культуры) на проблему и определить, насколько этьо усложняет проблему. 4. Переопределить проблему без влияния SRC и решить ее для иностранного рынка. Самое сложное – шаг 2. Требуется знать нюансы другой культуры. Альтернатива – работать с иностранными исследователями, с ранних стадий исследования. Тогда можно выявить специфически-национальные концепции. Хотя это бывает и дольше, и сложнее. К сожалению, кооперация начинается тогда, когда исследование уже спланировано, готова анкета! А иностранцы привлекаются только для проведения исследования. Возможно сильное pseudo-etic смещение. Как убрать SRС? Обычно при опросах используют те атрибуты, которые хорошо использовались в другой стране. Может быть, некоторые не важны, а важных - нет. Полезно использовать безатрибутные карты восприятия. См. практическое задание. Так что культурные различия – важный фактор! 6.4. Сопоставимость данных из разных стран Сравнимость конструкций и их измерения. Данные должны иметь одно и то же значение и интерпретацию, одинаковую точность, валидность и надежность для всех стран и культур. Сравнимость в различных аспектах. Это важно и для исследователя, и для принимающего решения менеджера (стратегия чаще разрабатывается глобально, а не по отдельным странам). Решения (в какую страну, какой товар подавать, какому сегменту) тоже требуют сравнимости. Даже при исследовании одной страны – скорее всего понадобится изучать и другую потом. Провели тест маркетинг товара в одной стране, успешно. Почему бы в другую не продавать? Но для решений надо сравнимые данные получить. 6.4.1. Дилемма Emic/Etic Есть уникальные паттерны поведения, ценности, паттерны отношения. Giri (Япония), послушание и уважение. Другие существуют в ряде стран, но по-разному. Агрессивность где-то – борьба, где-то - крик. Важно определить, насколько концентрироваться на уникальных чертах (тогда сложно сравнивать), а насколько – на общих (тогде потеряем точность). 6.4.1.1. Подходы Emic и Etic Получились из phonemic и phonetic, имею греческие корни. Emic - как люди сами понимают и описывают свое поведение. Etic – что люди говорят в свете общих социальных теорий. Отсюда и две школы исследований. Emic изучают специфичное, а сравнения – качественные или суждениями. Etic строят глобальные концепции отношения и поведения и разрабатывают культурно-независимые измерительные инструменты. Сравнение облегчается, но есть проблемы. Главная - ‘pseudo-etic’, при которой свой взгляд считается глобальным. Это два полюса. Для исследователя важнее все-таки etic, так как ищется сходство между странами. Etic подход Основа – прошлые исследования. Чаще идет из США. Там резработаны шкалы отношения, например, этноцентризм (см. ниже). Бывает, что не делается даже попыток адаптировать к другой стране. Это и есть ‘pseudo-etic’ феномен. Часто проверяется надежность инструмента, например, с помощью альфы Кронбаха (определяет состоятельность теста для измерения психологических атрибутов). Обычно считается, что мера и, соответственно, конструкция, одинакова, если одинакова ее структура. Так, если остались те же вопросы, характеризующие психологическую характеристику, то считается, что результаты эквивалентны. (см. практическое задание по тестированию предпринимательских способностей). Но тесты дают разные результаты раз от раза. Emic подход Исследования проводятся местными исследователями в каждой культуре. Иногда применяется этнографический подход, тут возможно обобщение между странами. Но обычно концепции уникальны. Так что обобщение сложно. Значит, надо как-то обобщить методики, чтобы взять плюсы от двух. Рис. Fig. Сначала проводится исследование в свое стране, потом с этим подходом идут в другую страну. Потом проводится emic исследование в другой стране изнутри. Результаты сравниваются. Сравнивать можно олько в случае, если есть что-то общее. Сложно определить то, что схоже. 6.4.1.2. Схема кросс-культурных исследований Надо обобщить. Есть 2 спаособа. Адаптированная etic модель Допущение: теория панкультурна, но требует адаптации. 1. Формулируем базовую теорию и набор концепций. Например: лидеры мнений, которые хорошо информированы, передают информацию и мнения населению и влияют на результаты голосования. Но в Швеции обнаружилось, что есть еще и горизонтальное влияние. Еще пример: развивающиеся общества и развитые. В развитых рынках есть неудовлетворенность от широты выбора. В развивающихся обществах – еще нет. 2. Изучаем связанные с теорией концепции и гипотезы. Тут либо проводим исследования в другой стране, либо используем результаты других исследователей. Если исследователи готовились интернационально, они дадут худшие результаты, чем чисто местные исследователи. По-другому. Явным образом строим альтернативные инструменты измерения концепций. 1. Считаем, что теория etic. 2. Методы измерений разрабатываются для каждой страны (emic измерения). Модель акцентируется на сходстве. Связанная emic модель Гибридная или связанная emic модель. 1. Несколько исследователей из разных стран проводят исследования в у себя. 2. Ищется набор одинаковых параметров и ключевых вопросов. Сложно, так как могут видеть по-разному. Особая сложность - терминология. Обычно английский – общий язык. Но перевод может быть сложным. Что такое лидерство, доверие, предпринимательские способности и т.д. 3. Достигнув консенсуса, каждый планирует исследование. Будет своя специфика. Все концепции должны стать измеримыми (operationalized). 4. Сходства и различия локализованных моделей обсуждаются. В результате получается: 1) общие концепции, 2) общие, но по-разному измеряемые, 3) различные взаимосвязи концепций, 4) уникальные концепции. 5. Гармонизируем. Либо будет общий список концепций, либо что-то комбинированное, применимое везде. Можно включить и все концепции, где-то они важны, где-то – нет. Могут получиться модели для регионов. 6. Проводим исследования везде на основе общей теории. Могуд появиться новые гипотезы. И т.д. Тут акцент на особенности. Может быть долго, но мала вероятность pseudo-etic ошибки. 6.4.2. Эквивалентность данных Изучалась эквивалентность концепций. Но. 6.4.2.1. Эквивалентность конструкций Из трех частей. Функциональная эквивалентность Концепции, объекты и поведение должны играть одинаковую роль (выполнять функцию) во всех странах. US: велосипеды = отдых, как теннис, гольф, Нидерланды и Китай – основной вид транспорта (как автобусы). Занятия. Некоторые страны спорт = регулярно, для здоровья. В других = досуг, развлечение. Покупки бакалеи. Развитые страны – мало времени – заказ через Интернет. А где-то ходят пообщаться с соседями, слушают советы продавца – важный момент жизни. Важность тоже различается. Автомобиль – роскошь или средство передвижения? В Африке если у парня есть велосипед и радио – он крут. Концептуальная эквивалентность Смысл, вкладываемый в объект, стимул или поведение. Тут главные споры – а можно ли сделать паннациональные концепции. Сохранить лицо ‘saving face’ (Китай). Philotimo (Греция) вести себя так, как ожидают члены группы (семья, друзья. Логом (Швеция) – без роскоши. Запад – индивидуализм, Восток - коллективизм. Ритуалы. Похороны в Индии многолюдны, на Запада чаще приватны. Эквивалентность категорий Beverage. Soft drink: баночные или бутилированные сода, минеральная вода (тоже различается в России и за рубежом), фруктовые соки, холодный чай, концентраты соков. В Нидерландах + молоко, так как запивают им. А где-то это детское питание или добавка к кофе. Семейное положение. Есть страны, где много жен, а иногда много мужей. Занятие. US attorney или British barrister. Важность. Китай – учитель = отец = король. В России - ? Каков статус у чиновников?. 6.4.3. Эквивалентность измерений Надо различать измеряемую концепцию и собственно процедуру измерения. Эквивалентность калибровки Валюта, вес, расстояния. Тут ясно, есть формулы и таблицы. Но и тут есть сложность при определении темпов роста: надо брать одинаковые периоды. Есть много других вербальных и невербальных стимулов: цвет, форма. Японские дети используют меньше цветов, чем австралийские и греческие. Школа влияет? Качество продукта – стандарты национальные. Эквивалентность перевода Задача: чтобы одинаково поняли. Исток: есть 2 вопроса. Они про одно и то же или нет? Что считать критерием эквивалентности и какой уровень должен быть. Невербальные стимулы вызывают ассоциации разные в разных странах. Метрическая эквивалентность 1. Шкала. Англоговорящие: 5 или 7 пунктов в шкале, Россия: 5, Где-то 10 или 20, есть 12. 2. Уровень ответов. Берем готовность к покупке. Обычно это 2 верхние позиции шкалы. Это одно и то же в разных странах? На Ближнем Востоке и на юге Европы ставят более высокие оценки. 6.4.4. Валидность конструкций Обычно конструкция имеет несколько измерений. Вопрос: как эти измерения связаны? Валидность конвергенции. Если измерить разными методами, в разных странах, будет ли одно и то же? Проверяется корреляция, должны быть высокая, тогда есть подозрение на валидность. Если в двух странах есть, но разная, подозрение уменьшается. Дискриминантна валидность. Отличается ли одна концепция от другой? Нужна низкая корреляция во всех странах. Номологическая валидность. Как конструкция связана с внешним критерием. Например, есть валидность CETSCALE, имеющая корреляцию между отношением, намерением купить и владением иностранными продуктами. Тут рассматривалась корреляция. Но используетс я также факторный анализ и структурные уравнения. 6.4.5. Надежность конструкций Надежность измеряется по-разному. 1. Те же результаты получаются при повторных измерениях. 2. Разные люди, чаще эксперты, дают сходные оценки. 3. Измерения имеют внутреннюю состоятельность (см ниже). 1 и 2 дороги, чаще используется 3. 6.4.5.1. Состоятельность во времени Повторный тест. Через 2…5 недель. Если меньше, помнят, если больше, изменяется отношение и намерение. General Perceived Self-Efficacy шкала имеет надежность .67 через 6 месяцев. В других исследованиях: через год от .75 до .55. В международных исследованиях дороговато. 6.4.5.2. Состоятельность по индивидам 2 или более судей, экспертов одни и те же оценки дают или нет? Например, если используются открытые вопросы, то как назначить категории ответов? Проблемы могут быть при измерении отношения и стиля жизни, поведения. Используется коэффициент ранговой корреляции Спирмана, Коэф Кэндала. См. презентацию. Собрали данные от мужей и жен в 5 странах: US и UK( англо) France, Belgium and French Canada (франко). Измерялось: - Демографические характеристики и соцэкономические - Самоотчет о поведении в виде рейтинга участия в деятельнсоти и решениях в семье - переменные психографики и стиля жизни Сравнили ответы мужей и жен. Надежность была выше в «жестких» переменных, чем в «мягких» (стиль жизни). Доход, возраст = согласие. Участие в жизни семьи – разница. Это тоже дорого, так как требуется собрать двойные данные от семей. 6.4.5.3. Внутренняя состоятельность шкал Альфа Кронбаха. Мера, с которой вопросы – про одно и то же. Низкое значение = плохо, высокое = хорошо, но для какой конструкции? Обычно начинается с факторного анализа для снижения размерности, а потом уже считается альфа. Если альфа <0,6 это плохо. Считается также корреляция ответа на вопрос к общему счету. Но эта валидность еще не означает двух предыдущих! Еще способ: деление пополам. Например, суммы ответов на четные и нечетные вопросы должны совпадать. Но это если много вопросов, от 16. В международных исследованиях еще и по странам сравниваются. 6.4.6. Эквивалентность процедуры сбора данных Инструмент, выборка, процедура опроса. Инструмент. Отвечают сами (через Интернет? Гибко, в удобное время) или записывает интервьюер (если образование низкое)? Слова (нужен перевод!) или образы (тоже нужен перевод)? Выборка. Выбираем домохозяйства или организации. Кого спрашивать? В арабских странах мужья играют большую роль при покупке продуктов. Доходы теперь от мужи и от жены, роли меняются. Если разные ответы, надо спрашивать больше. Список избирателей, телефонов. Если нет, то выборка кварталов (гнездовая). Нужен баланс сравнимости и репрезентативности. Например, выборка точно такого же возраста и дохода может быть нерепрезентативной в разных культурных контекстах. Процедура. Где-то лучше телефон. В других странах лучше личное интервью. Обычно самозаполнение анкет лучше, так как нет недопонимания при общении. 7. Сбор данных без опросов (Nonsurvey Data Collection Techniques) 7.1. Качественные исследования 7.2. Наблюдение и квазинаблюдение 7.2.1. Чистое 7.2.2. Изучение следов активности Посещение сайтов, чеки 7.2.3. Архивные исследования Данные о продажах, газеты, отчеты. 7.2.4. Протоколы ‘Think out loud’ 7.2.5. Проективные техники 7.2.5.1. Коллажи 7.2.5.2. Аналогии и метафоры 7.2.5.3. Завершение рисунков 7.2.5.4. Psycho-drawing Опишите бренд цветом. 7.2.5.5. Персонализация Бренд = персона. 7.2.6. Глубинные интервью 7.2.6.1. Индивидуальные 7.2.6.2. Фокус-группы, в т.ч. электронные 7.2.6.3. Расширенные креативные группы До 4 часов. Выявляются чувства. 8. Проектирование анкет Используются для сбора количественных данных. Инструмент важен, важно, чтобы собирались правильные данные в разных странах. Есть много источников ошибок. 8.1. Информация и вопросы Для начала: какую информацию собирать: - демография; - специально по продукту (использование и оценка); - специфично по стилю жизни, поведению и отношению (изучается разница). Например, в Индии и Китае большие семьи, как это влияет на потребительское и покупательское поведение? Одинаков ли смысл вопросов? 8.1.1. Формулировка вопросов Общие вопросы по сегментации проще, чем про отношение. 8.1.1.1. Базовые и демографические характеристики Пол, возраст – несложно. Доход, образование, род занятий, условия проживания – сложнее. Семейное положение. Стало сложно. ESOMAR рекомендует: (1) married/living together; женаты/живут вместе; (2) single; одиноки (3) widowed/divorced/separated; вдовы, разведенные, разделенные. Живут вместе. Немного различается ‘living together’, vivant comme mariй, leben mit einem Partner zusamme). Доход. У многих 2 работы. Говорят об основной. Фермеры не упоминают бартер, а только деньги. Не упоминают 13 зарплату. Более правильно делить зарплату по персентилям. Образование. Россия: 11 лет в школе, 4 года бакалавр, 2 – магистр. Европа: 12 лет школа, 3 года бакалавр, 2 года магистр. Болонский процесс – попытка унифицировать. Спрашивать лучше не уровень школы, а количество лет обучения. Занияте. lawyers in the US, solicitors and barristers in the UK advocats in France немного различаются. ESOMAR рекомендует: • A,B руководство и профессионалы • C1 образованные работники умственного труда и ВКР • C2 квалифицированные рабочие и служащие • D, E неквалифицированные работники физического труда и прочие. Жилье. Запад: квартиры или дома или секции таунхаусов; Африка – хижины; Ближний Восток – жилые комплексы: много комнат, выходят во двор, там живут расширенные семьи. Отдельные помещения для мужа, жены, детей, бабушек-дедушек, теть-дядь. 8.1.1.2. Поведение и продукт Зависит от социо-эконо-культурных особенностей. Если много супермаркетов, сетевых и дискаунтеров, то хотят побыстрее и подешевле. Там, где много мелких лавочек с ограниченным сааортиментом, слушают продавца, общаются. Нужны вообще разные анкеты! ?????Comparability in Purchase, Consumption, Usage and Disposal Behavior Since purchase, consumption, usage and disposal behavior are an integral part of day-to-day living, they are deeply embedded in the fabric of society and affected by sociocultural norms, cultural conventions and so on. As a result, the sociocultural context in which purchase and consumption take place, as well as the behavioral acts and processes that result in purchase decisions, often vary from one country or culture to another. This leads to significant differences in need states, the purchasing process, purchase occasions and the fit between products or specific product attributes and need states as well as the salience of different product benefits. Each culture, society or social group has its own particular conventions, rituals and practices relating to behavior on social occasions, such as entertaining family or friends, or behavior on festive occasions, such as marriage, graduation, Christmas or other cultural festivals. Rules relating to the exchange of gifts and products are, for example, governed by local cultural conventions (Levi-Strauss, 1965; Mauss, 1954). Thus, while in some cultures wine may be an appropriate gift for a dinner host, in others flowers are preferred. Japan, for example, has a unique system of gift giving, which plays a key role in maintaining the social structure (Fields, 1989). Gifts are given on two occasions during the year: midsummer (sheiben) and in the New Year. Gifts must be carefully matched to the status of the recipient relative to that of the giver, and appropriately wrapped. Consequently, in comparing gift-giving practices, questions relating to gift giving and positioning of products as gifts will need to be tailored to specific practices and behavior patterns. Similarly, attitudes with regard to the importance of different types of behavior vary from one culture to another. For example, among the middle classes in many industrialized countries cleanliness is considered next to godliness. Considerable importance is attached to activities and products that promote cleanliness, such as household cleaning products that keep the house spick-and-span and smelling fresh, or personal hygiene products. Frequently, advertisements for antiperspirants, deodorants, toothpaste or mouthwash promise the purchaser instant social success, or warn of the dangers of social ostracism without their use. In other countries, less attention may be paid to personal hygiene, clothes may be washed less frequently, and body odors or bad breath may not be considered offensive. The type of questions relating to product benefits and attributes asked in surveys of products such as household cleaning products or personal hygiene will therefore need to be tailored to the specific cultural context. The way in which purchases are made may also vary from one country or culture to another. In the US, for example, purchasing on credit and the use of debit cards reduces the need to carry cash. In developing countries in Africa and Asia, consumer credit is typically limited and credit cards are rare or nonexistent. In some countries, notably throughout Europe, use of debit cards is widespread and reduces the need to carry substantial amounts of cash. In some countries use of a mobile phone to pay for items such as soft drinks in vending machines has also been introduced. Use of prepaid cards for transportation and telephone systems is also widespread in many industrialized countries. Consequently these factors will need to be taken into consideration in formulating questions relating to different modes of payments, payment systems or services that rely on such forms of payment. Differences may also exist with regard to product disposal from one country to another due to differences in the strength of the environmental movement and environmental regulation. For example, in the UK bottles and cans are taken to bottle and can banks for recycling, and paper is separated from other garbage. In other countries a deposit is paid to the retailer for glass and plastic bottles, and in some cases aluminum cans. This is refunded on return of the bottle or can. The size of the deposit can vary from 5 cents up, depending on the size of the incentive that regulators wish to establish. Regulation on recycling of other items such as tires and batteries and on littering also varies. Again, these are factors that need to be taken into consideration in formulating questions relating to packaging and so on. Comparability in Product Class Boundaries In addition to such differences in usage, purchase and consumption behavior, competing and substitute products often vary from one country to another. For example, washing machines and other household appliances may be competing with domestic help and professional launderers, as well as with other brands of washing machines. In many Latin American countries in middle-class families domestic help does the washing, although the rising cost of domestic help has encouraged purchase of washing machines by wealthier families. Similarly, in Northern India dhobis, men who traditionally did the washing of middle- and upper-class families, have moved to better-paying employment opportunities in hotels, restaurants, office cleaning and so on. Consequently, purchase of small washing machines by the middle classes in India is on the rise. The range and type of items contained in a product class may also vary. For example, there are significant differences in the type of soft drinks available in different countries, and also in what is considered a soft drink. In the US, soft drinks consist predominantly of different varieties of colas (cherry, diet, caffeine free), lemon-lime sodas, ginger ale, iced teas, mineral waters, alternative beverages (mostly fruit-flavored sodas) and sports and energy drinks. In the Netherlands, milk is frequently consumed as a beverage at lunch and hence is included in the soft drink category. In other countries, fruit juices (apple, orange and grape) are popular as well as concentrates such as blackcurrant, peppermint or anise, which are then diluted. In Asia, freshly squeezed fruit juices, including mango, papaya and pineapple, are also popular soft drinks, and in some South East Asian countries, soybean milk flavored with honey, chocolate or strawberry is widely available. The appropriate definition of the product category and the product variants included will thus vary substantially from one country or region to another. In addition to the lack of comparability with regard to product class boundaries, competing product set and the type of products included or available within a specific product class, differences may be encountered with regard to brand availability or even the existence of branding. Some product categories tend to be dominated worldwide by large multinational companies such as Coca-Cola and Pepsi in colas, Kellogg’s and General Mills in cereal and Gillette and Schick in razors. Generally they compete with local and regional brands, though the strength and significance of these brands may vary from one country or region to another. For example Cott, a Canadian cola company, produces colas for private-label vendors in countries such as the UK, Germany and the Netherlands. In India, a local cola brand ‘Thums Up’, acquired by Coca-Cola through its acquisition of Parle, a local bottler, is extremely popular. In Iran, a locally produced cola called Zam Zam is popular and has been exported to Saudi Arabia and other Gulf countries as well as Denmark. In the razor market, the degree of local competition varies considerably from one country to another and local brands compete predominantly in the price-sensitive segment of the market. Again, awareness of local brands and relevant product attributes to examine is important in designing an effective questionnaire. In developing countries and in the former socialist countries, branding of consumer goods has only recently emerged as an important factor. In many former Eastern bloc countries, consumer choice was extremely limited. Consequently, consumers are not accustomed to making finely discriminating comparisons among brands and products. As a result, brands are often broadly categorized as ‘Western’ or local. Similarly, in developing countries, the market for consumer packaged or pre-processed goods is often limited to more affluent consumers or dual-income households. Such items have to compete with fresh products or food prepared in the home or by local merchants. For example, pasta or noodles may be made in the home, purchased freshly prepared from small merchants or itinerant vendors, or purchased pre-packaged from the store. Such differences in purchase, consumption and disposal behavior and in the comparability of product categories from one country or culture to another mean that careful attention needs to be paid to such distinctions in designing a questionnaire. Often this implies that some desk and qualitative research is needed to identify relevant factors, especially where management has limited knowledge or experience with attitude and usage behavior in the country. It is also important if the study is designed to assess reactions or purchase intentions relating to a product or service that is new to the country and could potentially stimulate substantial changes in existing consumption and purchase behavior. Attitudinal, Psychographic and Lifestyle Data The most significant problems in drawing up questions in multicountry research are likely to occur in relation to attitudinal, psychographic and lifestyle data. Here, as noted in Chapter 6, it is not always clear whether certain attitudinal or personality constructs – such as aggressiveness, respect for authority and honor – are equally relevant or equivalent in all countries and cultures. Even where similar constructs exist, the same question or attitude statement may not tap them most effectively. There is considerable discussion with regard to this issue and the best way to deal with it. This is examined in more detail in Chapter 10. Attitudinal, psychographic and lifestyle measures can be examined at two different levels: (1) general constructs, values or long-run orientations that hold across all product categories or areas of life; and (2) domain- or category-specific constructs that apply to specific product domains or life interests, for example food, clothing, sports or leisure activities. In developing measures of both general and domain-specific constructs, the relevant domain of the construct needs first to be specified in each country. General constructs might include personality constructs such as sociability or innovativeness, values such as materialism or self-achievement, lifestyle patterns relating to leisure behavior, attitudes toward work or family life, gender or shopping, and use of credit. Category-specific constructs, on the other hand, might relate to a specific product category such as detergents, automobiles or children’s clothing. In each case, the domain of the construct and how it is manifested in each country needs to be specified by examining qualitative data or through focus groups, consumer workshops and so on. Next, the specific items that best tap the various constructs in each country and culture should be identified. Even where similar constructs are identified in different countries, the specific items making up these constructs may not always be identical. In some cases, the same constructs may be measured by somewhat different items. For example, materialism might be measured in the US by a statement such as: ‘The things I own say a lot about how well I’m doing in life’. In France, the following statement might be more appropriate: ‘I like a lot of luxury in my life’ (Dubois and Laurent, 1993). Interest in identifying similar lifestyle segments on a regional or worldwide basis has been particularly marked among advertising agencies and market research companies. Euromonitor has, for example, recently profiled consumer lifestyles worldwide based on food and clothing habits, entertainment, education, sports pastimes, shopping habits, media and tourism (Euromonitor, 2004). It is, however, important to establish the link between the lifestyle segment and preferences or purchase behavior relative to a given product category or life interest. Consequently, use of attitudinal or lifestyle characteristics to profile cross-national segments identified on another basis, for example demographics, has been more typical. For instance, the Coca-Cola study of teenagers profiled differences and similarities of teenagers’ attitudes, values and lifestyles in different regions throughout the world. Studies examining domain-specific segments typically collect data relating to domain-specific attitudes and values, usage behavior and purchase criteria and then cluster respondents across countries based on these variables. One study of food cultures in Europe identified 12 different food cultures, seven of which were national, four transnational and one regional (Askegaard and Madsen, 1995). Another study of financial services in Europe (Bijmolt et al., 2004) grouped consumers based on attitudes and usage of financial services and then developed profiles of these segments based on financial service usage patterns. The degree of commonality may also depend on the specific countries or cultures and on the nature of the topic. There is likely to be more commonality among consumers in relatively similar countries, such as the industrialized Western nations, than between Westernized nations and emerging market countries. Allowance should, however, always be made for the identification of country- or culture-unique concepts, and also idiosyncratic measures of these, as well as pan-cultural concepts and measures, particularly where there are significant economic or cultural differences between countries. 8.1.1.3. Type of Question Another point to be considered is the form in which questions are asked. Questions can be either closed or open-ended. Closed questions require the respondent to reply according to a specific format and select from various alternative responses. Open-ended questions, on the other hand, allow the respondent freedom to provide his or her own response, without constraining the range of options. Similarly, questions may be posed directly, or indirectly so that the purpose of the question is not apparent to the respondent. Open-ended versus Closed Questions The most compelling argument in favor of the use of closed questions is that they facilitate analysis. Responses can be pre-coded and entered directly into a computer from the questionnaire. Closed questions also make it easier for interviewers to record responses directly into a laptop, or for respondents to record their response on a computer screen at home or in a shopping mall survey. On the other hand, closed questions mean that the researcher must specify in advance all relevant response categories. This may sometimes be difficult in cross-national research, especially where the researcher lacks extensive experience or familiarity with purchasing behavior or relevant determinants of response in another country or cultural context. Open-ended questions may be preferable in a number of situations. Since they do not impose any structure or response categories on respondents, they avoid the imposition of cultural bias by the researcher. Particularly where respondents fill in responses on a home computer, they may tend to provide lengthier, more expansive responses. In addition, they do not require the researcher to specify all possible responses. On the other hand, they do entail the somewhat tedious process of establishing coding schemes for responses, and tabulating them once the data have been collected. The levels of literacy or education of respondents also affect the appropriateness of using open-ended questions as opposed to closed questions. Open-ended questions are often used when a researcher lacks knowledge about factors underlying behavior or attitudes in the country or countries studied. Since a respondent has to respond to open-ended questions in his or her own terms, they require some sophistication and knowledge of the topic on the part of the respondent, otherwise responses will not be meaningful. Consequently, open-ended questions have to be used with care in cross-cultural and cross-national research where respondents have low levels of education. Open-ended questions are often appropriate in exploratory research, especially where the objective is to identify relevant dimensions, concepts or terminology associated with the problem studied. They might be used to elicit content domains relating to products, attitudes toward products or advertising stimuli, or the associations evoked by various stimuli. For example, in the case of products such as beverages, respondents might be asked to list all the items they perceive as beverages, and the most frequently occurring responses could then be used as the relevant product set. Similarly, in a study of attitudes toward products or advertising stimuli, subjects could be asked to indicate adjectives, words or phrases that best describe or characterize relevant stimuli such as packages or advertisements. This task might then be repeated using different scenarios, such as for family consumption, when entertaining for special occasions and so on. Direct versus Indirect Questions Another consideration is whether direct or indirect questions are utilized. Direct questions avoid any ambiguity concerning question intent and meaning. On the other hand, respondents may be reluctant or unable to answer certain types of questions. In addition, they may tend to provide responses perceived as socially desirable or those that they feel are desired by the interviewer. Use of indirect questions can help to avoid such biases. In this case, rather than stating the question directly, it is posed in an indirect form. For example, respondents might not be asked their own preferences, but rather the response they anticipate that most respondents, their neighbors or other relevant reference groups would give. This approach may be desirable in certain cultures, for example Japan where there are substantial pressures toward conformity. Alternatively, respondents might be presented with different types of purchase decisions or shopping scenarios, and asked which most closely correspond to their own. Another approach, especially useful where respondents may have difficulty recalling decisions or behavior, is to ask a series of questions leading up to the purchase decision or behavior. For example, to find out about brand loyalty respondents might be asked first how frequently they shop for groceries, where they typically shop for groceries and what brand they last purchased in particular product categories. This helps the respondent to recall the situation leading up to the actual purchase and increases the likelihood of an accurate response. Irrespective of the way in which questions are formulated, they need to be adequately pre-tested on an appropriate sample before being administered. While pre-testing is important in domestic research, it is crucial in international markets, due to potential problems of misunderstanding and miscommunication. Successive rounds of testing may be needed in order to ensure that sources of response bias are minimized and that respondents fully understand questions. This is critical if accurate responses and high item response rates are to be obtained. 8.2. Use of Nonverbal Stimuli Another important consideration in instrument design is whether respondents are shown nonverbal (as opposed to or in addition to verbal) stimuli in order to help them understand and respond accurately. Particularly where research is conducted in countries or cultures with high levels of illiteracy, such as Africa and the Middle East, it is desirable to use nonverbal stimuli. Questionnaires can be administered orally by an interviewer, but respondent comprehension will be facilitated if pictures of products or test packs are provided. Various types of nonverbal stimuli may be used in conjunction with questionnaires, including photographs, show cards, product samples or pictures. Nonverbal stimuli are also often used in conjunction with other data collection techniques, for example projective techniques, consumer workshops or synectic groups, as discussed in Chapter 7. Here the discussion is focused on the use of nonverbal stimuli in surveys to help respondents understand verbal questions, products and product concepts, or to express feelings. Show cards such as that in Figure 8.1 can, for example, be used to assist in answering product usage questions. This set was designed to aid consumers in understanding the different types of uses that might be made of a sewing machine. Sketches of products can also be employed to illustrate a questionnaire. In Figure 8.2 the questionnaire used in a study on health in an aboriginal community in Australia (Spark, 1999) was illustrated with sketches of outdoor plumbing, a washing machine and a refrigerator. This not only ensured that respondents clearly understood the question and the particular product covered in the survey, but also attracted their interest. Illustrations may be useful in rural areas when it is not certain whether the respondent has been exposed to the product and knows what it looks like or what functions it performs. In general, pictures or sketches need to be simple, so that respondents can understand them easily. Product samples can also be shown to respondents. A drawback of this is that respondents tend to become irritated if the samples are removed for the next interview. Consequently, it is advisable to leave a free sample or other reward for each respondent. Visual stimuli may also be employed to develop rich images and probe consumer associations with products and brands. In international research, they can have the advantage that they are not as directly affected by cultural and linguistic factors as verbal stimuli. In a study of young consumers’ attitudes toward global brands in Sгo Paulo, Mexico City and Venezuela, three types of visual stimuli – colors, geometric forms and photos – were used to examine consumer brand associations (Troiano, Costa and Guardado, 2002). Cards with 15 colors grouped into four categories – basic, intense/fashionable, cosmetic and undefined – were shown to consumers as well as 19 cards with geometric forms belonging to four categories – straight lines, regular forms, rhythmic lines and open forms – and 11 photos reflecting different atmospheres, such as lightning, a beach, an art gallery and skiers. The study revealed substantial similarities in the associations across the three cities, reflecting similar underlying values and visual brand identities. Even where literacy levels are high, use of visual stimuli in conjunction with verbal stimuli helps to provide a check on instrument equivalence and to identify potential biases from questionnaire translation or adaptation of the questionnaire to different linguistic and cultural contexts. The sample can be split, and half the respondents asked questions without the visual stimuli, while the other half are shown the visual stimuli. The results obtained from the two halves can then be compared to see whether there are differences. It is important to recognize that pictorial stimuli are not culture free, as perceptual associations and their interpretation differ from one country or culture to another. However, the types of biases arising from visual stimuli are likely to differ from those occurring in a verbal instrument. Consequently, comparison of the two procedures will indicate whether there is a need for further testing and instrument development. Figure 8.1 Interview show card used in consumer survey in South Africa (assists to answer question ‘What is your sewing machine used for?’) Figure 8.2 Questionnaire with sketches used in a study of Aboriginal health Source: Spark, 1999. Once the basic form of the questionnaire or research instrument to be used in the survey has been drawn up, the next step is to translate it so as to ensure that it is clearly understood by respondents, and to try to avoid any possible miscommunication. The issues involved in translating verbal and nonverbal instruments, and some standard translation procedures, are next discussed in more detail. 8.3. Instrument Translation Both verbal and nonverbal instruments need to be translated so that they can be used in different linguistic and cultural contexts. While the need for translation of verbal instruments is widely recognized, and examples of errors arising from mistranslation abound, the need for translation of visual stimuli is generally less well recognized. It is, nonetheless, important to realize that visual stimuli are not necessarily universal or pan-cultural. Consequently, if visual stimuli are developed in relation to a specific cultural context, the same misinterpretation and miscommunication problems can arise as in relation to verbal stimuli. Consequently, attention to how visual stimuli are perceived and interpreted in other contexts is required. 8.3.1. Verbal Translation In translating a questionnaire or verbal instrument, two principal methods have been identified in the educational and measurement literature: forward translation and back translation (Hambleton, 1993, 1994). In either case, a number of procedures, judgmental and statistical, can be used to evaluate the equivalence and quality of the translations. More recently, the problems arising from back translation, and particularly in detecting errors using this method, have led to advocacy of a committee approach, involving multiple individuals in the translation (Harkness, 2003). 8.3.1.1. Forward Translation In the case of forward translation, a single translator or group of translators prepares a translation from the source language into the target language. Several versions can be prepared and then compared. This approach is subject to the risk that the translation may contain errors and not accurately represent the meaning of the original or ‘source’ questionnaire. Difficulties may be encountered in finding equivalent words or phrases in the target language, resulting in errors. For example, an item relating to the president or head of state in a country will not have the same meaning in a country with a prime minister. Equally, the Spanish word paloma is equivalent to both ‘dove’ and ‘pigeon’ in English and amigo in Spanish does not always have the same meaning as the English word ‘friend’. Similarly, in a survey of health issues among Vietnamese immigrants in Australia (Small et al., 1999), in translating the phrase ‘I have been so unhappy I have difficulty in sleeping’ into Vietnamese, the term used for ‘unhappy’ incorporated the concept of irritable, which is not implied in English. However, this was felt more suitable than the word usually used to translate unhappy, as this was not considered strong enough to prevent sleep. While problems can arise because words do not match up across languages, more fundamental problems can occur when concepts do not match across cultures (Harkness, 2003). One example is differences in grammatical gender. Some languages such as Spanish and French have elaborate systems of grammatical gender. Others such as English have simple systems, and others such as Hungarian have none. English is typically gender neutral, leaving the sex of terms such as doctor, friend or secretary unidentified. In German, however, references to all of these are gender specific, requiring two forms – for example Freund/Freundin, Arzt/Arztin – to be used when a question may refer to either. Pronouns and adjectives have to agree accordingly. Comparability problems can arise in multilingual studies. For example, the German term for secretary is typically feminine, eine Sekretarin. The masculine form ein Sekretar either means a desk or a man with a senior executive position, requiring use of a different descriptor. While attention has been drawn to the need for comparable question wording in different languages, less concern has been exhibited in relation to translation of response scales. Yet this also can pose problems. For example, difficulties may be encountered in translating response categories such as often, sometimes, rarely and never. In Turkish, the equivalents of the terms ‘rarely’ and ‘never’ are frequently used interchangeably. Consequently, in one study few respondents picked the term corresponding to rarely, as both terms were understood as ‘not at all’ or ‘never’. Even where response scales are constructed specifically for a study, attention needs to be paid to comparability across languages. Often the scales traditionally used in each language in which the questionnaire is being developed are adopted, without regard to cross-language comparability. For example in the Eurobarometer, which measures social and political attitudes in the EU on a regular basis, the French and English scales differ in structure and to a lesser extent in semantics. Both use the semantic dimension of agree, but in the French scale the use of d’accord, pas d’accord suggests a unipolar scale, while the English scale uses a bipolar construction in which the wording is linguistically symmetrical with the endpoints modified by ‘strongly’. Equally, the ‘don’t know’ category is ‘can’t choose’ in English compared with ne sais pas or ‘don’t know’ in French. Such issues make it important to ensure that translators are knowledgeable about the subject matter and terms used, or the meaning of the source questionnaire is easily lost in translation. For example, the term ‘item pool’ was translated into Japanese as ‘item ocean’. Such errors are more likely to arise when translating linguistically dissimilar languages such as English to Chinese or Japanese, as compared with English to French or French to Italian. One means of checking this is to administer the translated questionnaire to respondents and then probe to ask about the meaning of each item and their responses. Respondents are asked how they interpret a given question and to suggest alternative phrasing or wording that might tap the same issue. However, this approach can be somewhat time consuming and cumbersome, and is only feasible where respondents have time available to discuss questions. 8.3.1.2. Back Translation The back translation method has traditionally been widely used in educational testing and psychological measurement (Brislin, 1980; Hambleton, 1993, 1994). It is also extensively used by both academics and marketing research companies. Following this procedure, a questionnaire is translated from the initial or source language by a bilingual translator who is a native speaker of the target language into which the translation is being made. This version is translated back into the original or source language by a bilingual who is a native speaker of that language. The two versions are then compared in the source language to check for errors and the quality of the translation. This approach is useful in identifying translation errors and the competency of the translator, but is subject to a number of problems (Brislin, 1980). Bilinguals often develop a particular language structure and usage. As a result, they do not always translate into the idioms commonly used by most people and may employ language or terms that are difficult for respondents to understand. Bilinguals tend to have standard rules for translating nonequivalent terms, for example always translating amigo as ‘friend’, which do not always capture the intended meaning. Since they know both languages, they may be able to make sense out of poor translations and do a good job of back translation, so that grammar and syntax errors in the target translation go undetected. Back translation starts with the assumption of a ‘source’ language and the evaluation of the adequacy of the translation is made in the source language. Consequently, there is always a question of the extent to which the structure and terms used in that language will dominate the questionnaire. One approach to alleviating this problem is known as ‘decentering’. This entails modifying both source and target questionnaire through successive iterations of translation and retranslation to eliminate the dominance of the source language (Triandis, 1972; Werner and Campbell, 1970). At the end of the process, both source and translated questionnaire are modified so that terminology is equally well understood and equivalent in each language context. While this procedure is likely to result in the best translation, it is time consuming and tedious. A further problem is that few international projects have the resources to effectively ‘decenter’ questionnaires. Many studies employ materials developed in a source language, which are then back translated by bilinguals, where the survey is to be conducted. The source materials are typically viewed as fixed or providing a base line, and little decentering occurs for fear of deviating too far from this base. Decentering is extremely difficult if not virtually impossible where there are three or more language versions (Bontempo, 1993). As a result, translators typically try to reproduce the source version as faithfully as possible in the target language. However, a totally faithful or literal translation may not always be desired in marketing research, especially where idiomatic or colloquial phrases are used. Not only are some phrases and terms difficult to render in other languages, but also in some instances it may be desirable to translate into equivalent colloquial phrases. Back translation will not necessarily achieve this. An idiomatic phrase can be faithfully translated and back translated without capturing the accurate meaning in the translated language. For example the phrase Das Leben in vollen Zugen geneissen in German might be literally translated into English as ‘Enjoy life in full trains’ and correctly back translated (Harkness, 2003). This would not, however, result in the English translation with the equivalent colloquial meaning, ‘Live life to the full’ or ‘Live life to the fullest’ (American/English). Similarly, just as errors can occur in forward translation, they can occur in back translation. For example, when ‘I have been so unhappy that I have difficulty sleeping’ was back translated from Tagalog, the back translation read ‘I became so lonely that I find it difficult to sleep’, which implies a very different meaning from ‘so unhappy’ in the original English. Further investigation suggested that the Tagalog might more accurately be rendered as ‘I felt so despondent that I had difficulty sleeping’, which was in fact very close to the original English. This example illustrates some of the difficulties associated with back translation. While on the surface it is a relatively simple procedure, further examination and discussion are typically required to avoid errors in back translation and ensure an accurate rendering of the term to be translated in the target language. 8.3.1.3. Committee Translation The limitations of back translation suggest the desirability of adopting a committee approach (Harkness, 2003). This can be organized either by parallel translation or split translation. In the case of parallel translation, several translators make independent parallel translations of the same questionnaire. A meeting is then held to review the translations and agree on a final version. An alternative approach, which saves time and effort especially if the questionnaire is long, is split translation. In this case, at least two translators and a reviewer are needed. The translation is then divided up between the translators in an alternating fashion. Each translator translates his or her own section and a meeting is held with the reviewer to examine the translations and agree on a joint version. Here, it is important to ensure that consistency is maintained across the translation and that translation of a similar phrase is harmonized. Once translated, whether by the parallel or split approach, the translation is reviewed by a committee, including the translators, reviewer and an adjucator, and a decision is made whether to accept the revised version. While somewhat time consuming, this approach may be particularly appropriate if questionnaires are complex or different expertise is needed for individual parts, for example different language versions. In countries that have multiple languages, such as Belgium, India, Canada, Switzerland and South Africa, or where different immigrant groups are to be interviewed, separate versions of the questionnaire or research instrument typically need to be developed. Similarly, if several dialects of a language are used, it may be desirable to treat each as a separate language. Even if the immigrant subgroup can understand the host-country language, translation into the local idiom will enhance willingness to respond and help to provide complete and accurate answers. This is particularly likely to be a factor with older and less well-educated respondents. Similarly, where different versions of the source language are used, adaptation to the appropriate version will be needed. For example, questionnaires originating in the UK or the US will need to be adapted for use in Australia, New Zealand or South Africa. Similarly, questionnaires in Spanish will need to be adapted to the different versions of Spanish spoken in various Latin American countries as well as within the US (Harkness, 2003). 8.3.1.4. Assessing Translation Equivalence Regardless of the specific translation procedure utilized, it is important to verify the quality of the translation and to assess the equivalence of various versions prior to use in the field. Here, two different approaches have been suggested. One is heuristic, based on testing various versions of a questionnaire on monolingual and bilingual subjects. The other applies statistical analysis based on item response theory. Various procedures for evaluating translation equivalence using both mono- and bilingual subjects have been suggested (Hambleton, 1994). These include: (1) evaluation by monolinguals of the source and target questionnaires based on clarity and comprehension of the items; (2) evaluation by bilinguals of both questionnaires based on possible meaning errors; (3) comparison of results obtained from administering the source and back-translated questionnaires to additional subjects; and (4) comparison of results obtained from administering both original and translated versions to bilingual subjects. The last two procedures are considered the most reliable insofar as the primary objective is to ensure that both the original and translated questionnaires result in the same pattern of response. 8.3.2. Translation of Nonverbal Stimuli In addition to translating verbal instruments, it is important to recognize that visual instruments and stimuli will also require translation and testing for miscommunication or misinterpretation in different countries and cultural contexts. This can arise as a result of differences in the interpretation of perceptual cues in different countries and cultural contexts, as well as the associations evoked by visual phenomena and objects. For example, not all African cultures recognize Western conventions using narrowing parallel lines or fainter colors to represent distance. In general, pictorial stimuli, especially where they are not supported by verbal stimuli, tend to generate a broad range of associations. Consequently, care is needed to ensure that the stimuli are interpreted unambiguously. Once translated, field testing of data collection instruments is necessary in order to identify any final problems and test for appropriate colloquial usage relative to the target respondents and so on. Many examples can be cited where a supposedly good translation of an instrument had to be revised as a result of a field test. This further underscores the importance of care in translation and use of multiple checks to ensure that errors not found in one stage are remedied in another. Where both verbal and visual instruments are used, comparison of results obtained with each method can provide a check on overall results, and problems arising due to miscomprehension be identified. 8.3.3. Potential Sources of Bias Associated with the Research Instrument No matter how carefully the instruments have been designed and how well they have been translated to avoid potential problems of miscommunication, some difficulties are likely to arise due to certain sources of bias characteristic of cross-national surveys. The major sources of bias in both international and domestic marketing research are the respondent, the nature of the topic, and the respondent’s interaction with the interviewer and with the instrument. Seven specific types of bias are particularly likely to occur in international research. These are: (1) yea-saying bias or tendency to respond in the affirmative; (2) extreme response bias or a tendency to use the extreme points of a scale; (3) social desirability bias or the desire to provide the socially acceptable response; (4) the nature of the topic being studied; (5) item nonresponse or failure to respond to certain types of questions such as income and education; (6) specific respondent characteristics; and (7) the response format. These are highly interrelated, for example yea-saying bias, extreme response style and respondent characteristics are all closely linked, but each gives rise to specific problems in cross-national research due to differences in the nature and the importance of the effects within and between countries. In the social sciences emphasis is placed on item nonresponse and social desirability as major sources of bias, while in marketing studies different types of response bias or response styles have received more attention (Baumgartner and Steenkamp, 2001; van Herk, 2003). Examination of five different types of response styles – acquiescent and disacquiescent styles, extreme response styles, midpoint responding and noncontingent responding – on data from 11 EU countries, has, for example, shown these to constitute an important source of bias that can distort correlations between survey items (Baumgartner and Steenkamp, 2001). This effect occurred systematically across all countries. Careful attention to the bias introduced by such response styles is, therefore, needed to assess the extent to which this may contaminate surveys. 8.3.3.1. Social Acquiescence or Yea-saying Bias Social acquiescence or yea-saying bias, i.e. the tendency to provide a positive response or check the highest agreement response categories in rating scales varies from country to country (Skjak and Harkness, 2003). It may stem from positive or optimistic attitudes in some countries and cultures or alternatively from the desire to be socially acquiescent, and provide the response perceived as desired by the interviewer or the sponsor of the study. It is particularly likely to occur on agree/ disagree items and ones that offer clear affirm/reject response (Smith, 2003) and also on 5–6- or 7-point Likert scales where the item is not worded in a balanced or neutral way, but reflects a positive or negative attitude. Yea-saying is common in Asia, where cultural values often lead the respondent to try to avoid distressing, disappointing or offending the interviewer in any way. At the extreme, this can lead to a tendency to agree to any assertion, irrespective of the respondent’s actual position. Japanese respondents, for example, always respond in the affirmative, as it would be considered extremely impolite to respond negatively. In addition, affirmative responses may be given where the respondent does not understand the question, but does not want to be impolite or display ignorance by not responding. In Europe differences occur between countries and regions within countries. Greeks, both men and women, have been consistently found to provide more positive responses (Baumgartner and Steenkamp, 2001; van Herk, 2003). A detailed study of yea-saying bias conducted in relation to both domain-specific and product-specific attitudes on topics such as cooking, washing and shaving as well as in values (using the LOVscale) consistently found Greeks to be most prone to extreme yea-saying bias (van Herk, 2003). Italians and Spanish were the next most prone to yea-saying bias, followed by UK, German and French subjects. At the same time looking at regions within countries in Europe, the further south a region, based on its location and its geographical coordinates, the more likely a respondent was to provide a positive response (van Herk, 2000). This suggests that climate and geographical location affect product-related attitudes (similarly to the way they affect mental health). In some instances the effects of yea-saying bias can be reduced by concealing sponsorship of the study and by more effective training of interviewers. Careful wording of questions in order to render them more neutral, for example by presenting a choice between two equally strong bipolar oppos-ites rather than using a 5–9-point Likert agree/disagree scale (or making Likert statements more neutral rather than positive or negative), may help to reduce this bias (Smith, 2003). Introduction of balanced Likert scales ensuring that half the response categories agree and half disagree can also be helpful. Translation may cause problems here, as captions for points on the scale may, as noted earlier, translate differently in different languages. 8.3.3.2. Extreme Response Bias Another source of bias is extreme response style. Extreme response bias is the tendency to use the extreme ends of the scale. This again has been found to vary by country (van Herk, 2003). In large-scale surveys of cooking, washing and shaving in six countries in Europe, as in relation to yea-saying bias, Greeks were found to have the most extreme response style. Italians and Spanish had the next most marked extreme response style, followed by the French, the British and lastly the Germans. The Germans were the least prone to extreme response style of any EU subjects, suggesting a more rational, balanced approach to responding to questions. Again, the further south a respondent came from in terms of region within the EU and the global coordinates of each region, the more extreme the response style (van Herk, 2000). This suggests a possible link between climate, hours of sunshine and extreme response style, as no differences were found on the East–West dimension. 8.3.3.3. Social Desirability Bias A source of bias closely related to social acquiescence or yea-saying is social desirability. This is bias relating to the topic studied and has been extensively examined in the social sciences where many topics, for example health education and life satisfaction, may reflect specific social norms and values (Johnson and Van der Vijver, 2003). Social desirability bias may also be triggered by interaction with the interviewer, and varies with the individual and cultural background. Individuals from lower-status backgrounds are likely to defer to the values of higher-status interviewers. Social desirability typically shows less impact with self-completion questions than when face-to-face interviews are used. Similarly, individuals from more affluent societies show low social desirability scores, while social desirability tends to be higher in collectivist societies and particularly among Asians (Johnson and van de Vijver, 2003). Responses may be intended not only to please the interviewer, but also to reflect behavior perceived as socially acceptable or normal in the respondent’s culture. For example, a respondent may say that he or she purchases or uses a product such as toothpaste or deodorant regularly, whether or not this is the case. Equally, questions relating to voting behavior, alcohol consumption or substance abuse are particularly likely to be susceptible to social desirability bias. Such biases can be reduced by making it easier to provide a socially nonacceptable response. Questions might be prefaced with phrases such as: ‘Some people feel this way, some people feel that way. How do you feel?’ Questions can also be matched with equally socially desirable responses, and respondents asked to choose the response that they think best describes themselves. Training interviewers to be neutral helps considerably in this regard as well as using interviewers of similar status or background to respondents. Increased use of observational techniques, videotaping of actual consumption and shopping behavior, and consumer workshops where consumers interact and discuss attitudes and reactions provides another means to identify and also avoid such sources of bias. In these methods, the interviewer is removed from direct interaction with the interviewee and hence the desire to please the interviewer has less impact on interviewee response. Elimination or rewording of items likely to be subject to social desirability biases can also help to ensure that response styles do not create spurious constructs (Baumgartner and Steenkamp, 2001). Equally, the randomized response approach can be used with sensitive questions (Reinmuth and Geurts, 1975). Alternatively, scale scores can be purified by calculating indices of all the response styles. Respondent scores are then regressed on these indices and analysis conducted on the residualized scores. Corrected covariance and correlation matrices based on purified scale scores can be used as input to structural equation models. 8.3.3.4. Topic Bias Some topics are more socially sensitive in some countries. Willingness to respond to questions such as level of income, or on topics such as sex or alcohol consumption, vary from one country or culture to another. In the Scandinavian countries respondents are considerably more willing to admit to and discuss alcohol consumption or use of contraception than in other countries. In Arab countries anything related to sex tends to be a taboo topic, and even personal hygiene can be sensitive. As a result, researchers need to identify what topics are socially sensitive in each country and cultural context. In addition, use of other research techniques such as collecting observational data, using projective techniques or developing improved interviewer probing techniques also needs to be considered. 8.3.3.5. Question Order and Scale Frequency Effects The order in which questions are placed will also affect response, as well as the recency and frequency of the behavior to which a question relates. Again, differences have been found across cultures both in response to question order and frequency estimation. Since Eastern cultures foster an interdependent perspective of self and emphasize relations with others and fitting in, one may expect Asians to be more knowledgeable about their own and others’ behavior. Westerners, on the other hand, view the self as different from others and favor an individualist perspective. Consequently, they are likely to be less attentive to their own and others’ behavior. Although publicly observable behavior occurs with similar frequency in both cultures, US students have, for example, been found to report high frequency for public observable behavior (i.e. visiting a library) when provided with high-frequency rather than low-frequency scales, while Chinese students reports were unaffected by the type of scale (Schwarz, 2003). Both US and Chinese students reported higher frequencies for private nonobservable behavior (i.e. having a nightmare) on scales where the response frequencies were higher. Differences have also been found in contextual effects between Asian and Western respondents that in turn have an impact on question order effects (Schwarz, 2003). East Asian cultures put a premium on maintaining harmony in social relationships and prefer indirect forms of communication that require reading between the lines (Markus and Kitayama, 1991). Asian respondents are, for example, more sensitive to the conversational context of questions, which in turn affects question order. In a study with Chinese and German students about life satisfaction and academic satisfaction, Chinese students were more likely to generalize from general to academic satisfaction questions, when questions were presented in that order rather than the reverse. Germans, on the other hand, were more inclined to generalize from academic to general life satisfaction when academic satisfaction questions were presented first. This suggests that careful attention is needed to the effects of question order and that the order in which questions are presented should be varied in order to minimize such effects. 8.3.3.6. Respondent Characteristics Certain types of respondents are particularly prone to yea-saying, extreme response and social desirability bias. Age, education and to a lesser extent gender are all factors frequently found to be related to response bias (Greenleaf, 1992). The less well-educated, those of lower socioeconomic status and women are all more likely to be prone to response style bias. Consequently, responses are likely to be subject to bias in lower-income countries where respondents also tend to have lower levels of education than in industrialized nations. Item nonresponse has also been found to be related to similar factors, namely gender, age and education, in different countries in Europe (Douglas and Shoemaker, 1981). In Chile, acquiescence biases have been found to be more common among less-educated respondents, suggesting that there may also be a relationship between tendencies toward acquiescence and socioeconomic status. Examination of the antecedents of yea-saying bias and extreme response style in six European countries – France, Germany, Greece, Italy, Spain and the UK – consistently found yea-saying and extreme response style to be negatively related to education and household income and positively related to age, if over 20 (van Herk, 2003). For gender there were conflicting results, but this may relate to the nature of the studies conducted, since two on cooking and washing were with home-makers and one on shaving was with males only. These findings imply that the distribution of different national samples on variables such as gender, income and education should be examined, to identify the extent to which such factors are likely to affect results. In some cases, apparent differences between nationally representative samples may reflect the impact of sample characteristics rather than true national or cultural differences. This is further discussed in relation to comparability vs representativity of samples in Chapter 9. 8.3.3.7. Item Nonresponse Item nonresponse is nonresponse that occurs through respondents not answering questions either deliberately or unintentionally. This is an important source of bias in cross-national surveys, as countries differ in rates of item nonresponse. Overall survey rates have been declining internationally for some years along with item response rates (Couper and de Leeuw, 2003). The decline has been highest in certain countries such as the Netherlands and the UK and is a cause for significant concern with regard to the quality of survey response. Cultures vary with regard to their willingness to talk and respond to questions (Lonner and Berry, 1986) and their interest, involvement and information regarding a particular topic. This will affect rates of nonresponse to different questionnaire items. Members of some cultures are more willing to be interviewed and to respond to all types of questions. Others exhibit greater reticence and tend to have high rates of nonresponse to all types of questions (Lonner and Berry, 1986). In Malaysia, the Chinese have been found to be more reticent than Malaysians or Indians in answering questions, giving a higher proportion of ‘no’ or ‘don’t know’ answers and fewer responses to open-ended questions. Item nonresponse also varies depending on the mode of data collection, both within and across countries (Skjak and Harkness, 2003). Item nonresponse has been found to be higher in self-completion formats, particularly mail questionnaires, than where the interviewer completes the survey. Respondents may inadvertently skip questions or fail to complete the questionnaire. If a large number of questions are unanswered, it may be preferable to eliminate the respondent from the survey. Poor design of response formats or inadequate instructions in self-completion questionnaires can also affect item non-response (Skjak and Harkness, 2003). Respondents will make more mistakes if questionnaires are poorly designed. Respondents may, for example, fail to answer questions or overlook items especially where these are filter questions, unless these are very clearly laid out. Inadequate instructions can also result in errors, especially among less-educated respondents. Poor design and poor instruction are especially likely to result in item nonresponse in emerging markets or in cultures where respondents and interviewers are less familiar with answering surveys. Willingness to respond to certain types of questions, such as questions relating to income or age or that are culturally sensitive, also varies from country to country. In the US there is a lower tendency to fail to respond to questions about personal income than in Europe (Young, 1999). A study of nonresponse to different items in a public opinion survey in eight European countries also found variation (Douglas and Shoemaker, 1981) in nonresponse to questions about income. Nonresponse to political questions was highest in Germany and Italy. Cultural differences also exist in the confidence with which people know or decide they have the ‘right’ not to answer (Skjak and Harkness, 2003). If respondents are reluctant to answer questions but are uncertain about the social acceptability of not answering, they may select the ‘don’t know’ category instead. In cultures where they are aware that they cannot respond, they will do so. Similarly respondents who do not want to appear ignorant will not respond rather than ticking ‘don’t know’. Thus both use of the ‘don’t know’ category and nonresponse can depend on the culture. Rates of nonresponse are highest in relation to the same types of questions irrespective of country or culture. As might be expected, rates of nonresponse are typically lowest with regard to background characteristics, such as sex or education; moderate with regard to behavioral variables; and highest in relation to complex attitudinal and opinion questions (Douglas and Shoemaker, 1981). This suggests the need for careful attention to wording of questions to ensure that they can be clearly understood by respondents. The impact of item nonresponse in international marketing research depends largely on the factors underlying nonresponse in different countries or cultures. This can be examined by assessing rates of nonresponse for various items across countries. Characteristics of nonrespondents to these items can then be compared with those of the overall sample, to check for differences. In general, however, the same factors – that is, sex, age and education – appear to be related to nonresponse and hence do not significantly affect comparability of results. A number of strategies may be used to reduce item nonresponse. First, threatening, monotonous, unclear or ambiguous items should be eliminated or reworded to gain respondent interest. Second, response can be increased by improved formulation of questions in a way that engages or involves the respondent. On the other hand, care is needed to ensure that where closed questions are employed, opinions are not elicited where they do not exist (Mitchell, 1965). 8.3.3.8. Response Format In designing questionnaires for use in different environmental and cultural contexts, an important issue concerns the adaptation of response formats to specific countries and cultures. In general, it has been found that adaptation of response format or stimuli to the environmental or cultural context enhances ability to respond or perform a task. This raises the issue of comparability of the instruments. Verbal rating scales are widely used in international marketing research and appear to be easily understood in many cultures. Even illiterate respondents appear able to express their feelings in words. Verbal rating scales are familiar and easily grasped by all types of respondents. These scales are quick to administer and require little additional explanation. Considerable care needs to be taken in their administration and interpretation. For example, some confusion may arise with interpretation of ends of the scale: ‘1’ may be considered the highest rather than the lowest point on the scale. Equally, Arabs and Israelis read from right to left, rather than left to right. Consequently, attention needs to be paid to providing clear instructions or, where scales are interviewer administered, providing adequate explanation to avoid misinterpretation. Difficulties can also occur in determining equivalents in different languages and countries of verbal descriptors for scale points (Voss et al., 1996). For example, numerical points can be used to develop measure equivalence scales rather than using verbal descriptors of scale intervals. However, these are not true equal interval scales and also may be context dependent. Continuous line or graphic rating scales can also cause problems. Respondents with low levels of education often have trouble conceptualizing a continuous scale from an extreme positive to an extreme negative, divided into equal intervals. Consequently, an interviewer may have to spend considerable time explaining the scale. This increases the time required to administer the questionnaire and limits its usefulness. Such problems suggest that the use of numerical rating scales (for example, pick x points out of 100) might be preferable. However, again, less-educated respondents have difficulty with such scales. In particular, they have difficulty in identifying the midpoint of a scale as well as interpreting distance between points on the scale. Consequently, considerable time is frequently required to explain numerical rating scales to less-educated consumers. This suggests that in countries with low literacy levels, considerable ingenuity will be required in developing mechanisms to record response. Interviewers can pose questions and record categorical responses concerning behavior and background characteristics, but difficulties can arise in obtaining responses to attitudinal questions, especially where indication of position on a scale by the respondent is required. One approach is to develop pictorial stimuli. Emotional response to a stimulus, for example, can be captured through the Self-Assessment Manikin (SAM). This is a pictorial device (Figure 8.3) that has been used effectively in a cross-cultural study of advertising effectiveness (Morris, 1995). The scale uses three different series of manikins to capture the dimensions of pleasure, arousal and dominance. The graphical portrayal of emotions helps eliminate the problems encountered when using photographs of people in cross-cultural research, where the apparent ethnicity of the models can influence responses. It can also be used among less-literate respondent populations and with children. Another device is the Funny Faces scale. Some research organizations have used this with illiterate or less-educated respondents in developing countries (Corder, 1978). It consists of five positions, ranging from very happy to very unhappy. Respondents are shown a concept, or read an attitudinal statement, and asked to indicate their degree of agreement or interest by indicating the corresponding position on this scale. For example, strong interest would correspond to being very happy. Some problems may, however, arise in using this type of rating scale. In particular, it may not be suitable in situations where (1) many attributes have to be rated or (2) a scale that is practical and quick to administer in the field is needed. In addition, the use of Western pictorial conventions in the Funny Faces scale may not be clearly understood. The Funny Faces scale has also been found to arouse negative reactions among better-educated respondents, who considered it childish and insulting to their intelligence. Figure 8.3 SAM: Self-Assessment Manikin Source: Morris, 1995. A similar type of scale has been used with aboriginal respondents in Australia (Donovan and Spark, 1997; Spark, 1999). A survey about community health in aboriginal communities used various pictorial scales to assess response to food consumption and health in the community. One response scale consisted of different-sized circles, from ‘lots’ – strong agreement – to ‘not at all’ – no agreement (Figure 8.4). Respondents were first given a practice scale to ensure that they understood the meaning of the different circles and then the questionnaire using the scale was administered. Another scale showed people on a hill (Figure 8.5). Those walking up the hill on one side were shown as healthy and strong, while those falling down at the bottom on the other side were ‘weak and sick’ in body. Respondents were then asked to indicate where they thought most men (women) in their community were. A similar scale was used to indicate whether people were happy or sad in mind. Figure 8.4 Questionnaire used in a study of Aboriginal health Source: Spark, 1999. Figure 8.5 Scale used to measure Aboriginal health perception Source: Spark, 1999. The most effective approach is to develop a scale that uses concepts familiar to the respondent. Steps of a ladder can be used as a scaling device (Cantril, 1965). Respondents are shown a ladder and asked to indicate their position in response to a given question, with respect to steps in the ladder. In the Middle East, use of an abacus provides an effective means of recording responses. The interviewer can describe the ends of the abacus as extreme points of a scale, for example agreement or disagreement. The respondent then moves the beads according to his or her degree of agreement or disagreement with the statement. In designing a response format, it is crucial to devise a format that the respondent can understand and that will enable him or her to respond accurately. As with other aspects of research design, it is important to consider comparability across different national and cultural contexts, but accuracy is of paramount importance. 8.4. Summary In designing instruments for use in survey research, the key issue is the development of a questionnaire that is clear, easily understandable and easy to administer. Questions need to be formulated so as to obtain the desired information from respondents and to avoid miscommunication between the researcher and the respondent. In multicountry research, one issue is the extent to which questions are formulated in precisely the same terms. This is more likely to be feasible for questions relating to demographic and other background characteristics than for behavioral or product market data. Greatest difficulty is likely to occur in relation to attitudinal and psychographic data. The way in which the question is posed, open-ended versus closed and direct versus indirect formulations, has also to be considered. Respondent comprehension is likely to be increased considerably by the use of visual stimuli such as show cards, pictures and photos. These can also be used to provide a check on bias arising from miscomprehension of verbal questions. It is, however, important to note that visual stimuli can be misinterpreted as well and need to be translated into the relevant idiom to ensure they are correctly interpreted. Instruments also need to be designed to minimize potential sources of bias in international surveys. This can arise first as a result of the topics covered in the questionnaire. Second, it can arise from the interaction between the interviewer and the interviewee. Third, it can arise from the characteristics of the respondent, such as his or her response style or socioeconomic and demographic origins. Finally, the format in which the response is obtained needs to be carefully considered. Here, a key issue is whether scales and response formats need to be adapted to specific countries and cultures. Particularly in developing countries with low levels of literacy, somewhat ingenious devices may need to be used in order to ensure accurate responses. In general, it is preferable to use different formats that generate equal responses, rather than the same format if this results in bias or less accurate responses. 9. Sampling And Data Collection Once a research instrument has been designed to collect the required data, the next step is to develop appropriate sampling and survey data collection procedures. Although sampling has to be considered in both survey and nonsurvey research, the main focus in this chapter is on sampling in survey research. Here, it is crucial to develop systematic procedures to ensure that reliable and comparable data are collected. This requires establishing a sampling plan based on the target population, developing a sampling frame or sampling list, and selecting appropriate survey administration procedures. These relationships are shown in Figure 9.1. Figure 9.1 Elements involved in developing a sampling plan In developing a sampling plan for a particular target population, decisions first have to be made with regard to the appropriate sampling frame, for example the world, country groupings, countries or units within countries. Next is the choice of sampling procedures. In the case of global and regional samples, the main problem is to find procedures that ensure representativeness of the target population. Since few comprehensive global or regional sampling frames are available, judgment or convenience sampling is often more practical and considerably less expensive than random sampling and also more likely to provide reliable and accurate data. This will particularly tend to be the case in business-to-business research, where the absolute population may be relatively small and known. In the case of national sampling units, in addition to finding an appropriate frame, researchers also have to consider whether the use of equivalent procedures for each sampling unit will yield comparable results. Differences in the availability and coverage provided by sampling frames or lists, and in the ease of reaching the target population, suggest that in some cases using different sampling procedures for each unit may provide better representation and be more cost effective (Hader and Gabler, 2003). A decision then has to be made as to how the survey should be administered. As in domestic research, four major alternatives can be identified, mail, telephone, personal interview or electronic. This latter method is becoming increasingly popular in Europe and the US as e-mail lists become more widely available. It also enables the use of pictorial or photographic stimuli, for example products or advertisements, as well as facilitating use of a combination of qualitative and quantitative techniques. Whatever the method used, the question of whether equivalent procedures from one country to another will yield comparable results has to be assessed. This, in turn, depends on the availability and adequacy of sampling frames for the target population. Mail surveys require the availability of a mailing list, and telephone surveys a list of telephone numbers although with sufficient density of numbers random digit dialing can be used. Personal interviewing provides greater flexibility, in that where convenience or quota sampling is used, a list of the target population is not required. Use of e-mail or the Internet, on the other hand, requires an adequate base of e-mail or Internet subscribers in the target population. The final sample, whether global, regional, national or within country, should be as representative as possible of the target population and, in the case of national and within-country sampling units, as comparable as possible across units. Cost considerations and sampling difficulties may limit the feasibility of obtaining representative samples. This chapter examines the various issues involved in sampling, particularly in obtaining sampling units that are as comparable as possible from one country or region of the world to another. First, the problem of obtaining sampling frames from which to draw the sample is examined. Second, the use of different sampling techniques is discussed. Third, the advantages and disadvantages associated with different survey administration methods in international markets are reviewed. Finally, issues associated with field staff selection and training are discussed. 9.1. Sampling In designing a sampling plan, the first step is to determine the appropriate geographical unit to be sampled – that is, world, country, region and so on – and to assess whether a sample frame for the target population exists. Appropriate respondents have then to be selected. For example, in a consumer survey of grocery products, the wife may be the relevant respondent; or in a survey of office equipment purchasing, the manager of the buying department. The next step is to determine the sampling procedure and size of the sample. 9.1.1. Selecting the Sampling Frame Once the target population to be sampled has been identified, the availability of a population list from which the sample may be drawn should be assessed. In international research, this frequently poses difficulties due to the limited amount of information available on industries, businesses or consumer groups in other countries. Even where sampling frames commonly used in Western countries are available, such as electoral or municipal lists, telephone books or listings, they frequently do not provide adequate coverage, particularly in developing countries, and can give rise to frame error. Lack of sampling frames often leads to use of nonprobability sampling in international marketing research. This can be used effectively in business-to-business marketing, where interviewing of certain key respondents may be more informative than systematic analysis of representative samples. In international marketing research, sampling may take place in relation to different geographical units. The most aggregate level is that of the world. The next level consists of geographical regions such as Europe or Latin America. Following this is the country level, and geographical units or other subgroups within countries, for example regions, cities, neighborhoods or city blocks. How the sample is drawn depends to a large extent on the specific product market research objectives and on the availability of lists for each type of unit. Some examples are shown in Table 9.1. The sequencing of research and whether, for example, one region or country is investigated first and then another are related to the availability of information for a given target segment. The issues associated with sampling at each level are next discussed in more detail. Table 9.1 The levels of the sampling frame Level Product market examples Examples of sampling lists World Financial institutions Machine tools D&B’s Principal International Business Power generation Gale’s Worldwide Business Directory Transnational consumer segments Subscribers to: Financial Times, Economist, National Geographic Regions Airlines Latin America 25 000 Automobiles D&B Europe Personal computers Regional trade associations Major Companies of Asia Countries Agricultural equipment National trade associations Construction supplies Credit card lists Consumer durables Population lists Consumer package goods Telephone listings Cities Upscale consumer goods Municipal lists Social services Church organizations Specialty foods Lists of government organizations Community organizations 9.1.1.1. The World The most aggregate sampling unit is the world. This is likely to be appropriate in business-to-business markets, such as injection molders, medical equipment, machine tools and so on. For these markets, worldwide lists of manufacturers can be obtained from sources such as Dun and Bradstreet’s Principal International Businesses (PIB), which has an online database listing 500 000 leading businesses in a variety of industries outside the US. CD-ROM versions available include 100 000 companies, 250 000 companies or 500 000 companies worldwide. An abridged print version is also available. Gale’s World Business Directory (2003) has information on 136 000 businesses worldwide and provides information on sales, net worth and number of employees as well as executive officers’ names and titles, e-mail and web addresses. Where a list for a specific country is required, in some cases trade associations are able to provide relevant and detailed information. Sampling at the world level is likely to be rare in consumer research unless the target population is a small global market segment. For example, subscribers to the National Geographic or American Express cardholders might be an appropriate target sample for testing a new foreign travel publication or travel goods, or readers of the Economist for business-class travel. Equally, Internet users might provide an appropriate sampling frame for specialized technology-based services. A major restriction on sampling at this level is the lack of information relative to the target population. 9.1.1.2. Country Groupings Samples can be drawn based on country groupings such as the Scandinavian countries. Again, this is most likely to be appropriate in business-to-business markets. Graham and Whiteside publish company directories such as Major Companies of Europe in six volumes, which list more than 24 000 of the largest companies in Europe. This also lists the names of 194 000 senior executives, their e-mail and web addresses, as well as a description of business activities, brands and trademarks and financial data for the last two years. More complete information on Graham and Whiteside’s directories can be obtained directly from its web site (www.major-co-data.com). Again, trade associations or financial associations also provide lists or directories for a specific industry. In a few cases, regional sampling of consumer markets may be undertaken. For example, the Europanel, established by a consortium of research agencies throughout Europe (GfK), provides a representative sample of consumers in Europe. In general, such samples are built up from nationally representative samples. Equally, Global Market Insite manages web-based panels in 200 countries on four continents, which can be accessed to develop customized regional or multicountry samples (www.gmi-mr.com). 9.1.1.3. Country Despite the globalization of markets, the country is still the most common level for drawing a sample in international research. In industrialized countries, sample frames such as electoral lists, population censuses and telephone books are commonly used for drawing samples. Such sampling frames are not always available or current in other countries and coverage will vary. In Germany, for example, the publication of household telephone numbers became optional in 1992 and the percentage of households listed dropped from 97% to 72%, substantially reducing the coverage provided by telephone listings. Similarly, in Switzerland when listing became optional in 1998, the proportion of households listed dropped to 88% in 2000 (Hader and Gabler, 2003). Some countries, notably in the Middle East, Africa and parts of Asia, lack any type of population lists. In others, street maps are not available, streets may have no names and houses may not be numbered. Lack of any commonly used sampling frames means that the researcher will have to construct a sampling list from scratch. In the Middle East, especially in Saudi Arabia, research organizations typically establish a sampling frame based on city blocks in the major cities. This is developed based on accumulated experience in previous surveys. Since different nationalities and socioeconomic groups live in particular areas, these blocks can then be used as a basis for quota or stratified sampling. Random sampling is not feasible, nor likely to be desirable given the stratified nature of the society. Different biases may be inherent in the use of different sampling frames. Outside industrialized countries, for example, use of telephone lists will provide a relatively skewed sample, consisting primarily of individuals or households of higher socioeconomic status. Similarly, use of city block data may result in underrepresentation of low socioeconomic respondents living in shacks or riverboats. 9.1.1.4. Units within Countries Samples can also be drawn based on subgroups within countries. These might be geographical units such as cities or neighborhoods, or alternatively ethnic, racial, cultural, age or demographic subgroupings such as Asian or Turkish immigrants in Europe or the Middle East, Vietnamese immigrants in Australasia, children or senior citizens in cities of over 500 000 in North America. Similarly, in an organizational context, specific industries or organizations of a certain size – that is, small versus medium-sized businesses – might constitute the relevant population. The availability of information from which to develop a sample is likely to vary with the specific subgroup. For geographical units, this will pose little problem if maps or local electoral lists can be obtained. Similarly, if ethnic, cultural or socioeconomic groups tend to live in certain neighborhoods, city block frames can be developed as a basis for sampling. Children can be sampled based on school districts and religious or ethnic groups based on church membership, religious or ethnic organizations. Members of such organizations will tend to be those with strong ethnic or cultural affiliation, and are not necessarily typical of all members of a particular cultural or ethnic group, but rather of ‘core’ members of the group. 9.1.2. The Choice of Respondent Once the sampling frame has been identified, the specific respondent to be sampled has to be determined. Here, an important consideration in studying families and organizations is to determine who is the relevant respondent; that is, the wife, husband or other person in a household, or in an organization the buyer, user and so on. In addition, a decision has to be made as to whether a single respondent is used or whether multiple respondents will be required, for example both husbands and wives and children in the family, or several members of a buying committee. It is important to identify the relevant respondent(s) in each country, since these vary from country to country. In some countries or cultures, for instance Asian and Latin markets, organizational decision making is highly centralized. Consequently, the relevant respondent may be the chairman, CEO or senior management. In other countries, for example Anglo-Saxon cultures, there is a greater tendency to delegate decision making, hence middle management may play a greater role. Similarly, in households, the relevant respondent(s) and roles of different household members in various purchasing decisions have to be determined. In extended families, for example, one person may be primarily responsible for groceries and other frequently purchased items. In some Middle Eastern countries, the husband may actually purchase groceries, but the wife specifies what items are to be purchased. In other cases, both may go together to purchase items. Husbands also play a major role in wives’ clothing decisions. In the case of consumer durables such as automobiles or household appliances, several family members may be involved. Consequently, focus groups to determine the relevant involvement of different family members may be desirable. The number of respondents to be interviewed from a given household or organizational unit has also to be considered. The cost of collecting data from multiple members is likely to be prohibitively expensive. As a result, especially for frequently purchased consumer goods, a single-family member, often the wife, may be used as a surrogate for other members. Similarly, in the case of organizations, the difficulty and cost of interviewing multiple managers will lead to selection of a single person as representative and as an informant about the decision process. 9.1.3. Sampling Procedures The next step is to determine appropriate sampling procedures. First, the researcher has to decide whether research is to be undertaken in all countries and contexts, and whether results and findings are generalizable from one country or context to another. Ideally, research should be conducted in all countries and contexts where marketing operations are planned. There is a trade-off between the number of countries in which research is undertaken and the depth or quality of research. Consequently, management may decide to use findings in one country as a proxy for another. For example, market response patterns in Scandinavian countries may be sufficiently similar to allow sampling in only one of these countries. Similarly, in the Middle East, strategy may be based on research conducted in Saudi Arabia and then rolled out to other Gulf states. It is important to realize that such a procedure is fraught with danger. Even though previous experience suggests that response patterns are the same or similar, this may change, or not be relevant in relation to the specific product or situation examined. In selecting sampling procedures, a key decision is whether random or purposive sampling should be used. This is closely linked to the availability of the sampling list and survey administration techniques. If sampling lists of the relevant population are not available, as is often the case in developing countries, probabilistic or random sampling poses some difficulties. Consequently, in much consumer research in international markets quota sampling is more likely to be used, often in conjunction with block sampling or random location sampling points. Similarly, in the case of business-to-business research, convenience or judgment sampling is most likely to be used, since the total population is often relatively small and known. 9.1.4. Sampling Techniques The next step is to select an appropriate sampling technique. Here, a major distinction exists between probabilistic and nonprobabilistic sampling. In probabilistic sampling, each respondent in the target population has a known chance of being in the sample. In nonprobabilistic sampling, some criteria are established on the basis of which respondents are selected. For further in-depth discussion of sampling techniques see Aaker et al. (2003) or Churchill and Iacobucci (2005). In industrialized countries, random or probabilistic sampling is generally considered desirable, though the same sampling design does not need to be used in each country (Hader and Gabler, 2003). Different sampling designs may provide equivalent results and representativity due to differences in sampling frames. Comprehensive lists of the target population are frequently available in such countries, though this is not always the case in other countries. Lack of published information about the relevant target population, limited availability of sampling lists and the cost associated with the development of such lists suggest that other methods such as judgment, convenience, snowball or referral sampling may be more cost effective. This may be desirable where personal interviewing is used rather than mail or telephone surveys. The various types of sampling are next discussed in more detail. 9.1.4.1. Nonprobabilistic Sampling Convenience Sampling In some cases, a convenience sample may be used. This implies selecting any respondent who is readily available. In emerging country markets, for example, convenience sampling in the market place provides a low-cost procedure for generating a sample. Given the difficulties and costs of developing sampling frames in such countries and in reaching the rural population, this procedure can be used to generate a sample that, while not strictly representative, may nonetheless be relatively free of any systematic bias. Judgment Sampling Another procedure is to select respondents based on judgment. Judgment sampling is based on the assumption that certain persons are better informed than others and have expert knowledge in a given field, or are specified in the research brief(s), for example customers of a client company. For instance, in international marketing research, judgment sampling may be used to identify area or industry experts for a given country or region in order to assess trends in the industry or region. In business-to-business research, this often provides a more efficient method of assessing the likelihood of new product acceptance, industry growth, market conditions and so on than use of quota or probabilistic sampling. The international sales force is also often a valuable source of information, both for identifying respondents and providing customer information, since they know customer needs and interests. Care should, however, be taken in using them to obtain quantitative estimates, such as of sales potential. Importers or export agents can also be used as ‘key informants’, though systematic bias may be introduced, reflecting the self-interest of the source. In emerging markets, questioning of village elders, priests or other local authority figures can be used to obtain information about the number of local inhabitants, their purchase behavior and other issues. In rural areas, or among respondents with high levels of illiteracy, this provides a relatively low-cost and rapid means of obtaining a general indication of relevant information, though not a precise estimate of actual figures. Quota Sampling A procedure widely used in both industrialized and emerging country markets is quota sampling. Quotas are established by specifying the number of respondents from within a given category, for example by age or socioeconomic group, occupation, nationality, urban versus rural, purchasers versus nonpurchasers and so on. In business-to-business research, quotas may be established by industry or the equivalent of an SIC (Standard Industrial Classification) category, firm size, location and so on. In industrialized countries, as markets become more segmented, quota sampling is increasingly used to ensure coverage of the target population. In emerging markets, quota sampling may also be used where there are known to be significant differences in behavior by nationality, income group, age and so on. For example, in the Gulf states there are significant differences in brand preferences and loyalty for cigarettes by nationality and income group. Consequently, quotas are established by nationality grouping. While this procedure ensures that the sample will be representative based on quota characteristics, there is a danger that these characteristics are systematically associated with other factors that will introduce confounding effects. Snowball Sampling Snowball or referral sampling is a technique well suited to international marketing research. With snowball sampling, initial respondents are selected randomly or based on judgment, and they are then asked to identify other members of the target population. This procedure continues until a large enough sample is obtained. This technique is helpful where the target population is difficult to identify, for example industrial buyers of a product, financial experts, or users of food stamps. In some cases the initial sample is a random sample, but in others it can be selected by judgment. In either case, the final sample will be nonprobabilistic, since respondents are likely to identify others similar to themselves. In the Middle East, referral sampling is widely used in order to obtain a sample of female respondents. A number of women are first selected, often of different nationalities, for example Arabs from different countries or of different socioeconomic classes. These women are then asked to indicate other women whom they think would be willing to be interviewed. Typically, these will come from the same socioeconomic group and be of the same nationality. This procedure is followed until the desired sample size for each group is obtained. 9.1.4.2. Probabilistic Sampling Simple Random Sampling Random sampling frequently poses a number of problems in international marketing research. In the first place, it requires the availability of a frame or list. Respondents are then picked at random from the list. A simple procedure is to use a random number table to establish a starting point and then select every nth name or person on the list until the desired sample size is obtained. These respondents then constitute the sample population. The lack of sampling lists often limits use of this approach, especially in emerging markets. If the survey is interviewer administered, an alternative procedure is to use the random-walk method. The interviewer then becomes responsible for sampling. He or she is provided with a route and instructed to interview a respondent in every nth house. This poses problems in some countries since the interviewer may have difficulty following the route, or determining exactly what is a ‘dwelling unit’ in villages, shanty towns or where buildings include multiple-family units. Furthermore, if there are no existing maps of an area, a mapping of dwellings in an area will need to be developed. The difficulties and costs associated with random sampling are a key reason why nonprobabilistic sampling is frequently used in international marketing research. Stratified Sampling In some cases it may be considered important to ensure that samples from different countries are representative on certain characteristics such as income, education, age, nationality, single-family or single-parent households, or business versus consumer users. This is particularly important where similar segments are to be targeted in different countries or regions throughout the world. A random sample is then taken from each group or stratum of interest in the population. Stratified sampling is either proportionate or disproportionate. In proportionate stratified sampling, the sample should be proportionate to the relative size of that stratum in the total population. In disproportionate sampling, other factors, for example differences in variance in each stratum, determine the size of sample taken from each stratum. In estimating market size for small power tools, for example, two segments may be of interest: ‘do-it-yourselfers’ and the professional segment, including craftsmen, repairmen and handymen. Greater variation may be expected in the professional segment in terms of frequency of use, type of user and so on than in the consumer segment. Consequently, it may be desirable to draw a larger sample from the professional segment relative to its size than from the consumer segment. Quota Sampling Quota and stratified sampling are frequently confused. While similar in that the objective is to obtain representativity on certain key characteristics, stratified samples are drawn probabilistically, while in quota sampling respondents are selected on a judgmental basis. In other words, any respondent who meets the specified characteristics may be picked until the desired quota for that subgroup are met. As a result, a quota sample may contain certain biases that are less likely to occur in a stratified sample. Cluster Sampling A related technique is cluster sampling. While in stratified sampling a random sample is selected from each stratum or subgroup, in cluster sampling the target population is first divided into mutually exclusive categories or clusters. A random sample of clusters is then selected, and either one-stage or multistage sampling conducted. In one-stage cluster sampling, all the population elements in each selected cluster are examined. In multistage cluster sampling, a random sample is drawn for each selected cluster. A common form of cluster sampling is area sampling. In this case, geographical units such as cities, regions, areas or blocks within cities are the clusters. Either one-stage or multistage sampling can then be conducted based on these clusters. This procedure is particularly useful and cost effective where no population lists are available. Area sampling can be either one-stage or multistage. In one-stage area sampling, the geographical units are selected and all relevant respondents within that unit studied. In multistage area sampling, further sampling within geographical units takes place. Suppose, for example, a manufacturer of detergents was interested in estimating consumption of powdered detergent in urban households in Taiwan. If one-stage area sampling were adopted, he might select four residential areas in three major urban areas, Taipei, Kaioshung and Taichung, and estimate consumption per household of detergent in each of these areas. The average for the four areas in each city could then be used to estimate overall consumption of detergent in the area. 9.1.4.3. Multistage Sampling Multistage sampling can also be used in conjunction with other sampling approaches and often involves a mix of techniques. In business-to-business research, specific industries may be selected for investigation based on judgment and then quota sampled, based on size within each industry. Similarly, in emerging market countries, area sampling may be used to pick major cities and villages to investigate and then stratified sampling employed. Alternatively, block sampling might be applied to select households to be interviewed within the cities, while in the villages judgment sampling of elders is used. Multistage sampling can help to reduce the costs of international marketing research. Efficiency is increased, since the initial stage(s) is used to pinpoint relevant respondents to be sampled subsequently. It is particularly likely to be appropriate for large-scale surveys and in developing countries, where sampling frames are not readily available. On the other hand, multistage sampling can also result in some loss in precision, since sampling errors accumulate from one stage to another. In addition, more time is likely to be required to identify the initial sample than if sampling were conducted in a single stage. In some cases where the initial stage sampling is based on secondary data sources, for example where the clusters are regions, states, cities or industries, multistage sampling can increase the efficiency and speed of the research process. A major cross-cultural survey using multistage probability sampling is the Eurobarometer (Hader and Gabler, 2003). The sampling procedure is a multistage random design based on a random selection of primary sampling units (PSU) according to the distribution of the national population by metropolitan, urban and rural areas. In the second stage, a cluster of addresses is selected from each primary unit and addresses to be sampled are chosen systematically using a standard random route procedure. In each household the respondent is selected by a random procedure such as the first birthday method. Up to two recalls are made to obtain an interview with the selected interviewee. As the preceding example illustrates, multistage random sampling requires extensive resources and is highly laborious. Consequently, it is primarily used by government and other large international organizations. In commercial research, two alternative forms of multistage sampling, double sampling and sequential sampling, are more commonly used. Double Sampling In double or two-phase sampling, the samples are drawn twice. Once an initial sample has been drawn, data are collected from respondents about characteristics such as purchasing behavior and frequency, demographic variables, location and availability for future interviewing and so on. This information is then used to develop a frame for drawing a second sample, based on respondents in the initial sample. These respondents are interviewed a second time in greater depth. In evaluating international market potential for web-site construction and management services, for example, an international telephone survey might be conducted to determine which industries and what size of companies are the heaviest purchasers of such services in different countries or regions throughout the world. A preliminary list of companies to contact could be established based on knowledge of heavy-user industries in the domestic market. Based on data collected from this survey, a list could be drawn up of companies and industries that appear to be heavy users and potentially cooperative respondents. A sample could then be drawn from the list, and these companies investigated in greater depth with regard to purchasing behavior, key criteria used for evaluating vendors, importance of service and training, key managers influencing the purchase decision and so on. Sequential Sampling In the case of sequential sampling, the total sample size is not determined in advance. Respondents are interviewed one after another, and the data analyzed simultaneously or at specific points in time. Depending on the reliability of results, a decision is made about whether more respondents should be interviewed. With the increasing use of laptop computers for interviews and computerized interviewing techniques, this procedure has become increasingly popular. Data are keyed directly into the computer and results updated after each entry. It is also frequently used for Internet-based surveys, where again results can be analyzed instantaneously. The decision when to stop sampling is then based on the stability of results obtained from each successive batch of respondents; that is, the stability of results obtained from the first 100 respondents relative to the second 100 and so on. The cost efficiency of research is increased as the minimum number of respondents to achieve reliable results is interviewed. 9.1.5. Sample Size Another decision concerns the appropriate sample size. Assuming a fixed budget, there is a trade-off between the number of countries or contexts sampled and the sample size within each country or region. Use of statistical procedures to determine the appropriate sample size is likely to be rare, as some estimation of population variance is required. Typically, the research budget will determine sample size. Management will decide that samples of 400 or 10 focus groups in each country are adequate, given budget constraints. If a sequential sampling procedure is used, then the sample size will be variable, depending on the stability of successive sample results. In determining samples within country units, population diversity should also be considered. Differences with regard to key variables or sampling characteristics such as income, age, education, nationality and so on may occur and affect results. For example, different linguistic groups or ethnic or nationality groups may have substantially different behavior and purchase patterns. Equally, there may be wide variation by income or socioeconomic group, as for example in India or Latin America. Consequently, it may be desirable to do quota sampling by these groupings in order to ensure population representativity. This may, however, require larger than normal sample sizes to test for the impact of differences in these variables on cross-national findings. This entails high sampling costs and can pose budgetary difficulties. Use of large sample sizes – that is, of more than 1000 respondents in each country – is rare except in government surveys or those sponsored by international organizations, such as the Eurobarometer. 9.1.6. Achieving Comparability in Sampling 9.1.6.1. Sample Composition A key issue in sampling design in international marketing research is the relative importance attached to representativity, as opposed to the comparability of the samples. If samples are representative of the target population, they are unlikely to be comparable with regard to key characteristics such as income, age and education. This will create a confounding effect if, as is frequently the case, such variables affect the behavior or response patterns studied. For example, income or education might be an important factor underlying interest in tropical fruit in different countries. If national samples are drawn in countries with different income levels, mistaken inferences could be made about national differences or similarities of interest in tropical fruits, when these reflected differences in income rather than ‘true’ national differences. One might, for example, conclude that there was lack of adequate market potential, when in fact a small, high-income segment constituted a potential spearhead for market entry in a low-income country. The relative importance attached to representativity vs comparability should depend to a substantial degree on the purpose of the research (Reynolds, Simintras and Diamontopoulos, 2003). In descriptive or contextual studies, where the primary concern is with external validity – that is, generalizing results to the country or cross-national group of interest – attention should be focused on the representativity of the sample. In comparative and theoretical studies, on the other hand, where the primary concern is internal validity and hence the homogeneity of the sample, emphasis should be placed on comparability. In addition to achieving comparability through a well-thought-out sampling plan, there are statistical procedures such as covariance analysis that can be used to evaluate the impact of different sample compositions on results and to adjust for these. These are discussed in Chapter 11. 9.1.6.2. Sampling Procedures Another issue is whether the same sampling procedures should be used across countries, if they vary in reliability from one country to another. In this case, rather than using identical sampling procedures and methods in each country, it may be preferable to employ procedures that have equivalent levels of accuracy or reliability. Suppose, for example, that in one country random sampling is of known validity, and in another country quota sampling. The results will be more comparable in terms of response rate and quality of response if random sampling is used in one country and quota sampling in another, than if the same sampling procedures are blindly applied in each research context (Hader and Gabler, 2003). Similarly, costs of sampling procedures may differ from country to country. Administrative cost savings achieved from using the same method in many countries may be outweighed by use of the most efficient sampling method in each country. In one country, area sampling may be the most effective, while in another country quota sampling may produce acceptable results at half the cost. Consequently, it may be more appropriate to use the quota method in the latter country, while employing area sampling in other contexts. Use of different sampling methods can also provide a check on the reliability of results and the potential bias inherent in different methods. For example, in an early study different sampling procedures with different sources of potential bias were used in five different countries (Webster, 1966). With one procedure a consistent bias was found in relation to one of the main variables studied, the percentage of firms in each size category owning the test product. If the same sampling procedure had been utilized in each country, this might not have been detected. In brief, use of similar sampling procedures will not necessarily ensure comparability of results, since each procedure is subject to different types of bias and these vary from country to country. Deliberate variation of procedures, on the other hand, if intelligently used, can provide a means of checking the validity of results and detecting biases inherent in different types of procedures. 9.1.7. Sampling Error Sampling is also a potential source of error in cross-national surveys. This arises because a sample, rather than a census, is employed to collect data. Sampling error can arise from frame error and from survey nonresponse. Where errors arise due to nonresponse, standard techniques can be used to increase response and to weight results to account for nonresponse. Where frame error occurs, there is little that can be done. As noted earlier, this is particularly likely to give rise to problems in international marketing research, due to the difficulty of obtaining accurate sampling lists, unless the researcher is willing to develop his or her own list or frame. 9.1.7.1. Frame Error The lack of adequate sampling frames or lists in many countries implies that samples may not be strictly comparable from one country to another. Consequently, sampling error may differ from one country or sampling unit to another. For example, the subscription list to the Economist provides better coverage of the business population in English-speaking countries than in French- or Spanish-speaking countries. If the limitations of the sampling frame are known, the sample can be weighted to account for these where random sampling procedures are used. If, however, judgment or convenience sampling methods are used, it may be difficult to evaluate the degree of bias arising from sampling error. 9.1.7.2. Survey Nonresponse Nonresponse to surveys is a source of major concern that threatens the validity of results no matter where research is conducted (de Leeuw and de Heer, 2001). As noted earlier (in Chapter 8), response rates in official statistical surveys have been declining internationally for a number of years, though there are differences between countries in the rate of decline as well as in the overall response rate (de Heer, 1999). Responses have for example been found to vary between 60% for the Netherlands and 97% for Germany. A similar trend appears to be occurring in other types of surveys. Response rates also differ depending on the mode of data collection. A meta-analysis of response rates in mail, telephone and personal interviews found a difference of 10% between response to personal interview and to mail surveys (Hox and de Leeuw, 1994). In personal interviews the most important factors underlying nonresponse are noncontact rates and refusal to participate. Again, differences have been found across countries. In some countries, for example Belgium and Denmark, noncontact rates were the most important, whereas in others, for example the Netherlands and the UK, refusal to participate was more critical. Nonrespondents appear to have the same profile from country to country and are typically low-income, less-educated consumers. Samples will be underrepresented with regard to the same segments in different countries. This can be corrected by double sampling on high-nonresponse segments. If other factors such as suspiciousness of interviewers or hostility to surveys underlie nonresponse, the relevant determinants and their impact will need to be investigated in each specific case. Relatively little research has been conducted into ways to increase rates of nonresponse in multicountry surveys, particularly among consumers. Use of monetary incentives, reminder postcards, sponsorship and sponsoring organizations’ stationery and personalization have typically been found to increase response in mail surveys, although the increase in response rate varies from country to country (Dawson and Dickinson, 1988; Jobber and Saunders, 1988). Japanese respondents have been found to have higher rates of response and to respond to incentives (Keown, 1985). Follow-up mailings are typically the most effective, particularly in consumer surveys (Nederhof, 1989). In the case of personal interviews, increasing preparation and effort in fieldwork may be desirable. For example, increasing the number of contact attempts, varying the time at which contact is made and following a prescribed schedule have all been found to reduce the number of noncontacts (Purdon, Campanelli and Sturgis, 1999). The most appropriate time to interview respondents may vary with the culture or country. For example, in the Netherlands it may be more effective to contact respondents early in the evening since the Dutch tend to eat early, while in Spain or Italy a later time may be preferable. In general, existing evidence suggests that standard procedures such as personalization, sponsorship and follow-ups, successfully used by research organizations in the US and Western Europe, are also effective in other countries and can be used to increase rates of response. 9.1.7.3. Nonsampling Error Surveys have to be examined for bias arising from nonsampling error. This is somewhat insidious in nature. Although the impact of the relevant factors has been extensively studied in the US and Europe, relatively little is known concerning factors that affect the quality of data provided by respondents in other countries. Few studies have been conducted and these relate primarily to a single country or area. Nonsampling error can arise as a result of a number of factors: the respondent; the interviewer– respondent interaction; item nonresponse; or recording error. In the case of the respondent, factors such as willingness to respond accurately and completely, response set bias, purposeful misreporting of information, faulty recall or respondent fatigue may all generate error. Item nonresponse is another potential source of error. Respondents may refuse to answer certain questions or state that they have no opinion about certain issues. These have been discussed in relation to instrument design (Chapter 8) and are not covered further here. The interviewer–respondent interaction, if not handled appropriately, can result in biased responses. Error from this source can be reduced by better research design. For example, questionnaires can be designed that require little or no clarification by the interviewer, and show cards and other stimuli used to assist the respondent. Improved interviewer training in how to administer questionnaires or conjoint tasks, or to create a relaxed interviewing environment so as to obtain better respondent cooperation, can also improve data quality. Recording error is an important source of error. This may arise either from carelessness or inaccuracy by the respondent where the survey is self-administered, or from lack of care by the interviewer. Recording error is of particular concern in contexts where there is a lack of experienced field staff and interviewer training is required. Use of computerized techniques can help to reduce some sources of error, particularly those due to interviewer–interviewee interaction, but can give rise to other recording errors and furthermore can only be used in industrialized countries. 9.2. Data Collection Procedures The next step is to select appropriate data collection or survey administration procedures. Here, a number of factors need to be considered. In domestic marketing research, the relative cost, target population, the length and type of survey and time constraints largely determine the choice of questionnaire administration procedures; that is, mail, telephone, personal interview or electronic. In international markets, however, other factors, such as the development of the marketing research or communications infrastructure, affect the decision, particularly with regard to telephone, email or web-based surveys. Levels of literacy, as well as the lack of sampling lists, imply that mail surveys, while potentially a low-cost means of reaching a large target population, are fraught with problems and hence personal interviewing may be preferable. The quality of the mail service and its reliability will further influence the cost effectiveness of mail surveys. Levels of telephone ownership and particularly private telephone ownership also limit use of telephone surveys, except for business-to-business or upscale target populations. 9.2.1. Mail Surveys In industrialized countries, mail surveys typically enable coverage of a wide and representative sample, they do not require a field staff, and costs per questionnaire tend to be relatively low. Respondents may be more willing to provide information about certain issues, for example financial matters, and have time to answer questions requiring thought or specific information at their leisure. On the other hand, nonresponse rates may be high, resulting in high costs per returned questionnaire and bias due to nonresponse. Control over questionnaire administration is also lost, and there may be omissions or lack of comprehension of questions. Furthermore, certain types of questions cannot be asked and interviewers cannot probe to obtain further information. In international marketing research, the absence of mailing lists, poor mail services and high levels of illiteracy limit the use of mail surveys. This does, however, depend on the specific product market and country concerned. Mail surveys can typically be used effectively in international business-to-business research. Mailing lists are often available. The key problem is to identify the relevant respondent within a company, and to personalize the address to ensure that the survey reaches him or her. Prior telephoning to identify the relevant respondent and to obtain his or her cooperation is likely to increase the probability of response. Sending the questionnaire by fax rather than regular mail helps to underscore the urgency of the survey and speeds up response. However, widespread use of fax numbers by direct marketing and sales companies has tended to result in a ‘junk fax’ phenomenon in many countries. Consequently, it is important to make prior contact regarding the survey in order to substantiate its authenticity and obtain respondent cooperation. In consumer research, particularly in developing countries, use of mail surveys may give rise to problems. Mailing lists comparable to those in the US and major European markets are often not available, or not sold, for example credit card listings. Lists that are available, for example magazine subscription or membership association lists, are often skewed to certain segments of the population, such as the more affluent. In addition, low levels of literacy and reluctance of respondents to respond to mail surveys limit their effectiveness. Mail surveys are extremely problematic in countries such as Brazil, where it has been reckoned that 30% of the domestic mail is never delivered, or Nicaragua, where all the mail has to be delivered to the post office. In addition, mail surveys are typically slow. Thus, mail surveys can be effectively used in business-to-business research, especially if administered by fax. In consumer research they are only likely to be effective in industrialized countries and where the goal is to reach specific customer segments, for example when conducting a customer satisfaction survey of patrons. While costs of administering mail surveys may appear low on a per questionnaire mailed out basis, low response rates or poor-quality data may limit the desirability of using them, especially in emerging countries. 9.2.2. Telephone Interviewing With the growth of telephone networks, more widespread ownership of telephones and increased ease of international communications through satellite links, use of telephone interviewing in international marketing research has increased considerably, especially in business-to-business research and in industrialized markets. The primary advantage of telephone surveys is that they enable coverage of a broadly distributed sample without requiring a field staff. Telephone interviews provide a quick way of obtaining information and nonresponse is generally low. In addition, control over interviewers and interviewer–interviewee interaction is facilitated. Interviews can be conducted at a central location and bilingual interviewers briefed about the survey. Typically, the interview is computer steered. The interviewer sits in front of a monitor and replies from each respondent are entered directly into the computer, eliminating many sources of interviewer bias and recording error. In addition, as noted earlier, results can be analyzed sequentially and sampling terminated when responses stabilize. Random dialing can also be computer generated, so that each telephone number within a given geographical area has an equal probability of being in the sample. If a respondent is not reached on an initial attempt, he or she can automatically be called back after a specified interval. Centralizing the administration of international telephone surveys significantly reduces the time and costs associated with negotiating and organizing a research project in each country, establishing quality controls, conducting callbacks, etc. Trained interviewers fluent in the relevant language are required, but with the growth of international research and increased migration of populations, many research organizations now have their own staff of interviewers fluent in key languages. While the additional costs of making international telephone calls are higher, these are typically outweighed by lower administrative costs and accuracy of response. International calls also typically obtain a higher response rate. Results obtained by using this approach have been found to be highly stable, the same results emerging from the first 100 interviews as from the next 200 or 500. Interviewer and client control is much greater. The questionnaire can be modified in the course of the survey and interviewing extended or halted to meet the client’s requirements. In business-to-business research, use of telephone surveys is often quite effective. The majority of businesses except in rural areas are likely to have telephones. It is important as in the case of mail surveys to be able to identify the relevant respondent(s). This can be facilitated by initial probing or prior notification by fax. Willingness to respond may, however, depend on relative time pressures at work and the desired target population. If the target population is upper management, some resistance is likely to be encountered unless substantial interest in the survey can be aroused. For example, suppliers may conduct surveys to identify potential customer needs or to determine customer satisfaction. Again, prior notification and establishment of an appointment to conduct the survey are likely to be helpful. On the other hand, telephone interviews cannot be too long, though in some cases telephone interviews of 30–90 minutes can be held. Questions have to be short and clear, and certain questions such as similarities or trade-offs cannot be used. Visual aids such as product samples or show cards also cannot be used. In some countries, particularly outside the US and Western Europe, low levels of telephone ownership and poor communication links limit the feasibility of telephone surveys, especially for consumer research. In the poorer countries of South East Asia, telephone mainline connections vary from 99 per 1000 in Thailand to 2 per 1000 in Cambodia and 10 per 1000 in Laos, and cellular phone ownership is equally low, largely precluding use of telephone surveys to reach a broad-based target market. In some countries, ownership of mainline telephones is low but is compensated for by higher levels of cellular phone ownership. For example in Venezuela, ownership of mainline telephones is 109 per 1000 but that of cellular phones is 263 per 1000, while in South Africa the figures are 111 and 242 respectively. Those who own telephones are typically in cities and/or of higher socioeconomic status, so that telephone surveys are only effective in reaching a relatively limited segment of the population. However, in many instances, this may be the relevant segment for other products as well. In many countries, particularly in Europe, mobile phone usage has increased dramatically in recent years. In the countries shown in Table 9.2, the number of cellular phone subscribers is higher than the number of mainline phones and in some cases more than one per person. This makes it important to consider inclusion of mobile phone users in telephone surveys, though this may depend to some extent on the target population. Mobile phone users, particularly those with no fixed phone, are more likely to be young, single and to move frequently. Consequently, if younger singles are an important component of the target population, it will be important to include mobile phones in a telephone survey. Yet in some countries, even families are becoming important users of mobile phones. In Finland, for example, the number of mainline subscribers is declining as more households are giving up fixed-line phones and relying solely on the use of mobile phones (Kuusela and Simpanen, 2003). Table 9.2 Number of phone subscribers and personal computers per 1000 in selected countries (2003) Telephone Mainlines Cellular Subscribers Personal Computers Luxembourg 797 1061 594 Israel 453 955 243 Hong Kong 555 1057 422 Italy 484 1018 231 Iceland 660 957 451 Sweden 736 889 621 Finland 488 901 442 Czech Republic 360 947 177 Norway 734 909 528 Greece 454 780 82 UK 591 841 406 Slovenia 407 871 301 Denmark 669 887 577 Portugal 414 904 135 Spain 429 916 196 Switzerland 744 843 709 Source: ITU, 2004 Surveying mobile phone users poses its own challenges, not the least of which is identifying mobile phone users. The owner of a mobile phone is often not the user, for example a large number of mobile phones are used by children. Mobile phone users may also be contacted while driving, or in a crowded location with ambient background noise, making interviewing difficult if not impossible. Consequently, it is generally advisable to make an initial phone call to establish a time to call back when the respondent will be in a more suitable location and be available to respond to questions. In many countries surveying mobile phone users adds to costs, as calls to mobile phones, particularly from a fixed phone, are more expensive than those to a fixed phone and roaming costs are added if the mobile phone user is out of the country. Because of these problems, as well as difficulties in obtaining lists of mobile phone users in many countries, mobiles are not usually included in random digit dialing (Jenkins, 2001). 9.2.3. Personal Interviewing Personal interviewing is the most flexible method of obtaining international research data. The respondent is clearly identified and hence the nature and distribution of the sample can be controlled. Nonresponse is typically low and all types of questions or data collection techniques can be used. On the other hand, personal interviewing does require the availability of a trained local field staff. Control of interviewer cheating is necessary, and procedures are required to administer fieldwork, to back check and to control the quality of fieldwork. In some cases large international research organizations, notably those with local branch offices, train and develop their own field staff. This provides greater control over the quality of the fieldwork and assures greater cross-country consistency. Research organizations providing coverage in the Middle East establish their field staff by a tiered system of referrals. The local field manager in each country identifies an area manager for each city and for each nationality or ethnic group. These managers, who may be couples or women managers, in turn recruit interviewers for their specific area. In general, interviewers have to be of the same nationality and ethnic group as the people they are interviewing. Both interviewers and interviewees are recruited through personal referrals, requiring detailed knowledge and personal contacts for each local neighborhood. Alternatively, companies may buy fieldwork from a local research organization. In some cases they belong to global networks, such as Global Market Research, and buy from member companies in the organization. In other cases they solicit bids from field organizations in the country and select the supplier based on previous experience with the company, its suitability for the job and so on. Many of these companies belong to international research organizations such as ESOMAR, and conform to their standards and code for conducting fieldwork. Personal interviewing is commonly used in emerging markets such as India, China, Eastern Europe, the Middle East and South America. Low wage costs imply that personal interviews are not expensive relative to other methods of data collection, as is the case in many industrialized countries. In addition, in some cases it may be the only way of collecting data, due to low levels of telephone ownership, illiteracy and absence of mailing lists. Personal interviewing provides an effective means of adapting questions to a specific company situation or individual and of probing for answers. In business-to-business research, willingness of management to cooperate and provide desired information will depend to a large extent on the competitiveness of the market environment, and also perceived time pressures. In certain product markets such as pharmaceuticals or electronics, management may be reluctant to provide information for fear this may be leaked to competitors. This may depend on the purpose of the survey. As in the case of telephone or mail surveys, management may be more willing to participate in customer satisfaction or product testing studies than usage surveys. In consumer research, as noted earlier, personal interviewing is commonly used outside the US and Western Europe. The ease with which the cooperation of respondents can be obtained does, however, vary from one country or culture to another. In Eastern Europe, for example, interviewers are often regarded with considerable suspicion. This is a legacy from the repressive nature of former communist regimes. Similarly, as also noted earlier, referrals are required in order to obtain cooperation of interviewees in the Middle East, and interviews with women have to be conducted in the respondent’s home by female interviewers. Thus, personal interviewing is often the most effective method of questionnaire administration in international marketing research, especially among consumers. In emerging markets in some countries, personal interviewing may be mandatory. Even in other situations, low wage rates coupled with higher rates of response, improved quality of data and representativeness of the target population largely offset higher administrative costs. 9.2.4. Electronic Surveys The growth of the Internet has also opened up opportunities to use new technologies for primary data collection, particularly in survey research (Worldwide Internet Conference, 1999). Since these technologies require respondents with access to a computer and who are reasonably comfortable with this mode of interaction, they are primarily applicable for consumer surveys in countries with high levels of personal computer ownership (see Table 9.2). They are also commonly used in business-to-business research in developed countries and industries where businesses rely heavily on computerized purchasing and information transfer, for example computer software, banks and other financial services, medical equipment, pharmaceutical products, airlines, hotels and other travel-related products and services. E-mail or web-based surveys can either be conducted on a one-time or ongoing basis. A key requirement is the availability of a list of e-mail addresses. Where a topic- or product-specific list is not available, national, regional or global lists can be obtained from commercial sources such as Global Market Insite or Claritas, who maintain panels of Internet subscribers. Global Market Insite, for example, maintains 200 country panels in the Americas, Europe, Asia/Pacific and the Middle East. Data is collected in the local language using a fully integrated, netcentric suite of software. The respondent is pre-notified, and in the case of a web-based survey, provided where necessary with a toll-free number to access the survey online. E-mail and web-based surveys are generally considerably less expensive than surveys by telephone, mail or personal interview (Harrell, Clayton and Werking, 1999). In comparison to mail surveys, cost of postage, mail processing and editing are eliminated. Equally, interviewer costs associated with phone and personal interview surveys are eliminated as well as telephone charges for pre-notification and prompting. Once initial set-up costs for questionnaire development have been incurred, the sample can be expanded at little extra cost and large samples are feasible. Accuracy is improved as respondents are in direct and immediate communication (Steffensen, 2004). Impossible or unlikely values, irrelevant answers and multiple answers to a question with a single answer can all be eliminated. Web-based questionnaires are extremely versatile. Structured questionnaires of any form, length or layout can be developed. They have the advantage that product details, pictures of products, brand and shopping environment can be provided as well as links to other sites. Graphics, sound and video can also be integrated, creating a more stimulating and realistic research environment for the respondent. Links can be set up to guide respondents from one part of the questionnaire to another where questions are not applicable to the respondent. In ongoing surveys items can be changed or eliminated without difficulty. Data are more timely and often responses are more reasoned and objective, especially in relation to open-ended questions, since people tend to think more before typing in a response. E-mail and web-based questionnaires are also less intrusive since people can respond in their own time. Equally, as noted earlier, both qualitative and quantitative techniques can be combined in the same questionnaire. The whole process is automated from the posting of the questionnaire to response, thus minimizing recording errors. Responses can also be monitored and readily checked as they are received. On the other hand, the sample is limited to those with an e-mail address and the response is not always high, especially where the questionnaire is viewed as ‘spam’ (unsolicited e-mail) by the recipient. Consequently, pre-notification, alerting the respondents and identifying the sponsor is desirable. Where e-mail questionnaires are sent as an attachment, system compatibility may pose some problems, especially in downloading, completing and returning questionnaires. Technical issues may also daunt some respondents, resulting in nonresponse bias. Consequently, use of an incentive and a contact for technical assistance are both desirable. Valid results will, however, only be achieved if response levels are adequate and there is no nonresponse bias. Response rates are typically quite low, often in the 0.5–2% range (Agrawal, 1999). Consequently, a large number of surveys need to be sent out to obtain a large enough sample to analyze. While lower cost and rapidity of response make this method attractive for use in international research, potential bias problems suggest that such methods should be used with caution. 9.3. Field Staff Organization and Training A key problem in data collection in international markets is the need for reliable and high-quality local research staff and local research organizations. As noted earlier, large international research organizations have branch offices and in some cases field staff. Other organizations buy or outsource field services from local research organizations. In some cases, there may be limited choice in the number of local field organizations available. This gives rise to issues of coordination and harmonization of fieldwork, as well as issues relating to interviewer training. 9.3.1. Organization and Coordination of Fieldwork An important issue in conducting international surveys is how to ensure that the same standards of data collection are used, and that the fieldwork is comparable and of the same quality worldwide. ESOMAR has established guidelines for fieldwork in an effort to further professionalize data collection and harmonize fieldwork standards. These relate to interviewer recruitment, interviewer training, survey control, quality control and back checking. These guidelines apply to business-to-business, consumer and retailer research. All member organizations of ESOMAR are required to conform to these standards and to the ICC/ESOMAR International Code of Marketing and Social Research Practice. (See Chapter 14 and the ESOMAR web site, www.esomar.org, for more detail.) Establishment of these guidelines helps to ensure that local supplier organizations that are ESOMAR members operate according to common, consistent and agreed quality standards. These standards have been developed for personal or face-to-face interviewing and telephone interviewing, as well as for group discussions and in-depth interviews in qualitative research, and retail audits and store checks. They are, however, very general. Consequently, many research organizations, especially where they purchase fieldwork outside the company or from a series of agencies, set up their own checks and controls over the quality of fieldwork. Where the research budget permits, an experienced research manager can visit the country or countries where the fieldwork is being conducted to brief the interviewers, conduct or supervise one or more pilot interviews, and in some cases monitor some field interviews. In addition, independent back checking of 10-20% of the interviews is typically conducted. In some cases where a company is dealing with a new supplier, it may check up to 50% of the interviews. In the case of firms that have local branch organizations, training programs for field managers are often set up and regular meetings of local research staff held to harmonize procedures and fieldwork standards. This is critical in relation to qualitative research due to the somewhat subjective nature of data interpretation and so on. Research International Qualitatif, for example, holds regular meetings and training programs for staff from its branch offices around the world, to discuss new ideas and uses of different projective techniques, and to compare experiences with various techniques. This not only helps in harmonizing techniques and their interpretation, but also provides a forum for discussing new techniques and improvements in existing techniques, as well as current trends in qualitative research. Where firms do not have local branches and either outsource interviews or employ local interviewers, they typically send an executive to set up the project and brief interviewers. Where feasible, the executive should be fluent in the appropriate language to brief interviewers on the project, be able to monitor a number of interviews, provide feedback and so on. Typically, local interviewers are bilingual. Where there is verbatim response, interviews can be translated and sent to the head office, where interpretation is conducted centrally, to ensure consistency across countries. 9.3.2. Training Interviewers Another issue relates to the training of field interviewers. This may be problematic in some countries where there is limited experience or history of market research, for example Eastern Europe. Again, ESOMAR has established guidelines for interviewer training in both personal or face-to-face interviewing and telephone interviewing. For face-to-face interviewing a day’s training is required, covering issues relating to the ICC/ESOMAR code of conduct, how to approach a respondent, how to cope with refusals, how to conduct an interview and ask certain types of questions. For telephone interviewing only half a day’s training is required. The guidelines are somewhat general, but establish a standard for minimum training requirements. It is important that interviewers should understand the need to establish a good rapport with the respondent and create a good climate for interviewer–respondent interaction. Here, the manner in which the interviewer introduces him- or herself is often a major factor in developing this rapport. Questionnaire administration also needs to be clearly understood, with the application of standard procedures and instructions. Introduction of any type of bias, such as the generation of inferences that a certain type of response might be preferred, needs to be avoided. In essence, the interviewer should be taught to remain as neutral as possible and to avoid interjecting any personal opinions. This is important in developing countries, where respondents are not familiar with surveys and will only respond if they feel comfortable with the interviewer. In some cases, establishing rapport while remaining neutral requires a delicate balance. A study in South Africa, for example, found that local interviewers tended to help and guide respondents rather than leaving them to follow the instructions in the questionnaire. As a result they influenced the respondent and showed how a task should be undertaken. For example, they might place the Funny Faces in order rather than let respondents do the task alone. Practice interviews can be helpful in order to ensure that interviewers have learned appropriate skills. Development of a detailed study protocol can be advisable when using interviewers recruited locally for a specific project. Recruiting interviewers locally may be helpful in surveying local ethnic groups and communities, as locals may more easily develop rapport with interviewees than trained interviewers, especially on sensitive topics such as health issues. However, careful attention to selection, training and provision of ongoing support will be needed. For example, a study of the maternity care of recent mothers in Australia from three different ethnic groups, Turkish, Vietnamese and Philippino, recruited interviewers from their local communities to conduct interviews in hospital (Small et al., 1999). Extensive training of the interviewers was provided, including developing skills in recruiting and interviewing, discussing cultural issues relating to the study, help in framing questions, discussing protocols relating to situations that might arise in the interview, and piloting and translating instruments used in the study. Even in industrialized countries, attention to interviewer training and to briefing and debriefing is essential to ensure maximal response rates and to avoid bias arising from the interviewer– respondent interaction. Particularly where the interview involves open-ended or complex questions and tasks, such as projective techniques or multidimensional scaling, attention to interviewer training is essential. Often this is key to ensuring the quality of the research. 9.4. Summary In determining appropriate sampling and survey administration procedures, a number of factors have to be taken into consideration. First, the relevant unit or level from which the sample is to be drawn has to be determined. Next is the selection of the technique to be used in drawing the sample. Related to this is the choice of survey administration procedures. In both cases, the impact of such decisions on the comparability and equivalence of the data from one unit to another has to be carefully considered, and weighed against the cost effectiveness of alternative plans and procedures. In drawing the sample, a number of different levels or units may be considered, including the world, country groupings, countries or groupings within countries. In making this decision, much depends on the purpose of the survey, the specific product or service concerned and the target segment, as well as the availability and adequacy of various sampling frames or lists. Surveys concerned with industrial products or with upscale mobile target segments, such as business people or foreign travelers, are more likely to sample at a global or regional level than surveys concerned with consumer packaged goods aimed at a mass market. The choice of technique to be used in drawing the sample is closely related to the level at which the sample is drawn. In many countries, especially developing countries, use of systematic random sampling techniques is likely to pose some difficulty. This is due to the lack of sampling lists, street maps and guides, and to sprawling urban developments and scattered rural populations. Consequently, convenience or judgment sampling procedures may have to be used in order to avoid excessive administration costs. In selecting among survey administration procedures – that is, telephone, mail, personal interview or electronic – careful attention has to be paid to the efficacy of each procedure in reaching the sample population, as well as the potential sources of bias associated with each. In many countries, especially the developing countries, telephone surveys will only tap a relatively limited segment and typically only short questionnaires can be administered by this means. Mail surveys are only effective in countries with high levels of literacy, or where they are in relation to the literate population. Personal interviewing is the most costly and time-consuming method of survey administration in cross-national research. This does require training of competent interviewers in order to minimize bias arising from the interviewer and interviewee interaction. Electronic surveys provide quick results over broad geographical areas. As rates of Internet penetration increase around the world, this method of data collection will become more common in international surveys. In brief, differences in market characteristics and the research infrastructure from one country to another imply that the development of the sampling plan and survey administration procedures may entail more creative thought and effort than in the case of domestic marketing research. Means may have to be devised to develop and identify appropriate sampling lists and techniques, and to administer surveys without incurring excessive costs in conditions where the available infrastructure is extremely sparse. In addition, attention has to be paid to the issues of comparability from one sampling unit to another, and to minimizing potential sources of sampling error. 10. Multicountry Scales At the core of much multicountry research is the development of operational measures of a construct to be examined in more than one country. Often a researcher is interested in testing whether a construct developed in one country holds in another country, or in examining similarities and differences in constructs across countries or geographical areas. Typically, an operational measure of the construct has already been developed in the base country and the task facing the researcher is to see whether the construct can be meaningfully measured in the same way in another country. Many of the conceptual issues and difficulties in establishing construct equivalence were examined in Chapter 5. However, operational measurement of the construct presents its own set of issues and difficulties. Constructs are often complex and multifaceted phenomena and require a multiple-item scale for adequate measurement. In addition, as noted earlier, constructs are not necessarily expressed in the same way in different countries. To complicate matters further, methods of measurement and measurement instruments are subject to different types of bias in different countries. Procedures for developing a scale to measure an underlying construct in a single country are relatively straightforward and well understood (Churchill, 1979). Developing a scale in a multicountry environment is considerably more complex and challenging and presents the researcher with two intertwined issues. The first and most fundamental question is whether the same construct exists in different countries. A particular construct identified in one country may not exist in another country or may not be expressed in the same terms. Consequently, attempts to assess the universality of a construct need to allow for the possibility that it may not take the same form or have different elements in other countries. Given this difficulty, some external criteria are needed to assess validity, rather than relying simply on internal criteria, such as reliability and convergent and discriminant validity. Once the underlying validity of the measure has been established, the second issue is to assess how individuals in a given country fall on the construct. Individuals in one country may score much higher, lower or the same as individuals in another country. Here, a key problem is whether this score reflects real differences in response or some form of metric bias. However, the issues only become relevant once the validity of the construct has been established. This chapter examines the various issues in developing scales to be used in multicountry research. A number of issues, such as types of data, assessing whether different scale items function in the same way and whether measures are reliable, pertain to all approaches to scale development and are covered first. To illustrate the different approaches to scale development, examples of how scales have been applied cross-culturally are discussed in detail. Finally, an approach to developing ‘decentered’ cross-cultural scales is suggested. 10.1. General Issues in Scale Development 10.1.1. Types of Measures A critical first step in developing a measure to be used in international marketing research is to select the type of response scale. The construct is an abstract concept that exists in the individuals’ minds. The first stage is to devise an instrument that allows respondents to express that construct verbally, or on paper, or directly in electronic form. The challenge is to develop an instrument that accurately captures the construct, expressing salient elements in a simple, familiar fashion. Responses can be collected in a variety of ways, but ultimately they reduce to the type of items used. These vary in terms of task difficulty and, hence, ease of use and reliability in multicountry research. Based on their underlying properties, three alternative types of items can be used in data collection: (1) nominal; (2) ordinal; and (3) interval. There is also a fourth type, ratio, but this is rarely applicable when collecting data from individual respondents. 10.1.1.1. Nominal Measures Nominal or categorical measures are often the easiest and seemingly most unambiguous means of collecting information in multicountry research. The respondent or interviewer simply indicates the presence or absence of a characteristic. For example, information about gender, occupation, ethnicity and type of dwelling unit are all collected using nominal measures. These can be used to categorize individuals or groups and may be useful in order to split the sample into groups and compare responses of different groups. Nominal measures can also be used to develop scales of economic development or standard of living using a Guttman scaling procedure (Manfield, 1971). The Guttman scale requires use of nominal data and is based on presence or absence of a number of objects. If the construct is standard of living, then ownership or nonownership of appliances such as radios, televisions, washing machines, refrigerators and so on would be the input to the scaling routine. Nominal measures are the simplest type of measure and place the least burden on the respondent. They are appropriate for illiterate respondents or those with low levels of education. The respondent simply has to decide whether or not the characteristic or category applies. Such measures do, however, require that the definition of a category is unambiguous and familiar to the respondent. Also, as discussed in Chapter 8, pictures or illustrations may be used with less-literate respondents. The major limiting factor of using nominal data is that, with rare exceptions, categorical data do not lend themselves to scale development. It is difficult to aggregate nominal data as the categories or objects, even within a country, are not comparable. 10.1.1.2. Ordinal Measures While nominal measures provide information on category membership, ordinal measures indicate whether an object is greater than, less than or equal to some other object. The most direct way to collect ordinal data is to ask respondents to rank order objects in relation to some attribute. Respondents could be asked to rank order a set of brands or rank order five different leisure-time activities from most preferred to least preferred. For respondents in most developed countries this is a relatively simple and straightforward task. However, when the research is conducted among less-literate populations, physical stimuli may be needed. As discussed in Chapter 7, such stimuli should be familiar to the respondent to ensure meaningful responses. Less-literate respondents will have difficulty reading the instructions or recognizing the written words for the different stimulus objects. Consequently, questions will need to be interviewer administered and physical stimuli used rather than words. Ordinal measures are relatively easy to collect. Even respondents with low education levels typically have little difficulty ranking objects, or expressing preferences in terms of rank ordering. Such data can be used as inputs in multidimensional scaling. However, unless the same set of objects is ranked in each country, rankings have limited use for comparative purposes. Ranking of one set of objects cannot be readily combined with that of another set to form a combined scale. Also, rankings between countries can only be compared if the same set of objects exists in both countries. When the same set of objects, activities or statements can be ranked in more than one country, rank order correlations can be computed to assess the degree of agreement in rankings between countries. However, as with ordinal data, there are limited instances where ordinal data can be used to construct a scale. When only two objects are rank ordered at a time, the procedure is referred to as paired comparison. Respondents are asked to indicate a preference or similarity judgment for two objects at a time. Data collected in this fashion can be used as input for multidimensional scaling routines. Ordinal or paired comparison data can also be used to create a unidimensional interval scale using Thurstone’s Case V Scaling model (Thurstone, 1959). An interval scale is created based on the extent to which certain stimuli dominate others. For example, if A is preferred to B 51% of the time, there is little perceived difference between the two stimuli. However, if A is preferred to B 90% of the time, there is a much greater perceived difference between the two stimuli. These relationships are used across all stimuli to create an interval scale that reflects the perceived difference between objects. 10.1.1.3. Interval Measures Most data collected in multicountry research is based on interval measures, or data assumed to be interval scaled. Technically, interval measures allow direct comparison of different positions on a scale. A frequently given example is that of temperature. For example, the temperature on an August day might be 97°F in New Delhi and 91°F in Beijing. Based on this, one can conclude that the temperature in New Delhi was 6°F warmer than in Beijing. The same data can be converted to °C and the same conclusion would be reached, although the absolute difference would be smaller. In addition, measurements of temperature from other cities in other countries at different times or seasons would be directly comparable to the data already obtained. For example, the temperature in Moscow might be 71°F, the temperature in Santiago, Chile 55°F, and the temperature in Singapore 88°F. Someone planning a trip to any of these cities would only have to compare their home climate with the city they were visiting to select appropriate clothing. In this example, an instrument used to measure the temperature in one city could be used in another city and would provide the same result. Thus, any measurement of temperature would be comparable. However, this is typically not the case when the same measurement instrument is used to collect questionnaire data in different countries. In multicountry research, responses to 5-point (or n-point) scales are often treated as interval data. Mean values are computed, comparisons made between countries and sophisticated analyses performed on the data. However, it is important to recognize that the underlying data do not possess the same properties as true interval data, as the distance between points may not have the same meaning in different countries. Consequently, comparability across countries is open to some question. A mean response of 4.3 on a 5-point scale in one country may or may not be equivalent to a mean response of 4.3 on the same question in another country. Also, the task presented to the respondent is more difficult and more abstract than for nominal or ordinal data. Collection of interval data requires a greater degree of sophistication and literacy among respondents. Also, providing meaningful responses to a 5-point scale that ranges from ‘strongly agree’ to ‘strongly disagree’ requires some familiarity with conventions used to collect these types of data. However, as the examples in Chapter 8 illustrate, with a little creativity in stimulus development these types of data can be collected among less-literate populations (see Donovan and Spark, 1997). 10.1.2. Differential Item Functioning The building blocks of any scale are individual items that together measure an underlying construct. The critical issue in international marketing research is whether the individual items function in the same way from one country to another. Unless the individual scale items measure the construct in the same way in each country, the sum of the scale items will not properly reflect the construct. For example, consider a scale that measures consumer ‘innovativeness’ in the US. It might employ 16 different items that are measured with a 5-point Likert scale. Each item taps a different aspect of innovativeness. However, these items may not necessarily function in the same way in other countries or cultures. The problem becomes that of arriving at a set of items that accurately and adequately measure the construct in multiple settings. An important aspect of this approach is that the researcher assumes that the same construct exists and can be measured in a similar but not identical fashion. Three approaches can be used to deal with differential item functioning or item bias: analysis of variance, the Mantel-Haenszel statistic and item response theory. The first approach involves applying the standard analysis of variance model. Respondents are divided into groups that have the same scores on the overall scale. With the scale for innovativeness the total score could range from 16 to 80 (sixteen items scaled one to five). Respondents with the extreme scores, 16 and 80, would not be included since individuals giving those responses would have responded the same in all cultures. Groupings of respondents would be formed using the remaining scores. Based on the response to the scale on innovativeness, respondents could be grouped into seven mutually exclusive categories. When the responses of the individuals from one country are plotted against the mean of another country, it is relatively easy to see whether there is any item bias present. Figure 10.1 illustrates situations where the plot reveals no bias (a), uniform bias (b) and nonuniform bias (c). Uniform bias (b) is the easiest to interpret in that respondents in one country consistently score higher or lower on a construct, in this case innovativeness, than individuals in another country. Nonuniform bias (c) is more difficult to interpret in that over part of the range some respondents score higher, and over part of the range they score lower than others. Figure 10.1 Hypothetical examples of (a) an unbiased item; (b) an item with uniform bias; and (c) an item with nonuniform bias Analysis of variance can be used to arrive at an analytical solution to differential item functioning. Country would be one factor in the analysis of variance and the score groupings, or level, the other factor. Significant differences in the score grouping are to be expected since each grouping is based on score differences and hence ignored. A significant country main effect would indicate the presence of a uniform bias. A significant country by level interaction would indicate a nonuniform bias. Depending on the extent and degree of bias, the researcher may decide to eliminate certain items, resulting in a reduced set of items. If too many items are eliminated it is also possible to address the problem in an iterative fashion, eliminating one variable at a time, starting with the most statistically significant first and proceeding until none is significant (Van de Vijver and Leung, 1997). The Mantel-Haenszel statistic uses as input dichotomous data and allows for pair-wise comparisons only. This is particularly problematic when a large number of countries are studied, as the number of pair-wise comparisons needed increases rapidly. Item response theory (IRT) is also typically applied to dichotomous variables. As a starting point, IRT assumes that a scale is unidimensional. If the scale is multidimensional, then each component needs to be examined separately. Item response theory identifies three parameters that need to be satisfied to determine whether an item is unbiased across countries. The three parameters are: (a) item discrimination; (b) item difficulty; and (c) the lower asymptote. The three parameters are illustrated in Figure 10.2 as they relate to item characteristic curves. Item discrimination refers to the ability of the item to discriminate between individuals who score differently on the underlying latent trait, for example an attitude, value or belief. Differences in the discriminant ability of an item across two countries indicate nonuniform bias (Figure 10.2a). Item difficulty, which is a major concern in assessing educational test scores, relates to the ability of an individual to respond correctly to an item. In other words, whether individuals who have high scores on a latent trait underlying the test, either an ability measure or an attitudinal measure, have high scores on a particular item. Differences in the item difficulty parameter suggest uniform bias (Figure 10.2b). The third parameter, lower asymptote, relates to the probability of guessing a correct response and pertains to ability tests. Figure 10.2 Hypothetical item characteristic curves of items that differ only in (a) the item discrimination parameter; (b) the item difficulty parameter; and (c) the lower asymptote (a) 1 - In applying item response theory to multicountry attitudinal and belief data, a two-parameter model consisting of item discrimination and item difficulty is most appropriate. The third parameter, lower asymptote, is not meaningful in this context, since guessing per se is not an issue. The model is estimated for each country to identify biased items. These are identified through the use of item characteristic curves and a chi square test (Lord, 1980). Items that are biased are deleted from the scale before comparisons are made between countries. For more detail on the procedures for detecting and eliminating item bias, see Van de Vijver and Leung (1997, pp. 62–88). Differential item functioning and item bias are closely related to measurement equivalence. In general, bias will tend to lower or may even preclude equivalence. Bias at the construct level – that is, the construct measured is not the same across countries or groups – precludes equivalence at any other level. Equally, bias at the method level precludes equivalence at the item level, while bias at the item level means that items will be nonequivalent. Eliminating items that function differently between two countries does not necessarily lead to elimination of differences in the average scores between groups (Poortinga and Van der Flier, 1988). However, taking out biased items does ensure that the score differences between two or more countries are free from item bias. The differences that remain suggest actual differences between countries. As noted earlier, when a large number of items have to be removed from the scale, this raises the more general question of whether the scale can be applied in more than one country. The application of item bias detection by researchers has revealed some interesting problems. First, it is often difficult to explain why a particular item is biased. Second, use of different statistical techniques suggests different results. Third, item bias statistics are not stable in test–retest studies or in cross-validations. Most problematic for those developing multicountry scales are low levels of agreement between expert judgment of biased items and statistical methods (Van de Vijver and Leung, 1997, pp. 84–88). Often researchers developing or modifying a scale for use in another country will use panels of experts to select or modify scale items. The lack of agreement between statistical approaches and judgmental approaches suggests that reliance on experts may not provide a scale that is unbiased in another country. This conclusion should, however, be tempered by the observation that different statistical techniques do not necessarily agree with one another, and that they show limited consistency when repeated a second time. 10.2. Reliability Issues in Scale Development Scales used to measure constructs should be reliable across all countries in which they are administered. Reliability, as it relates to construct measurement, was discussed in Chapter 6. These issues are particularly important in countries or contexts where little research has been conducted, or with which the researcher has little prior experience. In these cases, the reliability of different types of data or measures may not be well documented and, particularly in the case of attitudinal data, will need to be verified and established. There is a requirement to understand the limitations and possible errors associated with different types of measures. Furthermore, reliability does not exist in isolation, as the nature of the instrument and the method of data collection can influence the reliability of results. Linguistic and conceptual nonequivalence in measurement instruments used in cross-cultural surveys can produce differences in measure reliability. This poses a threat to the validity of conclusions reached. Examination of reliability, while costly and time consuming, is nonetheless critical and attention should be paid to include reliability checks as standard procedure in multicountry research. When examining reliability - in contrast to measure validity or equivalence - attention is centered on whether the same result is obtained when a measure is repeated in a different context, fashion or time. Despite efforts to design an instrument that is adapted to all countries and cultures, it may not be equally reliable in all contexts. Different types of measures, such as attitudinal, lifestyle and other measures, also vary in their level of reliability. The stability of data over time may also vary. It is thus important to compare the reliability of data obtained in different countries or contexts, since this may attenuate the precision of estimation and reduce the power of statistical tests. 10.2.1. Scale Dimensionality When a scale developed in one country is used in another country, an issue is whether the scale has the same number of dimensions in both countries. Differences in the number of dimensions begin to suggest a lack of comparability in the construct between countries. For example, if a scale is initially developed in one country to measure a construct and the construct is unidimensional, subsequent use of the scale in other countries should also reveal one dimension. If more dimensions were found, one would conclude that there is a lack of comparability between the two countries. The CETSCALE (for more details see later) has generally been found to be unidimensional (Netemeyer et al., 1991; Druvasal et al., 1993). However, Douglas and Nijssen (2003), using principal components analysis with varimax rotation, found two distinct dimensions in the CETSCALE in the Netherlands. The first factor explained 47.5% of the variance, contained all but two of the questions and corresponded to the core elements of consumer ethnocentricism. The second factor explained 11.1% of the variance and captured an element that may be unique to the Netherlands or a smaller market with relatively few domestic manufacturers of consumer durables. The two questions that loaded on the second factor were: ‘Only those products that are unavailable in the Netherlands should be imported’ and ‘We should buy from foreign countries only those products that we cannot obtain within our own country.’ The stability of dimensionality of the Maslach Burnout Inventory (MBI) was examined in a review of 35 studies that used it cross-culturally (Hwang et al., 2003). The original scale (Maslach and Jackson, 1986) had three dimensions and was designed to measure the burnout of individuals in various occupations. There are 22 items that are used to form the three scales: (1) emotional exhaustion; (2) depersonalization; and (3) personal accomplishment. Almost all the studies (91.8%) in the other countries identified a similar three-factor structure. However, only one study replicated the three factors with a similar pattern of strong loadings on the three primary factors. In addition, four studies identified four factors and three studies identified only two factors. This suggests that while dramatically different results will not be obtained when using scales in different countries, one must be prepared for differences that need to be explained. Further, a different dimensionality suggests that the construct is in fact different in the other countries. Another scale that has been widely used in the US as a measure of psychological wellbeing is Bradburn’s (1969) Affect Balance Scale (ABS). Macintosh (1998, p. 83) asserts, ‘If cross national research is to be meaningful, the validity of measuring devices must be demonstrated. Specifically, it must be shown that measuring devices developed primarily in the United States are consistently applicable in other cultures and regions.’ He undertook a cross-cultural assessment of the ABS and found that the full model did not fit, based on four different goodness-of-fit measures, in any of the 38 countries examined. Further, attempts to develop alternative models did not find any models that fit across the countries studied. The lack of fit was consistent with concerns that had been raised about the ABS over the past 20 years (see Macintosh, 1998). Some recent applications of the NATID scale further illustrate the difficulty of achieving the same factor structure with the same scale across multiple countries and time periods. Huntington (1993) identified four components of national identity: religion, history, customs and social institutions. These components guided development of the NATID scale by Keillor et al. (1996). Starting with 70 items, the 17-item CETSCALE and 53 new items, they used confirmatory factor analysis to reduce it to 17 items, including five from the CETSCALE. This was then applied in the US, Japan and Sweden (Column 1 of Table 10.1). Cui and Adams (2002) applied the NATID in Yemen. They added four emic items and used exploratory factor analysis and confirmatory factor analysis to evaluate and refine the scale. They tested two models, the original items (columns 2) and the expanded scale (column 3). To achieve the best fit they had to drop five items and restructure the four dimensions. Thelen and Honeycutt (2004) examined the NATID scale in Russia and also found four dimensions, but had to reassign one variable and relabeled two of the dimensions. Across the four uses of the NATID scale shown in Table 10.1 there were always four dimensions, but not necessarily the same four. The number of items in the scale varied from 12 to 17 and only four of the items were consistent across all four analyses. The construct of national identity clearly exists, but how it is measured appears to vary considerably from country to country. In a study of the Generalized Perceived Self-efficacy scale in 22 countries, Schwarzer and Scholz (2000) found that the unidimensionality of the scale held across all countries. There were differences in scores on the scale between countries and also by gender, with men typically scoring higher. An earlier study by Schwarzer and Born (1997) also found that the scale’s unidimensionality held across 13 different cultures. The scale was initially developed in German and subsequently translated into 26 different languages. More information on the scale and the different language versions is available at www.RalfSchwarzer.de. The dimensionality of a scale, and by implication the underlying construct it is measuring, is relatively straightforward and fairly unambiguous in its interpretation. Less evident is why the difference was observed and precisely how to interpret it. If the dimensionality is the same, greater confidence can be placed in the use of the scale in another country. Analytic approaches for looking at the comparability of scales are covered in the next two chapters. In the remainder of this chapter, examples of scale use in multiple countries are covered as well as suggestions for developing scales. 10.2.2. Using Multi-item Scales in Cross-cultural Research When using multi-item scales in multi-country research, a common procedure is to take a scale that has been developed in one country or context, translate it and then administer it in a number of countries, with relatively limited consideration of its validity or equivalence in other countries or contexts. This approach is based on the assumption that the underlying construct is both relevant and present in other countries. An additional assumption is that it can be measured using the same instrument. In some cases the internal consistency is examined using Cronbach’s alpha. Where high alphas are obtained, the scale is considered appropriate and applicable in that context. Increasingly, nomological or structural validity is also examined; that is, the relation of the measure to measures of other related constructs. Development of context-specific measures of constructs and examination of how and whether a construct is manifested in the same terms in other countries or contexts is rare. Table 10.1 Comparison of models, included items and dimensions Item Keillor etal.1996 Cui and Adams 2002, Model 1 Cui and Adams 2002, Model 2 Thelen and Honeycutt 2004 N1. Important people from the country's past are admired by people today. NH CR CR NH N2. One of country X's strengths is that it emphasizes events of historical importance. NH BT — NH N3. The country X has a strong historical heritage. NH — — — C1. A country X person possesses certain cultural attributes that other people do not possess. CH CH — NO C2. Country X citizens in general feel that they come from a common background. CH CH CH NO C3. Country X citizens are proud of their nationality. CH BT CH NH C4. People frequently engage in activities that identify them as country X citizens. CH CR CR NO C5. Country X citizens are proud of their [ethnic] and [religious] roots. — — CH — C6. One of the things that distinguish country X citizens from other countries is its traditions and customs. — — BT — B1. A specific religious philosophy is what makes a person uniquely a country X citizen. BT BT CH BT B2. It is impossible for an individual to be truly a country X citizen without taking part in some form of religious activity. BT — — BT B3. Religious education is essential to preserve the cohesiveness of the country X society. BT BT BT BT B4. A specific religious philosophy is not an important part of being an American, (reverse-coded) BT — — — B5. A true country X person would never reject his or her religious beliefs. BT BT BT BT E1. We should purchase products manufactured in country X instead of letting other countries get rich off of us. CE CE CE — E2. It is always best to purchase country X products. (Russian products, first, last and foremost) CE — — CE E3. Country X's citizens should not buy foreign products, because it hurts the country's business and causes unemployment. CE CE CE CE E4. It may cost me in the long run but I prefer to support country X products. CE CE CE CE E5. Only those products that are unavailable in country X should be imported. CE — — CE Notes: NH = national heritage, CH = cultural homogeneity, BT = belief tradition, CE = consumer ethnocentrism, CR = cuдtural heritage, NO = national homogeneity. Source: Thelen and Honeycutt, 2004. In this section alternative approaches to applying a scale developed in one cultural context to another are examined.* * Hambleton (1994) summarizes 22 guidelines formulated by an international committee for the translation and adaptation of psychological and educational instruments. The guidelines cover: (1) the context; (2) instrument development, translation and adaptation; (3) administration; and (4) documentation and score interpretation. These include assessment of single or context-specific constructs, single-context scale development, and cross-cultural assessment of single-context scales as well as approaches to developing shorter versions of the original scale. In addition, ‘decentered’ scale development in multiple cultural contexts is examined to illustrate a more culturally balanced design, in which no one country or context dominates scale or measure development. 10.3. Single-context Scale Development A scale that has been widely used in multicountry and in country-of-origin studies is the CETSCALE. It was developed by Shimp and Sharma (1987) to apply the construct of ethnocentrism to marketing and consumer behavior. More specifically, it represented an attempt by the researchers to measure consumers’ orientation toward the purchase of foreign products. Consumer ethnocentrism is rooted in the original construct of ethnocentrism, or the attitude that one’s own group (race or people) is superior, first discussed by Summers (1906) almost 100 years ago. A scale to measure the construct of ethnocentrism, the California Ethnocentrism scale, was developed by Adorno et al. (1950) and is closely related to patriotism and political-economic conservatism. The construct had also been measured among different populations such as black college students (Chang and Ritter, 1976) and in the UK (Warr et al., 1967). The first step in the development of the CETSCALE was to take the construct of ethnocentrism and apply it to consumers’ thoughts about foreign-made products. More than 800 US consumers were asked to ‘describe your views of whether it is appropriate for American consumers to purchase products that are manufactured in foreign countries’ (Shimp and Sharma, 1987, p. 281). Content analysis of the statements led to the identification of seven different aspects of consumers’ orientation toward foreign products, one of which was identified as consumer ethnocentric tendencies. These seven dimensions led to the generation of 225 statements, which were later reduced to 180. Six judges then evaluated all 180 statements and classified them into the different dimensions. Only those statements that were categorized in the same way by at least five of the six judges and were not redundant were retained. This resulted in 100 statements. A questionnaire consisting of 117 Likert-type statements (the 100 plus 17 items from Adorno et al.’s original scale) was then mailed to 850 households. A factor analysis of the responses resulted in 54 items loading 0.5 or better on the relevant dimensions. These 54 items formed a new questionnaire of 7-point Likert statements that was mailed to almost 4000 households in four different areas of the US. Confirmatory factor analysis was then used to examine the five-factor structure identified in the early stages and to eliminate items that were not considered reliable. One of the five dimensions was rejected and only one of the remaining conceptual dimensions was strongly supported. Of the original 43 non-Adorno items, only 18 were retained as reliable and related to the four constructs. Twelve of these items loaded on the consumer ethnocentrism dimension and two variables loaded on each of the remaining dimensions. However, these remaining six variables were also highly correlated with the 12-item scale. Scale development proceeded with the 25 items that were reliable, but focused only on one dimension, consumer ethnocentrism. Confirmatory factor analysis was performed on the pooled data (all four geographical areas combined) and for each of the four areas separately. The 17 items that loaded 0.5 or greater were retained and formed the CETSCALE (Table 10.2). The reliability of the scale was first assessed using Cronbach’s alpha. This ranged from 0.94 to 0.96 over the four surveys. In an additional two-phase study the test–retest reliability was assessed. The correlation between the two administrations of the scale separated by five weeks was 0.77. It was possible to assess convergent validity for one of the samples since respondents had answered a question two years earlier concerning purchase of foreign products. The correlation between this response and the CETSCALE was 0.54. Discriminant validity was examined by correlating the CETSCALE score with the related constructs of patriotism, politico-economic conservatism and dogmatism. While these constructs were correlated with the CETSCALE (range 0.39 to 0.65), the authors believe that there is discriminant validity. Finally, nomological validity was established by looking at how the CETSCALE correlated with attitudes, purchase intentions and ownership of foreign-made products. 10.3.1. Cross-cultural Assessment of Single-context Scales The development of the CETSCALE described in the above section illustrates a systematic approach to measuring a construct in one country. However, a central issue is whether the same instrument can be used in other countries or cultures. In making this assessment, the typical approach is to take the scale, translate it, administer it in a number of countries and then assess its reliability and validity. This approach bypasses a number of important steps in scale development. For example, the development of the CETSCALE started with over 800 US consumers providing responses to an open-ended question that probed their attitudes toward buying foreign-made products. This step generally is not included in the typical cross-cultural assessment. Further, all the purification and refinement steps used to arrive at the final 17 items were performed on US data. This allows the very real possibility that if another country had been the base, a different scale would have evolved. Examining the fit of a scale developed in one country in another country should be understood in terms of the types of inherent biases interjected. Comparing results obtained in country B using a scale developed in country A interjects a pseudo-etic bias (Triandis, 1972) and skews results toward commonality. It is also unlikely to lead to the same findings when the same procedures are followed to develop a scale to measure the construct in country B and then the results of the two scales (country A with country B) are compared. In particular, if the construct is not the same it is unlikely that any comparison is feasible or meaningful. The reliability and validity of the CETSCALE were examined in four countries by Netemeyer et al. (1991). In addition to the US, the CETSCALE was administered to students in France, Japan and West Germany. The sample sizes in the four countries ranged from 70 to 76. The English-language version of the questionnaire was first translated in the other three languages. The translated version was back translated and pretested on a small sample of US students as well as small samples of students from the three countries who were studying in the US. In addition to the 17-item CETSCALE, questions relating to attitudes toward purchasing foreign products, beliefs about the quality of certain products from foreign countries and preferences for products in general from those countries as well as cars and television sets were also administered. The results of a confirmatory factor analysis showed that the scale had a unidimensional factor structure across all four countries. Factor structure invariance was examined using two methods, multigroup analysis and the coefficient of congruence. Both approaches showed that the factor loadings were invariant across countries. The reliability of the scale was assessed in three different ways. First, composite reliability coefficients (Fornell and Larker, 1981) showed a high degree of reliability (0.91 or higher) in the same range as those obtained when the scale was originally developed in the US. Variance extracted estimates, which measure the amount of variance due to a construct’s measurement relative to random measurement error, were also calculated. Consistency was found across the four countries, although the levels in the French and Japanese sample were lower. Finally, corrected item-to-total correlations were calculated. These were higher for the US sample, but the ranges were similar for the other three countries. Collectively, the three measures suggest that the CETSCALE is reliable, not only in the US but in the other three countries studied. Discriminant validity, or the extent to which the CETSCALE is distinct from another construct, attitude toward home country, was assessed by three different methods. One-factor and two-factor models were estimated. In the one-factor model a unity correlation is assumed and in the two-factor model the correlation between the two scales is freely estimated. The fit of the two-factor model was better, suggesting discriminant validity. Second, the variance extracted estimates were greater than the squared parameter estimate (phi squared). Finally, the correlation between the two constructs was significantly < 1. Nomological validity was assessed by examining the correlation between the CETSCALE and the importance of buying foreign products, attitudes toward buying domestic products in general, attitudes toward buying foreign products in general, as well as attitudes toward buying specific products. While the overall pattern supports the nomological validity of the CETSCALE, the relationships were strongest for the US and weaker for the other three countries, with the least support in the former West Germany. More recently, the sensitivity of the CETSCALE to the context in which it is administered has been examined (Douglas and Nijssen, 2003). Earlier studies examine the CETSCALE in industrialized countries with large domestic industries and relatively strong feelings of national identity and patriotism. This study examined the CETSCALE in the Netherlands, a relatively small country with no or few domestic brands in many consumer durable product categories. As noted earlier in this chapter, their study revealed a two-dimensional factor structure rather than the single dimension found elsewhere. The first dimension captured core elements of consumer ethnocentricism, while the second dimension related to a distinction between products that were not available from domestic firms. In addition to finding a two-dimensional factor structure, the intensity of consumer ethnocen-trism was lower in the Netherlands than was observed in studies in other countries. The study also compared two different translations of the CETSCALE into Dutch. The first, a literal translation, followed typical translation protocols of back translation. Based on in-depth interviews dealing with the construct of consumer ethnocentricism, its meaning to Dutch consumers, as well as the meaning of different phrases and items, a second version of the questionnaire was created. This modified version was found to provide a better fit of the data as well as a lower correlation between the two dimensions of the CETSCALE found in the Netherlands. This study suggests that attention needs to be paid to contextual factors that may influence observed results. If there is a high degree of contextual similarity, then greater confidence can be placed on results that show similarity in responses. If there are differences in responses, it may reflect true differences, or differences brought about by differences in contextual factors. Further, the study suggests the importance of using qualitative research first to more fully understand the construct and the instrument used to measure it initially. 10.3.2. Examining Related and Context-specific Constructs The above description of the cross-cultural assessment of the reliability and validity of the CETSCALE provides a good illustration of the approach typically adopted by researchers. As indicated at the outset, it answers the specific question of whether a scale developed in one country exhibits the same properties when it is used in another. Assessment of the reliability provides reassurance that respondents are answering the questions in a consistent manner. Discriminant validity answers the question whether the scale is distinct from other constructs in the other countries. Nomological validity looks more broadly at the question of whether the scale measures the construct in the other countries. However, the approach to assessing discriminant and nomological validity does not address the more fundamental issue of whether a scale developed de novo in the foreign country would be the same as the one being tested. The complexity of establishing equivalence of constructs in multicountry research can be seen by examining the relation between the CETSCALE and an Animosity scale in the People’s Republic of China (Klein et al., 1998). The context for the study was Nanjing, China, a city that suffered during Japanese occupation from 1931 to 1945. This provided a respondent population that was likely to have residual feeling of animosity toward the Japanese. While the CETSCALE captures beliefs about buying foreign products in general, the Animosity scale is designed to measure attitudes toward a specific country. The Animosity scale has two major dimensions, Economic Animosity and War Animosity, both of which were administered relative to the Japanese (Figure 10.3). As expected, the Animosity scale and the CETSCALE were correlated (path coefficient 0.50). Interestingly, while both scales influenced willingness to buy, only the CETSCALE was related to product judgments. The Animosity scale and CETSCALE were also used to examine Dutch consumers’ attitudes toward Germans and German products (Nijssen and Douglas, 2004). One variation was to study two product categories, televisions, where there was a Dutch manufacturer, and cars, where there was none. The Netherlands provides a very different context from China in that it is a relatively small market with an open society. The findings were similar to Klein et al. (1998) with some differences in the product evaluations. In particular, German cars were evaluated more favorably than German televisions. These studies suggest that while a scale developed in one country (i.e. the CETSCALE) may be useful in another country, there may also be unique facets or constructs operating in the other country. Thus, it may be desirable to develop context-specific scales that are capable of measuring the indigenous constructs that affect the behavior of interest. 10.3.2.1. Developing Shorter Scales Scales originally developed to measure a construct are often quite long. In some cases it is desirable to shorten them to focus on core elements. Further, unlike initial efforts to develop a scale, in multicountry research a scale is typically part of a larger series of questions where a scale is used as a predictor or explanatory variable. In order to arrive at a more manageable questionnaire length, a shorter but psychometrically equivalent form of the scale is often used. For example, the Change Seeker Index (CSI) developed by Garlington and Shimote (1964) consists of 95 items that assess variety seeking. This is quite lengthy in its full form and would add considerably to the length of any instrument. An important consideration in reducing its length is whether a shortened version has equivalent psychometric properties. Steenkamp and Baumgartner (1995) have developed a seven-item version of the CSI that actually has better psychometric properties than the original scale. Figure 10.3 Structural equation model results * the variance was constrained. All coefficients are standardized. All solid line path coefficients are significant at P < 0.001. The dotted line coefficient is nonsignificant. Source: Klein etal., 1998. The full 95-item scale was first administered to a group of US subjects. Item-total correlations were run and the 67 items with correlations below 0.40 were dropped. The remaining 28 items were factor analyzed (principal components) and 13 items with a factor loading of below 0.50 on the first factor were dropped. The remaining 15 items were analyzed using LISREL. Two different models were evaluated: (1) all 15 items; and (2) the 7 items with a factor loading exceeding 0.7. While both models provided an acceptable fit, the fit for the 7-item scale was superior. The 7-item scale was then administered to another US sample as well as Dutch and Belgian samples to cross-validate and assess its nomological validity. These tests supported the superior qualities of the shortened scale. These results suggest that reducing scales to their core elements may enhance both their power and predictive efficiency. 10.3.2.2. Generalizability across Countries The emic/etic dilemma was discussed in Chapter 6. In the context of scales, it relates to whether a scale can be generalized across countries. A scale that consists exclusively of etic items can be generalized across countries, while a scale that consists of emic items cannot. Typically, a scale will end up containing both types of items. The earlier discussion of differential item functioning in this chapter begins to address this issue at the individual item level and helps to identify emic items. There are two different approaches that can be used to assess the generalizability of the entire scale across countries, confirmatory factor analysis (CFA) and generalizability theory (G theory). Sharma and Weathers (2003) examined both techniques to see whether the CETSCALE could be used to make valid comparison across countries. To conduct the comparison they used the data from the Netemeyer et al. (1991) study discussed earlier in this chapter. In using CFA, Sharma and Weathers followed Steenkamp and Baumgartner’s (1995) approach to testing equality of metric, factor and error variance, although they were unable to test for equality of covariance matrices because the number of parameters to be estimated was greater than the sample sizes. There was support for metric and factor variance equivalence, but not for error variance equivalence. They concluded that their analysis supports the notion that the CETSCALE is invariant across the four countries. To provide a comparison they analyzed the same data using G theory. The first step is to compute a generalizability coefficient that can be done for either random or fixed effects. The coefficient captures the variance due to countries, subjects and items as well as the interactions. Values of the coefficient greater than .90 suggest that a scale can be generalized across items (Shavelson and Webb, 1991). Sharma and Weathers obtained a generalizability coefficient of .936 and concluded that the CETSCALE can be generalized across items and, by implication, across countries. One of the main advantages of generalizability theory is that it can help determine the number of items and the number of subjects required to obtain a coefficient of at least .90. This is based on the initial study and can be used to guide the number of respondents and items for future studies. While CFA and G theory should be viewed as complementary approaches to assessing scales used in multiple countries, CFA is more useful in refining scales and identifying items that are problematic. 10.4. Multiple-context Scale Development The preceding approaches for assessing scales cross-culturally focus on examining the ‘fit’ of a scale developed in one country in another country or cultural context, and whether it exhibits high levels of reliability and internal consistency (or absence of item bias) in that context. While this provides evidence of reliability of the measurement instrument and equivalence of psychometric properties, it does not address the issue of whether the specific instrument provides an adequate measure of the construct being studied. The construct may, for example, be expressed in different terms, requiring an emic-specific instrument. Equally, it may consist of different components or dimensions. In either case, prior investigation of the construct in each country or context is needed before developing an instrument adapted to that country or context. A time-consuming but more theoretically sound approach is to develop a new scale for a given construct in each country studied to see whether the same dimensions are uncovered. Aaker (1997) developed a scale to measure brand personality in the US. Her scale revealed five different dimensions of brand personality: (1) sincerity; (2) excitement; (3) competence; (4) sophistication; and (5) ruggedness. The next issue was to determine whether the same dimensions existed outside the US. To examine this issue she adopted an emic approach and developed a scale for Japan following the same procedures used initially to develop the scale in the US (Aaker, 1998). The Japanese brand personality scale also yielded five factors. To determine how unique the set of five dimensions was, the two scales were administered to two additional samples, a group of US subjects and a group of Japanese subjects. From the set of 40 brands used in the initial studies, a subset of 10 brands that were highly familiar and for which there were no cultural differences in familiarity and liking was identified. Further, the brands represent the range of brand categories: symbolic, utilitarian and symbolic/utilitarian. The 10 brands were rated on 70 traits (42 English traits, 36 Japanese traits, minus the 8 overlapping traits). The two groups were factor analyzed separately to arrive at two sets of five brand personality dimensions. The first four dimensions were highly correlated (> 0.80) with each other (sincerity, excitement, competence, sophistication). However, a Japanese dimension labeled dependence and a US dimension of ruggedness were not strongly related to each other, suggesting that both were unique to their respective countries. Developing the scales separately in each country ensures an emic-centered measurement instrument and avoids the pitfalls of an etic approach. Since scales are developed independently of each other, they each provide insights into the unique components of brand personality in the US and Japan. There are still issues related to why the differences occur, but at a minimum, this approach avoids imposing a structure that is biased toward finding similarities. 10.4.1. Developing Cross-cultural Scales One of the greatest challenges facing cross-cultural researchers is the development of scales that measure a construct in multiple countries. In addition to all the issues related to achieving comparability and equivalence in the instrument, sampling frame, survey administration procedures and analysis, there is the underlying issue discussed in Chapter 6 of whether the construct exists and can be measured using the same or similar instruments in more than one context. Approaches based on existing scales are first examined. Much of this discussion relates to assessing the application of single-context scales in additional contexts. Then an alternative approach is covered based on ‘decentering’ measure development, which can be used to arrive at cross-cultural scales with fewer confounds. 10.4.1.1. Approaches Based on an Existing Scale Most published research dealing with cross-cultural scales report the results where a scale that has been developed in one country, typically the US, is applied in one or more additional countries. Few, if any, modifications are made to the original scale, with the exception of dropping items that do not exhibit high levels of reliability. In taking this approach, the researchers are assuming that a construct found in one country is manifested in the same form in another. Researchers may also adapt the scale by adding items to enhance their ability to identify culture-specific constructs. Assumed Etic Approach The most common approach is the ‘assumed etic’ approach. This is similar to Berry’s (1969) notion of ‘imposed etic’ and Triandis’s (1972) notion of ‘pseudo etic’. The choice of the term ‘assumed’ is meant to draw attention to the implicit assumptions being made by the researcher. The impact of these implicit assumptions on the soundness of the conclusions depends on how the scale is being used. In some cases, the researcher is primarily concerned with establishing the universality of the construct. The researcher assumes that the construct identified in country A exists in country B and can be measured in the same way in country B. Scale items are translated into the local language, the scale administered and results analyzed. In other cases, the scale is used as an independent variable to see whether it can predict some behavior or outcome in another country. Here, the emphasis shifts to focus on the predictive ability of the scale, rather than the universality of the construct. In both cases, minor adjustments will be made to the scale based on internal reliability. For example, items that do not contribute to improved levels of reliability based on Cronbach’s alpha are dropped. In all other respects, the scale is assumed to apply to the other countries. The criteria for item elimination are reliability and whether the item is correlated with other items that are supposed to measure the same construct, or more simply whether the item ‘fits’ with the other items. A related approach is to take a scale that has been developed in country A, administer it in country B and eliminate items that are biased. The biased items are dropped from the scale, so that the new scale has a reduced total number of items. Items are eliminated if they are found to be biased based on item response curves. This is a more rigorous criterion and can be done in an iterative fashion. The prevalence of the assumed etic approach is explained in part by the widespread availability of scales that have been developed to measure constructs in a single country, typically the US. Researchers become intrigued by the scale and attempt to apply it elsewhere. Underlying this process is the explicit belief that concepts are universal and that a properly modified scale can measure them. An implicit belief is that the concept is expressed in the same way in different contexts. A fundamental problem is that the starting point of the research is a scale anchored in one context and that elimination of items will reduce its length and improve its reliability, but not shift it to encompass another culture. Adding Emic Components Another approach is to add new items to a scale developed in one country in order to capture elements that are unique to the second country or context. Here, it is assumed that the core construct is etic or universal, but may be expressed in somewhat different terms in other contexts. Consequently, the measurement instrument may require some modification or the addition of culture-specific items. For example, Schwartz (1992), in developing his value survey to measure universal value types, identified values drawn from the world’s major religions and previous research on values, as well as inviting collaborators in many countries to add values. A total of 56 values were selected for the survey to represent value types. These were subsequently reduced to 45 based on an examination of the intercorrelations among the values (smallest space analyses [SSA]; Guttman, 1968). Adding emic elements results in a new scale for country B consisting of the items that measured the construct in country A, together with some items that are unique to country B. This approach provides a much richer set of scales, but at the same time creates a scale that is different from the original. It is also more complicated to analyze, as there are both common elements and elements that are unique to each country. The analysis becomes highly complex as more countries are added to the research. In this case, the procedure is repeated for each country studied, resulting in a common core across all countries and emic elements in each country. 10.4.2. Developing Decentered Scales An alternative and more desirable approach is to adopt a ‘decentered’ approach to scale development. Collaborators in other countries with culturally diverse backgrounds are asked to participate in defining emic or culture-specific dimensions of the phenomenon studied and provide input into instrument design and development of items relating to their specific country or culture. This item pool can then be analyzed to develop an appropriate scale identifying both common and culture-specific elements. This approach helps to eliminate the dominance of a specific country or cultural context in the operationalization of the construct and design of research instruments and procedures. 10.4.2.1. Item Development Development of a decentered scale must start with a clear initial definition of the construct that is to be examined. After the construct has been defined, the team of researchers should be broadened to include researchers from each of the countries to be studied. The domain of the construct should be refined and a determination made of whether or not the construct is operative in each of the different cultural contexts. At this point the initial construct definition may be redefined to encompass different manifestations. Once the construct is defined, the next step is to generate an item pool in each country or context studied. Suppose that a multicountry study is conducted to develop a scale of xenophilia or ‘love for things foreign’. This would be hypothesized to be positively related to the purchase or a willingness to buy foreign-made products and a tendency to rate these as superior or of better quality than domestic products. The construct might also be hypothesized to be negatively correlated with consumer ethnocentrism. Collaborators in the different countries would conduct studies to identify relevant dimensions of xenophilia in their own countries. Part of this process would be to identify context-specific factors that could affect feelings of xenophilia. For example, macroeconomic factors such as per capita GDP, the availability of locally made substitutes and regulations limiting imports would all affect the results. In addition, surveys of consumers in each country might be conducted. Consumers might be asked to describe their attitudes and behavior toward foreign items or to mention the first words that come to mind in relation to foreign products. Further probing might relate to images as well as words associated with foreign things. Projective techniques might also be employed. For example, respondents might be asked to describe a person who typically purchases foreign products. These country-specific studies could then be content analyzed to identify relevant dimensions of xenophilia in each country or research context. The next step is for a team of multicultural judges to determine which dimensions are common across countries (or regions) and which are emic or culture specific. If no common dimensions are identified, then it may be determined that the construct is best measured in different ways in each context, and no direct (or statistical) comparison is feasible. This is an important step that precedes actual data analysis. This step also adds to the understanding of the construct. In either case, a pool of items to tap the various dimensions needs to be identified. 10.4.2.2. Analyzing and Comparing Decentered Scales Data analysis presents its own unique challenges. For the sake of simplicity of exposition, the discussion will focus on the situation where the construct of xenophilia is being examined in only two countries, A and B. Adding more countries typically means that an iterative process will need to be followed. The researcher can either successively compare each unique pair or, when a large number of countries are being studied, one country can be treated as the target and each additional country can be compared to the target. If a construct is truly universal, changing the target should produce similar results. To start the analysis process, the combined item pool is administered in all countries studied. The items are analyzed first within each country to purify the scale and eliminate items showing item bias or low levels of reliability in each country. Next, within-country analysis is performed to identify emic or country-specific items. Factor analysis can be performed within each country on respondents’ standardized scores to identify factors common across countries and those specific to each country. Items with low alpha levels within a given country are dropped and the remaining items from each country used to determine emic-specific and etic components of the scale. To test whether the factor structures are similar, a Procrustes or target rotation can be performed. With two countries, either can be selected as the target. With more than two countries, the target should be rotated to see if the factor congruence coefficients hold for all unique pairs. This procedure is discussed in Chapter 12. In addition to items designed to measure xenophilia, some additional scales should be included so that discriminant and convergent validity can be established as well. Given that xenophilia is a love for things foreign, the researchers might want to include other scales such as a cosmopolitanism scale or the CETSCALE. The EAP scale mentioned earlier might also be included, to determine whether xenophilia is distinct from the desire to try new products in general. In attempting to establish convergent and discriminant validity, there is also the issue of whether the additional scales have etic properties. Ideally, the researcher would find scales that had already been developed or validated in each of the countries being studied. However, this is unlikely to be the case and will present the researcher with an additional set of validation problems. At this point, confirmatory factor analysis should be applied to the data. As part of this analysis it is desirable to have sufficiently large samples to be able to split the within-country samples into two samples for cross-validation purposes. Since confirmatory factor analysis is best with samples of 200–300, a sample of approximately 500 is desirable in each country. One subsample is used to estimate the model and the other is used to validate it. The confirmatory factor analysis will provide the best indication of whether the same model holds across both countries. If it does hold, the researchers can conclude that xenophilia is an etic construct. Alternatively, certain factors that are identified may be common and certain may be unique, suggesting emic components. So far the analysis procedures have established the reliability of the xenophilia scale and convergent and discriminant validity. Nomological validity would also have to be established. Information on the purchase, purchase intention, willingness to buy and ownership of foreign-made products should also be collected. The extent to which the xenophilia scale predicts these variables would establish its nomological validity. Differences between countries might suggest that the construct is not as etic as believed, does not bear the same relationship to purchase behavior, or that certain macro-context variables, such as availability or cost of foreign products, inhibit individuals’ ability to act on their xenophilia. The above discussion suggests some approaches to developing a decentered scale to measure a construct, in this case xenophilia. There are, however, different approaches that will allow the researcher to derive a decentered scale. The critical aspect is not to let one culture or perspective dominate the identification of the construct and the development of items to measure it. Thus, the most critical steps are the ones that precede the analysis; that is, construct definition and item identification. If the construct and the items are heavily influenced by one perspective, then examination in another context will simply mirror, albeit imperfectly, the original. Further, it will help perpetuate the belief in the universality of a construct that is more limited in its applicability. 10.5. Summary Developing scales to measure constructs that exist in multiple countries is one of the greatest challenges facing researchers engaged in cross-cultural research. Issues of nonequivalence that perplex researchers dealing with single items are magnified as single items are combined to form multiple-item scales that measure complex constructs. In this chapter approaches to developing better cross-cultural scales were covered. Critical issues in multicountry scale development are the types of measurement techniques used to gather the data as well as the underlying reliability of the data. As scale construction begins, the researcher must assess whether individual scale items function in the same way in each context or culture. Further, an assessment must be made as to whether the measures are in fact equivalent. Typically, in multicountry research a scale developed in one country is applied and evaluated in another country. While this is a common practice, it gives rise to a range of problems. The researcher needs to be cognizant of these, particularly if the original scale was developed in a single context. More theoretically correct approaches involve developing the scale with input from multiple contexts at the outset, thus decentering the perspective adopted in scale development and developing a culturally unbiased scale. 11. Experiments 11.1. Marketing Experiment Specifics Experiment in science is a process of fixing all state variables of some system, introducing some signal to system input and fixing the outcome. Then the result is explained within some theory. Experiments are not widely used in marketing research, because there are many uncontrollable variables practically in every marketing phenomenon. They are still more seldom used in International Marketing fresearch, as the scale is too large. They may be used in separate countries to study separate phrnomena and get qualirtative or qualitative data for further comparison. Data processing is similar to data, obtained by other sources. Thus, to get full panorama of international marketing research it is necessary to consider experiments. Для получения от экспериментов полезных и достоверных результатов, они должны быть тщательно спланированы. Описание проекта дается по [Черчилль]. 11.1.1. Experiment Validity В полевых экспериментах, которые проводятся в реальных условиях, нельзя полностью освободиться от влияния посторонних факторов. В лабораторных экспериментах создаются особые условия, позволяющие минимизировать все посторонние влияния. Результаты лабораторного и полевого экспериментов могут отличаться по различным причинам. Валидность результатов эксперимента – степень уверенности в их правильности. Существует множество видов валидности. В лабораторных экспериментах исследовалась зависимость цена – спрос. Испытуемым (добровольцам, собравшимся в лабораторию) давались карточки с названиями товаров и ценами. Эти цены обсуждались, составлялся список «псевдопокупок». Эта же зависимость исследовалась в полевых условиях, в супермаркетах. Записывалась недельная продажа перед изменением цен и через две недели после изменения. Никакой рекламы об изменениях цен не было. Результаты двух исследований показали тенденцию к преувеличению роли цены в лабораторных экспериментах. Для лабораторных условий, помимо указанного эффекта, следует иметь в виду и то, что участвовать в таких экспериментах соглашается определенная категория людей (прежде всего, имеющих свободное время), что может исказить результаты. Внешняя валидность – степень уверенности в соответствии результатов исследования действительности. Из приведенного выше примера видно, что в лабораторных экспериментах эта валидность довольно низка. В полевых экспериментах наблюдается реальная ситуация, а не ее модель, поэтому внешняя валидность обычно выше. Внутренняя валидность – степень уверенности в том, что наблюдаемый эффект обуславливается именно экспериментальной переменной, а не другими факторами. В рассмотренном полевом эксперименте не было рекламы, специально оформленных витрин и тому подобных посторонних для эксперимента факторов. Но не исключено, что одновременно с ним проводилась реклама другого товара или магазина, что могло повляить на результаты. Так что внутренняя валидность полевых экспериментов обычно низка. Она выше у лабораторных экспериментов. Таким образом, чтобы результаты эксперимента оказались полезными, следует тщательно учитывать возможные источники ошибок. 11.1.2. Unwanted variables История. Это события, ситуации, обстоятельства, которые происходят или возникают одновременно с экспериментом, являются внешним для эксперимента и могут влиять на его результаты. Например, для эксперимента, O1 X O2 , где O1 – измерение объема продаж в магазине; X – изменение цен, в данном случае – снижение; O2 – измерение нового объема продаж по новой цене, в момент X могло произойти открытие нового магазина по соседству, что привлекло покупателей туда. Это отрицательно сказалось на объеме продаж. Результатом такого исследования может быть вывод, что при снижении цены товары меньше покупают. Чаще всего история проявляется не столь явно, поэтому проблема оценки влияния этого фактора достаточно сложна. Общий подход сотоит в том, чтобы выделить контрольную группу объектов исследования, на которую влияет только история. Созревание. Это процессы в объектах эксперимента, зависящие от времени. Люди устают, стареют, изменяются их семейное положение и вкусы. Интерьеры магазина выходят из моды, устаревают, периодически обновляются. Продавцы набираются опыта. В связи с созреванием возникает извечная проблема маркетинговых исследований: когда их проводить. Если опрос провести сразу после рекламной кампании, не успеет проявиться эффект распространения слухов. Если же ждать слишком долго, то воздействие исследуемой рекламы заслонится другими эффектами, изменится отношение покупателей к рекламируемому товару. Даже в процессе двухчасового интервью люди устанут, проголодаются, будут проявлять нетерпение, что скажется на их ответах. Поэтому опросы следует тщательно готовить и проводить достаточно быстро. Статистическая регрессия. Это тенденция к сглаживанию выбросов. Типичный пример: некоторая семья закупила большое количество упаковок сока, что было обусловлено семейным праздником. В дальнейшем объем покупок семьи стабилизируется на среднем уровне. Смещение выборки. Иногда нельзя сказать, что группы были одинаковы до проведения эксперимента. Например, утверждение типа: 60% слышавших рекламу по телевидению положительно относятся к товару, в то время как 60% не слышавших относятся отрицательно неверно, так как рекламу могли слушать те, кто уже и без нее хорошо относился к товару и интересовался им. Они лучше запомнили, что реклама была. Способ устранения этой ошибки – тщательный подбор элементов исследования. Например, для оценки эффекта от выставки нового пищевого продукта в магазине экспериментальная и контрольная группы элементов исследования выбираются среди магазинов, близких по параметрам: площади, району расположения и др. Другой способ подбора – выделение пар похожих магазинов и определение одного из каждой пары в экспериментальную группу, а другого – в контрольную. Но подбор элементов – довольно трудоемкое дело. Кроме того, не всегда ясно, по каким именно параметрам его производить. Для магазина это может быть площадь торгового зала, численность персонала, объем продаж за месяц, ассортимент… Хорошие результаты дает случайный выбор элементов исследования – рандомизация31. Она обычно предпочтительнее при большом числе элементов исследования. Потери в ходе эксперимента. Это исчезновение тестируемых единиц в процессе проведения эксперимента. Например, если для эксперимента было случайно выбрано 30 магазинов, но пять из них отказались участвовать в исследовании, то нельзя сразу сказать, что их отказ не изменит среднего значения результата. Требуется устранить сомнения на этот счет. Может быть, это магазины, испытывающие финансовые затруднения, и они по-другому отнеслись бы к экспериментальному воздействию. Фактором, искажающим результат эксперимента, является и сам процесс тестирования32. Он вызывает следующие эффекты. Основной эффект тестирования заключается во влиянии априорного наблюдения на последующие наблюдения. Например, если студентам дается тест на сообразительность, то результаты повторного теста будут лучше, даже если им не сообщать о результатах первого теста (здесь О1 влияет на О2). Типы задач, методы их решения, практические приемы будут уже знакомы. В анкете ответы на начальные вопросы влияют на последующие ответы респондента (то есть О1 влияет само на себя). Интерактивный эффект тестирования – влияние предварительных (априорных) измерений на эффект воздействия (О1 влияет на Х). Если опрашиваемых попросили ответить, что они думают о некоторой марке машин, то после этого они будут замечать рекламу данной марки, интересоваться ею. Соответственно, воздействие рекламы на них будет выше, чем на людей, которых заранее не опрашивали. Инструментальная ошибка. Любой физический прибор (весы, термометр, рулетка) имеет определенную погрешность измерений. В маркетинговых исследованиях наблюдается аналогичный эффект. Ответы респондентов зависят от формулировки фопросов, от порядка вопросов, тона, которым их задает интервьюер и многих других факторов. Не исключены и ошибки при записи ответов. 11.2. Experimental designs as a source of market insight [Ryak L., Wilson H. Experimental methods in market research. // International Journal of Market Research, 2005, Vol. 47 Issue 4, p. 347-366]. For simplicity experiment designs that tend to occur in marketing research are divided into four groups, as shown in Table 1. Table … Experimental designs in marketing research Type Setting Group assignment Dependent variable measurement 1. Laboratory experiment Laboratory a) Random, or: b) Systematic Quantitative 2. Field experiment Field a) Random, or: b) Systematic Quantitative 3. Ex post facto study Field Naturally occurring Quantitative 4. Quasi-experimental qualitative design Field Naturally occurring Qualitative 11.2.1. Laboratory experiments In the classical laboratory experiment, the subjects perform some task or activity within a carefully controlled physical environment. This can help to reduce the number of extraneous variables - factors other than the independent variables being studied - that could be affecting the dependent variable. The most common design is a ‘before-and-after, with control group’ design, in which a control group of subjects and an experimental group differ only in that a ‘treatment’ is given to the experimental group only: that is, the experimental group has a different value of the independent variable. The dependent variable is measured both before and after the experimental treatment. Subjects are allocated between the groups either randomly or systematically - that is, by selecting control group members to match the experimental group members on specific potential extraneous variables (for example, age, profession, and so on) and thereby to control for these variables. Typical usage of laboratory experiments: Computer software testing Gadget testing Evaluating consumers’ channel preferences, particularly comparing online and offline options. Taste tests and the like have also long been conducted in laboratory conditions. Toy testing. About validity. One wonders, for example, quite how much insight can be gained into the minds of popcorn-eating teenagers at the multiplex from experiments on MBA students about two fictional brands of popcorn described only on paper, however impressive the methodology or learned the journal. 11.2.2. Field experiments Real examples. Comparison of well-being in an experimental group of consumers that had moved location, with well-being in a control group. Customers are divided into two parts. One half received the marketing treatment, the other half did not. Reactions were monitored. Research sample was divided into three. One third received a discount voucher, another third received a promotional item of equal value, while the control group received nothing. Psychological experiments: 2 groups, one half received a prize in 2 variants: mobile phone or equal sum. Most choose money. In another group all were given a phone and then were suggested to change for money. Few agreed. There are problems in experimenting with large groups, but they are less whtn dealing with small groups. Research: 30 insurance sales people were given certain training and another 30 - matched on previous performance with the experimental group -were not. The performance of the experimental group increased immediately and continued to increase over time. In business-to-business research there were compared executive compensation in two small groups of companies, matched on industry type but differing in compensation style. Field experiments address the laboratory experiment’s weakness of ecological validity, allowing genuine insight to be gained into customer reactions and commercial implications that continues beyond the experiment. The most common practical challenge, though, is ethical. 11.2.3. Ex post facto studies We will probably always be faced, though, with situations where we wish to examine the effectiveness of actions that have already occurred. The ex post facto study applies the logic of the experiment to this situation. The group to whom the intervention occurred is compared retrospectively to a similar, naturally occurring group where the intervention did not take place. As with systematic group allocation, the control group is chosen so as to match the experimental group as far as possible on any hypothesised extraneous variables. The case for the negative effect of smoking on health was made primarily on the basis of such ex post facto studies, mortality rates of smokers and matched groups of non-smokers being compared along with a check that the groups did not differ on other relevant criteria such as diet and exercise. As with the field experiment, a challenge is defining genuinely comparable groups. A further drawback is that pre-tests - such as surveys of the attitudes of the two groups before the intervention - are clearly not possible. Nevertheless, this logic is considerably stronger than the common approach of simply surveying customer attitudes among those exposed to the intervention, as insight can be gained into what shifts were due to the intervention and what shifts would have occurred anyway. 11.2.4. The quasi-experimental qualitative design We have so far focused on quantitative approaches. There are many circumstances, though, where we wish to gain insight into the impact of a customer management strategy without making a prior assumption of what that impact might be. The logic of the experiment is applied to this situation. An experimental and control group are allocated as in the previous approaches, but the post-test measurement is carried out qualitatively, typically through interviews or focus group discussion, and not quantitatively. Evaluation of a decision support system through interviews was performed with both those who had used the system and those who had not, to provide a check for such problems as the ‘history’ error - ‘any variables or events, other than the one(s) manipulated by the experimenter, that occur between the pre- and postmeasures and affect the value of the dependent variable’. 11.3. Cases 11.3.1. Introduction to cases Three cases from [ibid] are summarised in Table 2. Table 2 Summary of cases The first two cases share a common theme: the introduction of customer profitability analysis into a business, and the impact of this intervention on customer management practice. In the insurance company of the first case, a quasi-experimental qualitative design was used to investigate what change in key account management (KAM) practice occurred when an experimental group of key account managers was given access to customer profitability data. Interviews were held with both the experimental group and a control group to increase confidence that the changes observed in the experimental group came about as a result of the project and were not down to other company factors or to chance. The bank studied in Case 2, by contrast, was used to thinking in terms of mass marketing. A project to introduce customer profitability and segmentation analyses aimed to improve the profitability of the overall customer portfolio. The primary research question was whether segment membership impacts on customer profitability and, second, on the likelihood of success in the application process for the company’s personal loans, the product range studied. This can be regarded as an ex post facto design in which natural variation in segment membership represents the independent variable - just as natural variation in smoking habits is used in the smoking studies we referred to earlier. In this case the sample size was extremely large (the research was conducted on a sample of more than 95,000 customers). The ‘control group’ in this case was in fact a validation sample, consisting of 100% of applicants for a personal loan during a specified period. Thus, the validation sample included all of the experimental group. The analysis of the experimental group was compared and contrasted with an analysis of this entire validation sample. The importance of this validation sample for the bank was that it was a proxy for the market as a whole, since the validation sample included loan applicants who were turned down by the bank and loan applicants who were offered a loan but decided not to take one out, as well as all successful applications during the period. Case 3 is an example of a field experiment. An ICT (information and communication technology) vendor wished to test the hypothesis that the introduction of desk-based account managers (DBAMs) to work alongside the existing field sales force would save cost without decreasing revenue or customer satisfaction. A pilot study was defined to test this hypothesis, introducing 12 DBAMs into a set of accounts for a trial period, with measurement in both this experimental group and a control group of employee satisfaction, customer satisfaction, sales cost and revenue. 11.3.2. Case 1: an insurance company The research for this case took place at the London office of one of Europe’s leading insurance companies. This company had a relatively small number of key accounts and had recently established a Key Account Management (KAM) team to handle relationships with its largest customers. There was little history of performance measurement in this office. In particular, the profitability of the key accounts being handed over to the KAM team was not known for certain. This was a pressing issue with the creation of the KAM team, since intensive investment in customer relationships would increase the cost of serving the key customers. Hence a project was instituted to provide the KAM team with account profitability data, with assistance in action research mode from one of the authors. 11.3.2.1. Purpose of the quasi-experiment The purpose of the quasi-experiment was to provide a check on changes in customer strategy that might be occurring during the project but for reasons not connected with the project. The control group would indicate what changes would have happened anyway without the project. There were three particular reasons to be on the lookout for such extraneous variables. First, because relationship management and the development of customer-specific strategies were actively being discussed within the company, the establishment of the KAM team was the most visible effect of an ongoing cultural change. Second, moving the very largest accounts into the Key Account Management team might have had an impact on the resources available to service other major accounts. Third, other external factors, such as changes in the marketplace or customer demands, might have led to changes in customer management strategies that would have come about irrespective of the initiative on customer profitability. 11.3.2.2. Group allocation and management The experimental group included eight account managers managing 18 accounts. The control group comprised accounts that were major but not key accounts, managed by three account managers not involved in the customer profitability work. Both groups of managers were interviewed individually, with some modifications to the project entry questionnaire and project exit questionnaire for the control group. Other than that, contact with account managers in the control group was avoided. The control group account managers were not made aware of the results of the customer profitability project until after it had been completed. The control group was matched with the experimental group on three criteria: industry sector; the product portfolio owned by the accounts; and the degree of internationalisation of the accounts. They differed, however, on account size, as the company wished all the largest accounts to have access to the new profitability data. This is an example of the impurity in systematic group assignment that can occur due to practical management criteria - an issue we discussed in our literature review. 11.3.2.3. Findings As the KAM team learned more about the profitability of their customers, they made several modifications to their customer management strategy, of which three of the most significant were as follows. Differential service: The key account managers became more aware of service levels to customers who were marginally profitable or unprofitable, and sometimes refused additional free services to these customers, beginning to negotiate instead. Selective sales effort: The depth of sales coverage was increased in those accounts that were either more profitable or were believed to have the potential to become so, in order to increase ‘share of wallet’ of this profitable business through new product and service sales. Selective customer divestment: In some cases, accounts were removed from the ‘key account’ list as not deserving the resource-intensive KAM approach. While maintained as customers of the company, it was accepted that this reduced focus might result in a reducing share of wallet or indeed a total loss of business from these companies. All of the experimental group account managers reported one or more of these changes in their accounts. By contrast, the only change reported in the control group was with one account where the client had had a change in management team. This led to the need for more input from the account manager, who held a series of workshops as part of a process of getting to know the new team. However, the account manager did not consider there had been any changes to profitability or risk in the relationship as a result. The company regarded this exogenously driven short-term change to the way the account was managed as very different from the long-term proactive changes to customer strategy seen in the experimental group. This simple qualitative use of a control group gave the insurance company confidence that the changes in customer management strategies that had been observed during the project were as a result of better knowledge on the part of the Key Account Management team about the profitability of their customers. The KAM team and other senior managers at the insurance company expressed considerable satisfaction with the project. 11.3.2.4. Limitations There were two main limitations in this quasi-experimental design. The first, which we have mentioned, was the danger of selection error -differences in the two groups that are relevant to the dependent variable. Specifically, the control group differed somewhat in account size. The interviews attempted to counteract this by checking whether account size was significant as a driver of either profitability or customer management. More equal groups in account size would clearly have formed a more ideal design, but might well have fallen foul of the holdout problem we described earlier. The second limitation is the danger of contamination or ‘spillover’ of the control group due to its proximity to the experimental group. In this case, the control group account managers were located in the same open-plan office as the experimental group team. The control group interviews explicitly checked for this point, finding that two of the three control group account managers were not even aware that the profitability work was taking place. The third was the account manager whose one reported change in strategy we have already discussed, and who believed this change to be instigated by the client rather than by internal factors. We concluded that this danger had not in fact contaminated the research findings significantly. 11.3.3. Case 2: a bank Case 2 concerned a business-to-consumer project that took place within the personal loans division of a major bank. The division had a customer base of several hundred thousand people. Again, the wider project concerned the introduction of customer profitability analysis, but instead of analysing the profitability of individual customers as in Case 1, Case 2 took a top-down approach and aimed to break down a research sample of the current customer base into differing profitability tranches. 11.3.3.1. Purpose The project aimed to identify tranches of personal loan customers with similar profitability, with the objective not just of developing customer management strategies appropriate to the profitability of the customer, but also of identifying the most profitable customer types in order to target and acquire new, more profitable customers. In the evaluation of this project, the main research questions were, first, what impact segment membership has on profitability and success in the loan application process and, second, what impact this process has on the credit score rating of the customer portfolio. 11.3.3.2. Research design Several ideas for research design were considered by the researcher and client. The first was to run the project in three stages (pre-measurement group, results evaluation and strategy testing, and post-measurement group), where the middle stage would involve modifications to customer management as a result of the customer profitability analysis - a simulated before-and-after design (Winer 1980). The second idea was the use of closely matching customer profitability tranches. Customer profitability tranches thought to be of great interest were: • the tranche containing the highest proportion of profitable customers • the tranche delivering the greatest amount of profit • the tranche containing the highest proportion of unprofitable customers • the tranche delivering the least amount of profit, and • an average tranche. Under this option, the research team intended to identify a similar tranche to each tranche selected for detailed study. This second tranche would act as the control group. For example, the tranche containing the second highest proportion of profitable customers might act as the validation sample for the tranche containing the highest proportion of profitable customers. Both tranches would be profiled in detail and differences between them noted. A third approach the researcher suggested was to set aside a control group of customers for whom profitability data would not be calculated. Differences in customer management of the control and experimental groups would then be assessed as in Case 1. The company decided, though, that it did not wish to forego the potential benefits of the profitability analysis for any customers (the holdout issue again), so this was ruled out. In the event, the project team decided on a fourth approach - to use the total sample of applications for a loan as a validation sample - for the following reasons. Using the first two approaches, the number of customers in some tranches would be too low for statistical significance. The approach of monitoring any subsample against a set of all customers was the bank’s normal approach for statistical analysis of customer data. Using 100% of total loan applications would provide a proxy for the entire market. The benefit of testing the customer profitability tranches against the entire market would be in the identification of relatively profitable segments in which the bank might be under-represented in the customer portfolio, as well as relatively unprofitable segments that might be over-represented. 11.3.3.3. Findings The project analysed the profitability of customer tranches, the risk of the customers in those tranches, and the relationship the customers in each studied tranche had with the bank as a whole. Sociodemographic data such as ACORN were also overlaid on this analysis. For each key variable, results were presented as a ‘compare and contrast’ with the validation sample (see Table 3). One of the key conclusions from this analysis was that the bank’s customers were of higher quality than the validation sample. They were slightly older and more likely to own their own home, to work in a professional occupation and to have higher earnings. This resulted in a better credit rating for the research sample as compared with the validation sample, confirming the bank’s decision to focus on these customers (see Figure 1). Similarly, the division’s marketing department had recently completed a market segmentation exercise for which it had only indicative data for the percentage market share of eight segments it regarded as potential targets. These eight segments were mapped onto the validation sample and research sample to see what proportion of customers fell into each segment, providing some confirmation as to market share. The validation sample indicated that the bank was relatively heavily represented in one segment comprising younger unmarried customers. However, this segment was discovered by the research to be relatively unprofitable. Prior to the research, it had been assumed that the lifetime value of this particular segment was high. The profitability research, coupled with the use of the validation sample, convinced the bank to alter its policy towards this segment. Recent aggressive targeting of this segment was discontinued. The validation sample also indicated an opportunity. Another segment, comprising certain older customers without children, turned out to be considerably more profitable than the bank had realised. The bank did have a substantial proportion of customers in this segment but the validation sample suggested that people meeting this profile were more heavily represented in the population as a whole. Further analysis suggested that the bank was able to attract prospects of this type but was not able to convert them from prospects to customers. To do so entailed developing a new proposition to appeal to this profitable segment. Table 3 Experimental and validation groups by ACORN segment (Case 2) ACORN ACORN description Project sample (%) Validation sample (%) A Thriving 11.4 10.2 B Expanding 14.3 12.1 C Rising 6.8 8.4 D Settling 26.1 23.3 E Aspiring 14.9 15.1 F Striving 22.8 26.1 Data missing 3.7 4.8 Figure 1 Credit score by experimental and validation groups (Case 2) 11.3.3.4. Limitations The validation sample was, as explained above, 100% of the applicants for a loan during a period of several months. It was taken as a proxy for the market as a whole. For this reason, the sample was drawn from a period during which there were no special marketing promotions or pricing deals that might have skewed the sample. However, it was not clear how closely the validation sample did in fact approximate to the market as a whole. This would be amenable to statistical analysis that may form part of an ongoing marketing research activity. 11.3.4. Case 3: an ICT provider This company sold IT and telecommunications products and services to large corporate clients. As with Cases 1 and 2, the change it wished to evaluate was not the market researcher’s typical problem of a new product or promotional method, or a change in customer behaviour, but rather a change in customer management: the introduction of desk-based account managers (DBAMs) to complement the existing field sales force. It was hoped that the introduction of fully professional account managers who worked purely from the office, making extensive use of remote media such as the telephone, would lower the cost of sale due to their lower salary and expense costs and greater contact time, and it was hoped they might actually be preferred in some circumstances by customers for simpler matters. The company had the good sense, though, to recognise that some form of pilot testing using an experimental approach would be wise before investing too heavily in its intuitive strategy. 11.3.4.1. Purpose A trial was set up to provide a set of field account managers with the assistance of a small number of DBAMs for the trial period. The purpose of the trial was to establish whether DBAMs could indeed sell some of the product range, whether the cost/revenue equation was indeed improved, and whether the people involved - employees and customers - were happy with the approach. 11.3.4.2. Research design The company selected a matched set of accounts to act as a control group, as it was entirely possible that such metrics as profitability and customer satisfaction might change in the customer base as a whole due to extraneous factors such as competitor activity or new product introduction. Both groups were assessed before and after the trial by employee and customer surveys to measure employee and customer satisfaction respectively. An analysis of internal data measured the cost of each approach as well as the revenue gained, to enable a comparison on the measure of cost-to-revenue ratio, the proportion of revenue that was needed to cover marketing and sales costs. 11.3.4.3. Findings The experimental group achieved a very significant improvement in the key metric of cost-to-revenue ratio, reducing it by around 6% of revenue. The company’s interpretation of this result was that DBAMs proved able to sell products to a much higher value than had been anticipated; furthermore, in many other sales that were completed by the field sales force, the DBAMs played an important role at earlier stages of the decision-making process. As a result, replacement of a proportion of the field sales people by DBAMs on a one-for-one basis decreased cost at the same time as increasing revenue, as the field sales force had more time available to chase the greatest opportunities where they were most needed. Employee satisfaction did not improve significantly during the trial, but at least it did not actually drop - an achievement in a situation of radical change, which some employees were bound to find threatening. The company was particularly pleased that customer satisfaction actually went up, as customers found they could contact a member of the account team more readily, as DBAMs tended to have shorter continuous contact times and of course no travel time, and so were easier to contact. 11.3.4.4. Limitations This field experiment with systematic group assignment had many strengths in research design. Notably, the control group was more directly comparable to the experimental group than in the first two cases. The benefits in the credibility of the results were considerable, with the DBAM approach being rolled out throughout the division over the following three years. It is worth noting one weakness, though, due to the famous ‘Hawthorne effect’: the way in which any novel circumstances are likely to influence the behaviour of experimental group members (Roethlisberger & Dickson 1939). In the Hawthorne studies, the lights were turned up in a factory and performance went up. The lights were turned down and performance still went up. The researchers soon realised that virtually any change would result in at least a short-term improvement. In Case 3, the equivalent danger was that such metrics as customer satisfaction were increased due to the specific attention those in the experimental group knew themselves to be the subject of. In this regard, it is interesting to note that the company’s subsequent experience has confirmed the financial benefits of DBAM introduction but has had a rather less significant positive impact on customer satisfaction than the pilot suggested; it is not clear whether this is purely a function of the Hawthorne effect or whether it is due to differences in the rollout programme as compared with the pilot. 11.4. Are experimental designs valuable in gaining marketing insight? Insight forms a bridge between market research data and managerial action. Without insight - in the heads of marketing and general managers, not in the minds of internal or external researchers, as Wills and Williams (2004) point out - the data will sit on the shelf, proverbially and literally. One test of whether research delivers insight (though not of the validity of that insight) is therefore whether managerial action results from the research (which incidentally would therefore form an interesting metric by which market researchers could measure their performance). On this measure, at least, the experimental approach seemed successful in the three case studies we have described. The insurance company study focused the project team around the value that they gained from their customer relationships. This led to changes in customer strategy as confirmed by the project team exit interviews (and by comparison with the control group exit interviews). Astonished by some of the profitability data, the bank made significant changes to its loan application process to attract more of the customers it wanted and fewer of the less profitable customers. It also reoriented its marketing communications to target the most attractive segments. Differential pricing was refined to ensure that low-risk, attractive segments were indeed attracted to the bank’s offering. The ICT vendor has reduced its cost-to-revenue ratio significantly by replacing around 20% of its field sales staff by desk-based account managers. It also estimates an eight-figure increase in annual revenue as a result of the change of strategy. Researchers need to broaden their methodological armoury to include experimental and quasi-experimental approaches that should be at least considered when faced with a research brief. To a man with only a hammer, everything looks like a nail; to a market researcher who does surveys and focus groups, every research problem is perfect for a survey or focus group. Some jobs need many tools, and some ‘customer management insight’ projects need many data inputs, and a holistic analysis of those inputs in which the logic of the experiment can play a part, we believe, more often than current usage would suggest. To those who are inspired to rethink their research design choices, we hope we have also given some useful material from a varied set of three studies that have, to a greater or lesser extent, used the experimental method to shed light on some of the common pitfalls and how they can be avoided. 11.4.1. Test Marketing Тестовый маркетинг стал широко использоваться начиная с 1960 года. Тестовый маркетинг представляет собой управляемый эксперимент, осуществляемый в ограниченной, но тщательно выбранной части рыночного пространства с целью предсказания объема продаж или размера прибыли в абсолютном или относительном выражении как реакцию на маркетинговые действия фирмы. Многолетний опыт показывает, что 3 из 4 новых товаров, имевших успех в тестовом маркетинге, имели успех и впоследствии, а 4 из 5, провалившихся в тестовом маркетинге, впоследствии также не пользовались спросом33. Однако тестовый маркетинг дорог, занимает длительное время, раскрывает конкурентам планы фирмы и не всегда дает точный результат. При планировании тестового маркетинга следует учитывать следующие параметры. Стоимость. Если обычное маркетинговое исследование включает затраты • на проектирование инструмента сбора данных; • на осуществление выборки; • на проведение опросов; • на обработку данных, • то тестовый маркетинг включает дополнительно затраты • на рекламу; • на персональные продажи; • на организацию и проведение выставок; • на раздачу бесплатных образцов. Если к тому же исследуется новый продукт, то средства тратятся на изготовление пробной малой партии товара. При этом себестоимость единицы товара гораздо выше, чем при серийном производстве. В США, например, на тестовый маркетинг, охватывающий 2% национального рынка, затрачивается в среднем $3,1 млн. Продолжительность. Точность тестового маркетинга возрастает с возрастанием его времени. Если из двухмесячных маркетинговых экспериментов только 13% дали правильный прогноз, то из десятимесячных – 83% 34. Рекомендуемый срок для тестового маркетинга – не менее года (с целью учесть сезонные колебания). Управление экспериментом. Требуется определить, какой регион и какие города будут охвачены, какие каналы распределения (оптовые или розничные) будут задействованы. В тестовом маркетинге товару уделяется повышенное внимание: товар производят особенно тщательно, обеспечивают его наличие на полках магазинов, отводят ему место на складах. Показательно, что 80% провалов после успешного тестового маркетинга обусловлены реальными условиями хранения, которые отличались от условий в тестовом маркетинге. Действия конкурентов. Конкуренты могут снизить цены на свои товары во время вашего тестового маркетинга и тем самым сильно уменьшить объем продаж вашего товара. С другой стороны, они могут и скупить ваш товар, создавая у вас эйфорию успеха. Некоторые примеры результатов тестового маркетинга даны в [31]. • Сухая краска для волос, наносимая на гребешок, в жару потекла на лоб потребителей. • Подушечки для мытья посуды оказались слишком скользкими и падали с полок на кухне. • В холодную погоду жидкое детское питание расслаивалось на воду и осадок. • В жаркую погоду сигареты пересыхали, теряя вкусовые качества. • Еда для животных вызывала у них аллергические реакции. • Совмещенный по времени со снижением цены, перевод продукта в жидкое состояние воспринимался покупателями как разведение его водой. • Из-за экономии клея при изготовлении упаковки, она разваливалась при транспортировке. • Покупатели спутали жидкость для мытья посуды с лимонным соком; в результате 33 взрослых и 45 детей заболели. • Концентрированный сок некоторые потребители употребляли для питья, а некоторые – для окраски квартиры. • Конкурентам удалось скопировать рецепт нового соуса. В этих случаях тестовый маркетинг позволил избежать больших затрат, которые возникли бы при производстве этих товаров сразу для национального рынка. Методы тестового маркетинга классифицируются главным образом по типу используемого в эксперименте рынка. Они будут рассмотрены в соответствии с обычной последовательностью их применения при выводе на рынок нового товара. 11.4.1.1. Test Marketing on Model Market Вначале следует определить отношение потребителей к новому товару: нужен ли он им, нравится ли он им. Хотя все это должно учитываться еще при проектировании новых товаров, необходима проверка того, что получилось в результате разработки. Для этого требуется изготовить одну или несколько единиц товара. Покупатели интервьюируются дома, у магазина, около витрины, где представлен образец товара или в специально отведенном помещении (лаборатории). Им показывают новый товар, и они высказывают свое мнение о нем. Сообщаются цены на предлагаемый товар и цены конкурентов. В лабораторных экспериментах используются псевдоденьги или купоны для того, чтобы посмотреть, каким товарам отдается предпочтение. Иногда товар можно купить, но он может даже предлагаться бесплатно. Через некоторое время (например, через 2 недели) получившим или купившим товар звонят и спрашивают их о том, нравится ли он им. Все собранные сведения вводятся в компьютерную модель, которая прогнозирует уровень повторных покупок и долю рынка. Главное в этом методе – компьютерные модели. Современные модели позволяют достичь 10% точности прогноза объемов продажи в 80% случаев. Этот вариант хорош и для оценки покупок на пробу и повторных покупок. На этом этапе конкуренты мало знают о ходе тестового маркетинга. Метод хорошо показывает слабые продукты, при этом затраты на исследование невысоки. Ахиллесовой пятой метода является невозможность предсказать, сможет ли фирма обеспечить торговлю товаром и невозможность определить реакцию конкурентов. Для этих случаев имеющиеся модели работают очень плохо. В этой связи, метод хорошо применим для оценки модификаций товаров и хуже работает с новыми товарами. 11.4.1.2. Test Marketing on Controlled Market Следующий вопрос, на который следует найти ответ, – будут ли покупатели покупать новый товар. Для этого его необходимо представить на полках магазинов. Управляемым называется рынок с усиленным распределением. В этом случае производится оплата розничным торговцам за место на полках, устраиваются специальные выставки, применяются различные методы продвижения. Фирма часто стремится обеспечить тем самым заданный объем продаж. На управляемом рынке обеспечиваются: • заданный уровень представления товара в магазине; • размещение в заданном отделе; • заданное количество мест на полках; • заданная цена; • заданные методы продвижения. Недостатком метода является то, что согласие или отказ торговли принять новый товар часто определяет успех, а здесь этот вопрос решается в принудительном порядке. В реальной ситуации товара может не быть в магазине, он может быть плохо выставлен, может быть не решен вопрос с торговлей о цене и мерах продвижения. Новый товар, критичный к условиям торговли, нецелесообразно испытывать этим методом. 11.4.1.3. Test Marketing on Real Market Последняя большая задача – организовать торговлю новым товаром. Требуется проверить, как идет торговля, выявить и устранить возникающие проблемы. В этом случае компании продают свой товар по обычным каналам, но только в определенном городе или районе. Метод используется, • когда требуется узнать, действительно ли торговля примет товар; • когда капитал ограничен и необходимо точно прогнозировать возможные затраты; • когда осуществляется выход на новую территорию и необходим реальный опыт. 11.4.1.4. Test Marketing Sequence Самый дорогой из рассмотренных методов – тестовый маркетинг на реальном рынке. Использование управляемого рынка несколько дешевле. Самый дешевый метод – модельный тестовый маркетинг. Основная последовательность выбора процедуры тестового маркетинга показана на рис.Рис. 5. Если получены очень хорошие промежуточные результаты, то можно рискнуть перескочить через один из этапов. Рис. 5. Последовательность применения методов тестового маркетинга 12. Analysis Of Multicountry Data Once procedures for collecting data have been determined using either survey or nonsurvey methods, the next step is the choice of appropriate methods and procedures for data analysis. These two steps are interrelated as certain types of analysis, for example multidimensional scaling, require collection of specific types of data. There are a number of issues to be considered in relation to how data analysis is conducted and organized. For nonsurvey research data analysis is typically qualitative. Survey research requires some type of quantitative analysis or, at a bare minimum, tabulation of the responses. When research is conducted in a single country, these issues are the same as in domestic research. In multicountry research, the issues become more complex due to the existence of multiple units of analysis. The analysis is often conducted in different phases, starting with a within-country analysis and progressing to an across-country analysis. To ensure that the appropriate data are collected, a data analysis plan has to be established beforehand. The complexity of multicountry data analysis is due to the multitier character of the research design, entailing analysis not only at the country level but also across regions and eventually at a global level. In domestic marketing research, decision problems and analysis typically relate to a single national sample. In multicountry research, management is concerned not only with developing marketing strategy or tactics relative to a single national market, but also with assessing the extent to which such strategies or tactics can be standardized across different geographical areas. Consequently, analysis needs to be conducted within countries and across countries. This poses a number of issues with regard to the level of aggregation and procedures used to analyze data. In this chapter the problems associated with both within- and across-country phases of data analysis are examined, as well as procedures for minimizing difficulties. The discussion focuses primarily on the analysis of quantitative data. It is assumed that the reader is already familiar with standard marketing research texts (Churchill and Iacobucci, 2005; Kumar et al., 2002; Lehmann et al., 1998) and with the procedures commonly used to edit, code and analyze data. Methods of analyzing cross-national data that focus on the differences in the level of variables are examined, including use of both univariate and multivariate techniques. Here, knowledge of statistical methods commonly applied to analyze domestic marketing data is assumed, and emphasis is placed on how analytical procedures and methods are used to test for similarities and differences between data from different countries. Standard texts on multivariate analysis can be consulted for more background on data analysis techniques (Lattin et al., 2003; Hair et al., 1998; Johnson and Wichern, 1998). 12.1. Multicountry Data Analysis 12.1.1. Analysis of Data at Different Levels of Aggregation A first concern is the specific level at which the analysis should be conducted. This issue needs to be considered before data are actually collected, as the ability to conduct analysis for more than one unit depends on how the data are collected and the size of the sample for each unit. Given adequate pre-planning and sufficient sample size, there are three basic levels at which multicountry data can be analyzed: intra-, inter- and pan-country. 12.1.1.1. Intracountry Intracountry analysis is the most direct and straightforward. Data are analyzed within a country and inferences made about the nature of the relationship between variables within the country. The approach is identical to marketing research carried out in a domestic market where the unit of analysis is the individual. To the extent that intracountry analysis is carried out in multiple countries, comparisons can be made about the relationships in each of the countries. However, any comparisons across countries are made with the knowledge that there may be elements that are not comparable across countries. Observed differences may reflect real differences between countries or simply be a function of differences in sampling, measurement or background factors. 12.1.1.2. Intercountry Intercountry analysis shifts the focus away from relationships within a country to examination of relationships between countries. Analysis is at the country level and the observations are the means of variables from each country. The most critical assumption for this type of analysis is that the individuals sampled are representative of the country. Since the country is now the unit of analysis, it is desirable to have as many countries as possible. However, even if all possible countries are included in the analysis, the maximum sample size would be in the neighborhood of 210. Realistically, the number of observations would be far less due to both the time and cost of obtaining additional data and the high degree of heterogeneity across a wide range of countries. Even within the 20 original OECD countries, there are not only sizable differences but also considerable heterogeneity. A critical issue in this type of analysis is that the variables chosen for the analysis should be representative of the country. 12.1.1.3. Pan-country With pan-country analysis all respondents are grouped together and analyzed without regard to their geographical location. The focus shifts back to analysis at the individual level and inferences are made about the entire sample without regard to the country. If no relationship is found, there still remains the possibility that there is a relationship within specific countries. Further, if a significant relationship is found, it does not necessarily hold for all countries represented in the sample. Each of the three approaches to analyzing the data allows for different inferences and is appropriate for different circumstances. Intracountry analyses are most suited to situations where the firm needs to develop strategies on a country-by-country basis and there is minimal concern for achieving standardization across countries. Intercountry analysis is appropriate when the researcher wants to make inferences about countries and is less concerned about the individuals that make up the country. If there is a high degree of homogeneity in the composition of the country samples, then it may also be possible to make observations about individuals in that country. Finally, the pan-country analysis is useful to see whether a relationship can be found that holds across all individuals and countries. However, the results will tend toward the mean and conceal important variations that exist in different countries or within different segments. 12.1.2. Approaches to Data Analysis The different units of analysis suggest that multicountry data analysis can focus on making inferences within a country or across multiple countries. Analysis can also examine whether or not variables differ in terms of level, typically differences in mean values. This can be done within a country, looking at different groups (within countries) or across countries. Analysis may also be conducted to examine the structure, or relationships, of variables. This can be done either within or across countries (Figure 11.1). Fig. Multicountry data analysis 12.1.2.1. Analysis within and across Countries In analyzing cross-national data, the existence of multicountry, or other data units, implies the desirability of adopting a two-stage or sequential approach to analysis. In the first stage, data are analyzed within countries or other relevant organization units. In the second stage, the comparability of findings across different countries or organizational units is investigated, and the significance of observed differences and similarities examined. In the first, within-country phase of analysis, relationships are examined among the independent variables, for example among different attitudinal or socioeconomic variables; or among various dependent variables, such as purchases of different brands or preferences for different product benefits. This enables identification or verification of relevant constructs to be examined in subsequent stages of analysis. In the case of lifestyle variables, it may be desirable to reduce the number of variables examined; or in the case of preference or purchase data, to identify relevant bundles of benefits or purchases. Then the association between the dependent and independent variables can be studied using standard statistical techniques. In the second, cross-national phase of the analysis, the comparability of findings from one country to another is investigated. An important consideration is whether the comparison is made on an implicit, judgmental basis, or whether differences and similarities are explicitly analyzed. In the first case, the interpretation of the significance of observed variation across countries is based on subjective judgment incorporating previous management or researcher experience. In the second case, analytical techniques are applied to test the magnitude of differences and similarities between countries or units of analysis. Various different procedures may be used. Subsequent discussion focuses on the advantages and limitations of these techniques in multicountry analysis. 12.1.2.2. Differences in the Level and Structure of Variables When making cross-country comparisons, the analysis can focus on either the differences in the level of variables or the structure of variables (Van de Vijver and Leung, 1997). The majority of analyses focus on answering the question: is there a significant difference in variable X between country A and country B? This type of analysis is generally straightforward and employs techniques that are easy to use and interpret, although more complex multivariate techniques can also be employed. Typically, if no difference is found, the researcher concludes that they are the same. If a difference is found, then the researcher seeks to find out why a difference exists. When addressing structure issues the researcher asks a more complex research question: is there a difference in the relation of variables X and Y between country A and country B? Answering questions relating to structure requires the use of more sophisticated statistical techniques and is often more difficult to interpret. The structure of variables often represents complex patterns of relationships and interactions. Some of this complexity is simply the result of the number of variables being considered and some is due to the inherent complexity of multicountry relationships. The discussion so far has implicitly assumed that the country is the relevant unit of analysis. This may be the case when a company is organized on a country-by-country basis and plans strategy accordingly. However, in many instances it may be desirable to aggregate the data along other than country lines. Countries may be aggregated to form regional groupings that are relevant to the firm’s operation. Alternatively, subgroups within a country may also serve as the relevant level of aggregation. It may be desirable to aggregate consumers living in geographical sections of a country. Within Canada, it makes sense to aggregate consumers from different provinces, particularly since Quebec is different in many respects from other parts of Canada. To deal with these inconsistencies, the plan of analysis should start with a clear definition of the appropriate unit of analysis (Douglas and Craig, 1997). In some instances, the country is the appropriate unit. More typically, there are smaller units within a country that form the appropriate unit of analysis. With the units defined, the analysis can be structured accordingly and, more importantly, inferences made at the appropriate level of aggregation. 12.2. Assessing the Differences in the Level of Variables between Countries The remainder of this chapter is devoted to a discussion of statistical techniques used to assess differences in the level of variables between countries. Examples are used to illustrate the use of the techniques in multicountry research. In Chapter 12, techniques that are used to examine the structure of variables between countries are examined. A number of the techniques, for example multiple regression, can be used to consider differences in structure as well as level. The different ways to use each technique are illustrated. The discussion assumes that the reader is already familiar with the underlying statistical formulas and computational procedures, and merely illustrates use of these formulas and procedures in international marketing research, based on examples drawn from available published sources. Readers not familiar with these techniques are referred to standard sources, such as Hays (1994), Winer (1991) and Siegel and Castellani (1988). 12.2.1. Cross-tabulation One of the most common approaches to data analysis in marketing research is the cross-tabulation of data. Typically, two variables are cross-tabulated with each other to see whether there are differences. If the researcher were interested in the relationship between income and age of respondents, the first step would be to code income and age into mutually exclusive discrete categories. The two variables would then be cross-tabulated to see if income differed by age category. The responses to questionnaire data can also be split into mutually exclusive groups, such as users versus nonusers, male versus female consumers, and consumers living in one country versus those living in another. A chi-square statistic can be computed to see whether there are significant differences in the distribution of responses between the two groupings. This is a technique for determining the probability that differences between the expected and observed number of cases in each cell are significant. Where results are cross-tabulated by national sample or by different subgroups within a country, chi-square can be used to test for independence between national samples or subgroups. The researcher might want to determine whether the income distribution varied between two countries. The chi-square statistic would indicate whether the differences were significant. Table 11.1 Comparison of characteristics of commercials in which men and women appear in the US, Mexico and Australia Ad characteristics US (%) Mexico (%) Australia (%) Women Men Women Men Women Men Product user Female 13.0 2.3 25.0 4.0 7.7 4.1 Male 0.6 3.0 6.8 11.9 0.0 4.1 Either 86.4 94.7 69.2 84.2 92.3 91.8 X2 = 13.52* X2 = 19.73* n.s. Setting Home 33.5 22.7 25.9 20.2 28.8 16.3 Store 9.0 6.8 15.5 14.1 5.8 14.3 Occupational 3.6 15.2 0.9 3.0 0.0 4.1 Outdoors 11.4 7.6 12.1 18.2 5.8 14.2 Other 42.5 47.7 45.7 44.4 59.6 51.0 X2 = 16.30** n.s. n.s. Voice-over Female 12.4 11.5 10.9 Male 67.2 66.5 76.9 Chorus 9.5 13.5 9.4 None 10.9 8.5 2.9 * P < 0.001, ** P < 0.01. Source: Gilly, 1988. Table 11.1 contains data on characteristics of commercials in three countries, the US, Mexico and Australia, in which both men and women appear (Gilly, 1988). The table shows at whom the products were targeted (columns) and the characteristics of the ads (rows). In the US there were differences between the product user and the portrayal and gender in commercials, with women appearing in commercials for women or for either gender, but infrequently in commercials for men’s products. There were also gender differences in the setting, with women more likely to be portrayed in the home and men more likely to be shown in occupational settings. For Mexico there were also differences, as indicated by a significant chi-square, in gender portrayal based on the product user. Women were more likely to appear in commercials aimed at women and men in commercials targeted toward men. There was no significant difference in setting in Mexico. In Australia, none of the differences was significant. While the chi-square statistic indicates that there is a significant difference, it does not provide an explanation. One can conclude that there is a greater equality in sex roles in Australian ads. However, the reasons behind the differences observed in the US and the lack of differences in Australia are open to a broader interpretation and would require more extensive data collection to answer. Chi-square analysis is a test of independence and as such provides no indication of the degree of association between two variables. Additional statistics, such as the contingency coefficient, phi, tau or gamma, can be computed to provide an indication of the association between variables. Most importantly, the variables used for the cross-tabulation must be ordinal so that an interpretation of the measure of association is meaningful. If one variable is nominal or both are, then interpretation becomes problematic (see Siegel and Castellani, 1988, for computational procedures and limitations). The example illustrates two-way cross-tabulations. In analyzing cross-national data, the researcher may want to use three-way or n-way cross-tabulations. For example, gender, purchase rate of a particular product and country can be cross-tabulated. The computed chi-square statistic provides an indication of the overall degree of independence. Partial gammas can then be computed to provide a measure of the relationship between two variables, controlling for a third (or more) variable. In the above example, this would allow determination of whether the relationship between gender and purchase of a particular product is affected by the country. 12.2.2. t-tests Often the researcher is interested in whether the mean values of a variable are different between countries. The t-test provides a relatively simple way of testing for differences in means obtained from national samples. A t-test is a statistic that provides a measure of the significance of differences between means drawn from two sample populations. It thus indicates whether there is a statistically significant difference between values on a given variable for two samples. In international marketing research, t-tests can be used to test whether mean scores on a given variable are significantly different between countries. Alternatively, they can be used to test whether different subgroups or segments within countries exhibit significant differences in behavior, attitudes and so on, and the level of significance in each country can then be compared. When using t-tests, or for that matter any time a comparative statistic is used when variables are not comparable, there are a number of cautions. For example, if a survey were conducted in China and the Lebanon examining the relationship of purchasing behavior to income, there would be a significant difference between the two countries on income. In 2002 the per capita GNP in the Lebanon was over four times greater than the per capita GNP of China (Lebanon US$3990 and China US$960). However, if the income levels are adjusted to account for purchasing power parity (PPP), then the differences largely disappear. The PPP adjusted GNPs for the two countries are virtually identical (Lebanon $4660 and China $4520). Thus, the reported mean incomes may be statistically different, but the ‘real’ incomes of the respondents do not differ. The same type of problem can exist when attitudinal or purchase intention data are collected, particularly in one-time cross-sectional research. For example, Germans tend to understate purchase intentions for new products relative to Italians (Bhalla and Lin, 1987). With a single study conducted in both countries, one is likely to conclude that a new product launch in Italy would be more successful than in Germany. In situations where a company has conducted multiple studies over time, it is possible to norm the purchase intentions against past responses to assess how the purchase intentions compare to previous studies. Thus, the higher mean purchase intention in Italy and the lower purchase intention score for Germany may both predict the same level of sales. A potentially more serious problem in using t-tests in international marketing research is the issue of multiple t-tests. Often there are a number of variables that the researcher is interested in comparing between countries. Typically, in international marketing research whenever there is a large set of variables, there will be differences between some variables. The traditional approach is to set a significance level, typically P < 0.05, and test whether differences are significant between all pairs of means. In many instances this is an acceptable procedure, particularly where the variables being tested do not represent a logical grouping of variables and can be construed as being independent observations. However, in situations where the variables cover a range of interrelated variables, the use of multiple t-tests with an alpha level of 0.05 will artificially inflate the likelihood of getting significant results. These variables are not independent of each other but represent a family of variables. To provide an adequate protection level from type I errors requires establishment of the familywise error rate. To control for the familywise error rate, Bonferroni-adjusted tests can be conducted (Hays, 1994). The procedure adjusts the alpha level for the number of t-tests being performed. The critical value of alpha used in all the tests is alpha divided by the number of tests being performed. For example, if the alpha level were set initially at P < 0.05 and there were five t-tests, then the adjusted alpha would be 0.01 and this value would be used for all five t-tests. This procedure controls for the overall type I error rate or the probability of rejecting the null hypothesis that there is no difference between means. Other procedures devised for multiple comparisons, such as Tukey’s procedure or the Scheffe method, can also be used (Winer, 1991). Not adjusting for the number of t-tests being performed increases the likelihood of obtaining significant differences between some of the means. This is magnified by the fact that researchers are typically looking for differences between countries and consequently are not surprised when they find some. The extent of the upward bias is fairly severe. For example, when there are 10 sets of variables and 10 t-tests are performed with an alpha level of 0.05, the probability of obtaining a significant difference increases to 0.5. When there are 20 sets of variables and the alpha level is not adjusted, the probability of finding at least one significant difference increases to 1. 12.2.3. Analysis of Variance Analysis of variance is a very useful technique in international marketing research as country differences can be included as a factor in the design. It can be used to test for the significance of differences within a national sample or between different national samples. One-way analysis of variance tests whether the means of several additional samples are significantly different for a single variable. Two-way analysis of variance extends this logic to the situation where there are two influencing variables. Two-way analysis of variance thus tests for the effect of two variables, as well as interactions between those two variables. (See Winer, 1991, for detailed discussion of procedures and limitations.) The use of two-way analysis of variance in international marketing research is illustrated by a study conducted by Pieters and Baumgartner (1993). In their study, they compared attitudes toward advertising of three groups (practitioners, homemakers and students) in two countries (the Netherlands and Belgium). The results of their study indicate a significant effect of group, country and a country-by-group interaction. Additional analysis in the same study compared the mean scores on 20 attitudinal variables toward advertising (Table 11.2). Since all variables reflect attitudes toward advertising, the significance levels were Bonferroni adjusted. This procedure insures that the alpha level of P < 0.05 is adjusted for all the comparisons being made and that conclusions about statistical significance accurately reflect the number of comparisons being made. Table 11.2 Mean scores of groups in countries on the attitude toward advertising items Mean scores of countries and groupsa Label Advertising Effectsb The Netherlands Belgium 1 2 3 1 2 3 CHE: Makes society cheerful C,G,I 3.3 4.0 2.9 4.1 4.5 2.9 INF: Information about new products C,G,I 4.2 2.7 3.9 4.9 3.7 3.8 NOT: Makes buy things not needed C,G,I 2.3 3.9 3.5 4.1 3.6 4.1 PRO: Is progressive C,G 3.2 3.1 2.6 3.6 4.2 2.9 FAL: False and misleading info C,G 1.6 3.5 3.0 4.1 2.2 3.2 FOS: Fosters the need for products C,I 3.4 3.8 4.0 4.4 4.4 4.2 IRR: Irritates C 3.4 3.6 2.9 2.8 2.5 2.6 AWA: Can influence without awareness C 4.0 3.9 4.2 4.5 4.5 4.4 AMU: Amuses C 3.9 3.6 4.1 4.1 4.3 4.1 ART: Is art C 2.6 3.3 3.5 3.3 3.7 3.8 PRE: Confirms all kinds of prejudices G,I 2.9 3.0 3.6 3.4 2.1 3.5 LOW: Plays on lower needs of people G 2.4 2.9 2.7 3.0 2.7 2.7 CUL: Positive influence on culture G 3.5 3.0 3.0 3.1 4.1 2.9 EXA: Exaggerates advantages G 3.7 2.6 2.6 2.5 4.0 2.3 DIS: Makes people dissatisfied G 2.2 3.1 2.9 3.3 1.9 3.2 BET: Helps to make better decisions G 3.7 2.6 2.5 2.0 4.0 2.3 KNO: Stimulates self-knowledge G 2.7 2.5 2.2 2.3 2.7 1.7 DUM: Keeps people dumb G 1.8 2.6 2.4 2.3 1.2 2.3 EXP: Makes products expensive G 1.9 3.3 2.8 3.7 1.8 3.0 OFF: Is offensive to certain people — 2.5 3.0 2.9 2.5 2.1 2.8 a Groups: 1, Practitioners; 2, Homemakers; 3, Students. Responses: 1, completely disagree; 5, completely agree; 3, don’t know/no opinion. n per group = 60. b Letters indicate statistically significant effects in the univariate ANOVAs: C, country effect is significant; G, group effect is significant; I, interaction effect is significant. All effects are significant at p < 0.05, Bonferroni adjusted (original P/20). Source: Pieters and Baumgartner, 1993. 12.2.4. Analysis of Covariance Analysis of variance is a powerful method of analyzing data from multiple countries. It allows the researcher to examine mean differences between countries and treatments and make inferences about significant differences as well as adjusting for variables that may affect results. In multicountry research there is often the need to account for dramatic differences that exist between countries in contextual factors on sample characteristics that may have an impact on results. There may be concerns that subjects’ responses are influenced by their income levels, age, educational attainment or other factors. Often, it may be these factors that account for the observed differences, rather than the more fundamental differences between countries. There are two ways to deal with this problem. The simplest and most direct is to match the samples from different countries so that they are the same on the relevant background variables. This ensures that the samples are matched on the variables of interest, for example income, but at the same time it may cause one of the samples to be nonrepresentative of the country from which it is drawn. Consider a situation where subjects in the US and Thailand are being compared. The per capita income in the US is more than 17 times that of Thailand (US $35 400; Thailand $2000, per capita GNI; World Development Indicators, 2004). If a research project is being conducted in Thailand and the US, one option would be to select Thai subjects that match the income of the US subjects. Taking this approach will result in a group of Thai subjects who are at the upper end of the Thai income distribution. So in controlling for income level by matching, the researcher has lost the ability to generalize to one of the populations from which subjects were drawn. It should be noted that the per capita gross national income numbers used above are not adjusted for purchasing power parity. The US and Thai income figures adjusted for PPP are $36 110 and $6890, resulting in the US per capita income being only five times greater than that of Thailand. While the income disparity between the two countries is not as great as the unadjusted numbers suggest, it is important to remember that income figures that respondents provide to a survey question do not reflect any adjustments. One procedure to deal with this issue is to express income data relative to the strata into which respondents fall within their country. The other approach involves analysis of covariance, which allows for the effect of a particular variable to be accounted for directly in the analysis. In the example above, the experiment would be carried out using two groups of subjects, one from the US and one from Thailand. In addition to collecting data on the dependent variable, information would also be collected on the subject’s income. The analysis of covariance model provides for statistical control and allows adjustment of the influence of income. The most critical assumptions of the analysis of covariance are that the relationship between the covariate and the dependent variable is linear and that the degree of relationship does not depend on the experimental variable (Hays, 1994). In applying covariance analysis, each national sample is treated as an experimental group. Although this implies nonrandom assignment to treatment groups, this has been found not to generate biased estimates (Overall and Woodward, 1977). The relevant socioeconomic and demographic or other sample characteristics likely to affect treatment response (that is, the dependent variable) are then used as covariates. Next, the group response means and associated F statistics from the analysis of variance are compared with the corresponding means and F statistics in the covariance analysis (that is, after adjustment for the background characteristics). This indicates the extent to which the results are affected by sampling characteristics. National or group response means are adjusted for variance in the covariates, and the impact of such variables is accounted for. The magnitude of this adjustment can also be examined to gauge its overall impact. Caution should be exercised in the interpretation of results where there is the possibility of an association or common factor underlying the treatment means and the covariates. The significance of the F statistics associated with each of the covariates can also be examined to identify which specific sampling characteristics, such as age or income, are most strongly related to the dependent variable. Here again, caution in interpretation needs to be exercised, due to the likelihood of multicollinearity among sample characteristics such as age and income. Also, these are often discrete variables that violate the linearity assumptions of the underlying model. Some preliminary analysis and data reduction procedures, for example factor analysis or a step-down MANOVA procedure, may be desirable (Homans and Messner, 1976). Univariate analysis of covariance was applied in a study examining the impact of administrative heritage on acquisitions of British and French firms made by either British or French firms (Lubatkin et al., 1998). Six covariates were used: relatedness of the merger, growth prospects of the acquired firm, firm’s relative size, local resources used, age of the merger (years) and the merger type (domestic or cross-national). The authors first examined the relationship between the six covariates and the five dependent variables used in the study. A significant relationship was found between the two sets of variables, suggesting that overall the British and French firms appear to establish different headquarters–subsidiary control linkages. The main effect of the firm’s nationality was examined for each of the five dependent variables (managerial transfer, structural control, strategic control, resource control and socialization) while controlling for the six covariates (see Table 11.3, part 2). There were significant differences between French and British firms for two of the dependent variables, managerial transfer and strategic control. The study did not report the results for a test of differences without adjusting for the covariates. Once the authors established that the covariates had a significant effect on the dependent variables, that step would not have generated meaningful results. As a general rule, whenever countries are one of the factors in an experimental design, it is advisable to conduct an analysis of covariance in addition to an analysis of variance. Additional data should always be collected on variables that vary between countries and that might have some impact on the dependent variable. For example, consider a hypothetical analysis of variance where there is a significant main effect for an intention-to-purchase measure for a consumer durable between the US and Thailand. One would conclude that there is a significant difference between Thai and US subjects in terms of their intention to purchase the durable. However, when an analysis of covariance with income as a covariate was run, the main effect was not significant, although the covariate was. The latter analysis portrays a more accurate picture of the phenomenon and allows for more valid inferences about the relationships being investigated. In this type of application, there is the assumption that the regression lines within each cultural group are parallel (Lord, 1967). If the regression lines are parallel, then the relationship between the variables is the same in both countries, although the intercept may vary. This indicates that the nature of the relationship is the same, but the level of the variable, for example strength of intention to purchase, differs. Hsieh et al. (2004) examined the effect of brand image perception on automobile brand purchase behavior across 20 different countries. To deal with concerns about country differences, they used three covariates. Brand awareness and the brand’s market share in each market were used to control for level of familiarity and the popularity of the brand. Whether the brand was foreign or local was also used as a dichotomous covariate. Market share and local origin were significant, but brand awareness was not. 12.2.5. Multivariate Analysis of Covariance Multivariate analysis of covariance is a generalization of the analysis of the covariance model to the case involving more than one dependent variable (Winer, 1991). It explicitly takes into account intercorrelations among dependent variables. Thus, differences significant in a univariate analysis may disappear in an overall multivariate analysis and, conversely, differences that do not appear in a univariate analysis may emerge in a multivariate analysis. In cross-national research, multivariate analysis of covariance can be used when analyzing within-country data to control for the effect of sampling characteristics on the interaction between other variables; for example, different measures of response to advertising commercials for brand users and nonbrand users. Results for different countries can then be compared. As in univariate analysis of variance, multivariate analysis of covariance can also be applied to data pooled across countries, to take out the effects of sampling characteristics or other variables such as attitudinal characteristics when making comparisons between countries. The technique is particularly suited to a more elaborate experimental design with multiple dependent variables. It allows for the impact of experimental factors to be assessed on more than one dependent variable simultaneously. Gibson (1995) conducted a study to examine the emphasis that female and male managers place on leadership behaviors and styles across four countries (Norway, Sweden, Australia and the US). Five leadership behaviors and six leadership styles were examined in the study. The analysis consisted of a 2 Ч 4 MANOVA (gender by country) with 11 dependent variables. The results of the analysis are shown in Table 11.4. Overall significance was tested using F-test approximations. Country and gender were significant, but the interaction was not. After finding significance for the main effects, univariate F-tests for each covariate were conducted to determine the specific factors accounting for the differences. To see whether company size influenced the outcome, Gibson then conducted an analysis of covariance with goal setting as the dependent variable, gender as the independent variable and company size as the covariate. Even after accounting for the effect of company size, gender still had a significant effect on goal setting. A multivariate analysis of covariance was then performed with three dependent variables (goal setting, benevolent autocratic style and laissez-faire style), country as the independent variable, and industry and company size as covariates. Even after adjusting for the covariates, there was still a country main effect. 12.2.6. Multiple Regression Regression analysis is a robust statistical technique that has a variety of uses in multicountry research. In simple regression, the dependent variable is assumed to depend on a single independent variable. In the example in Chapter 4, sugar consumption was the dependent variable and GNP the independent variable. Overall there was a good fit, particularly when the high-income and low-income countries were examined separately. In multiple regression, a number of independent variables are assumed to underlie variance in the dependent variable. In multicountry research, multiple regression can be used to examine the extent to which certain variables account for variation in one variable within countries, and the results compared from one country to another, either qualitatively or by explicit statistical testing. Data from different countries can also be pooled and countries entered as dummy variables in the regression. For further discussion of regression techniques, see Greene (2003). 12.2.6.1. Regression Coefficients One of the issues in using multiple regression in international research is whether to use standardized or unstandardized coefficients. According to Singh (1995) the choice depends on the specific situation facing the researchers. In situations where the researcher has established construct equivalence and the task is that of comparing regression coefficients across two or more groups, the unstand-ardized coefficients should be used. In situations where equivalence has not been established and the objective is to make within-group comparisons, then standardized coefficients should be used. Using standardized coefficients often makes it easier to interpret the results, since all the coefficients are on a common metric and not influenced by the actual scale used to measure the variable. Further, standardized coefficients make possible emic comparisons, since the individual regression coefficients have been adjusted to account for within-sample variability. It is important to point out that this is a within-sample adjustment and uses the same metric within a sample, but not across samples. Use of unstandardized regression coefficients allows for etic comparisons, if construct equivalence has been established. Further, the unstandardized coefficient can provide a more comparable measure across samples since they have not been adjusted for within-group variability. Also, unstandardized coefficients are more likely to be structurally invariant; that is, they are more likely to be the same from sample to sample within a country and hence provide a better basis for comparison. For a more complete discussion of the issues and an example see Singh (1995). Another option that can be useful in multicountry research is to conduct the regression so that the resulting coefficients are elasticities. Elasticities can be computed by expressing the dependent and independent variables as percent change. Alternatively, as the example in Chapter 4 from Rao and Steckel (1998) illustrates, the variables can be transformed into logarithms. Using the logarithms of the variables will also result in estimates of elasticities. The regression coefficients are then interpreted as the percentage change in the dependent variable that would result from a 1% change in an independent variable. For example, regressions could be run in two or more different countries on the relationship between sales and advertising expenditures. The regression coefficients for advertising expenditures, computed as an elasticity, would provide an indication of the likely effect of increasing or decreasing advertising spending on sales in each of the countries. A 1% increase in advertising might result in a predicted sales increase of 0.4%, while a 1% increase in another country might result in a 1.6% increase in sales. With limited funds, the manager would be inclined to allocate more money for advertising in the second country. 12.2.6.2. Using Regression in Multicountry Research When regression analysis is used in international marketing research, it can be employed to address issues related to the differences in the level of variables between countries or issues related to structure or interrelation of variables. When used to look at level issues, hypotheses are tested by examining whether the intercept of a regression equation is the same across different countries. Regression can also be used to examine structure issues as well. In evaluating structure issues, a pan-country regression should be run first using data from all countries. As a second step, a dummy variable is added to the regression for each country. The second regression includes the country dummies, as well as an interaction term for the dummy variable interacting with the independent variable. If the intercept and the regression weights are similar across all countries, then it can be concluded that the relationship holds not only within a country, but also across countries (Van de Vijver and Leung, 1997). In a study that examined the performance of US films in eight foreign markets (Argentina, Austria, Australia, Chile, Germany, Mexico, Spain and the UK) multiple regression analysis was used (Craig et al., 2005). Foreign box-office revenues were predicted based on a country’s cultural distance from the US, the number of McDonald’s outlets per capita, the language spoken locally (English vs a language other than English) and the film’s genre (13 categories coded as dummy variables). Per capita income was used as a control variable. Cultural distance had a significant effect on the performance of films (see Table 11.5). The coefficient was negative (-0.157) and highly significant. Films released in countries that were culturally closer to the US were more likely to perform well. Conversely, films released in countries that were farther from the US in terms of cultural distance did not perform as well. The coefficient for the number of McDonald’s per capita was also significant (0.0399). Countries that exhibit greater acceptance of McDonald’s also tend to place a greater value on American films. In this case it was important to control for a country’s per capita income so that a greater number of McDonald’s restaurants did not simply indicate a higher level of economic development. The genre of the film also had an impact on its performance, with action, fantasy, adventure, animated and horror doing better and family films doing worse. Overall the variables used in the regression explained 58% of the variance in performance of US films in foreign markets. Table 11.5 Estimated regressions for log per capita box office, all countries and by language group (estimated standard errors in parentheses)a All All English German Spanish Nonrandom Parameters Drama -0.0933 -0.0832 -0.269 -0.260 -0.154 (0.0919) (0.0933) (0.085)** (0.060)** (0.176) Romance 0.177 0.171 0.362 0.770 -0.118 (0.129) (0.132) (0.124)** (0.088)** (0.229) Comedy -0.0942 -0.0927 0.0208 0.0362 -0.395 (0.0901) (0.092) (0.084) (0.058) (0.171)** Action 0.179 0.185 0.132 0.303 0.0531 (0.089)** (0.092)** (0.084) (0.058)** (0.171) Fantasy 0.557 0.558 0.462 0.816 0.252 (0.133)** (0.136)** (0.124)** (0.085)** (0.252) Adventure 0.197 0.204 0.0896 0.0356 0.139 (0.107)* (0.109)* (0.101) (0.0682) (0.205) Family -0.287 -0.189 -0.027 0.0216 -0.764 (0.101)** (0.103)** (0.099) (0.0701) (0.189)** Animated 0.281 0.284 0.168 -0.351 0.450 (0.108)** (0.110)** (0.101)* (0.069)** (0.208)** Thriller 0.0827 0.081 -0.132 -0.202 0.172 (0.113) (0.115) (0.107) (0.075)** (0.209) Mystery 0.685 0.679 0.256 1.176 0.803 (0.201)** (0.207)** (0.185) (0.145)** (0.375)** Science 0.162 0.166 0.207 0.251 -0.0478 Fiction (0.117) (0.119) (0.111) (0.076)** (0.228) Horror 0.279 0.294 0.132 0.251 0.276 (0.108)** (0.111)** (0.100) (0.026)** (0.206) Macs Per 0.0399 0.070 0.0274 0.0329 0.238 Cap. (0.0033)** (0.001)** (0.002)** (0.0031)** (0.017)** Cultural -0.157 Distance (0.012)** English 0.140 Speaking (0.0593)** Random Parameters Constant Term = a0 + a1 logIncome + oawa Intercept 1.611 0.845 -2.252 -1.715 -1.211 a0 (1.366) (1.402) (3.022) (0.419)** (0.761) Income -0.386 -0.392 0.181 0.0291 -0.249 a1 (0.169)** (0.174)** (0.419) (0.0521) (0.093)** Std. Dev. 0.172 0.158 0.035 0.509 0.121 oa (0.013)** (0.014)** (0.013)** (0.011)** (0.026)** Coefficient on PCUSBox = b0 + b1logIncome + sbwb Intercept 4.226 4.054 3.328 3.497 1.439 P0 (0.967)** (0.995)** (1.890) (0.301)** (0.526) Income -0.384 -0.363 -0.286 -0.279 -0.047 P1 (0.121)** (0.124)** (0.263) (0.037)** (0.066) Std. Dev. 0.406 0.403 0.445 0.653 0.368 sb (0.011)** (0.011)** (0.009)** (0.009)** (0.019)** Regression Disturbance Standard Deviation Std. Dev 0.940 0.967 0.454 0.319 1.178 (0.0075)** (0.008)** (0.008)** (0.006)** (0.012)** Log L -3042.035 -3094.537 -526.765 -510.2834 -1680.854 Sample 2198 2198 597 559 1042 a * (**) Indicates significant at 95% (99%) significance level. Source: Craig et al., 2005. Multiple regression can also be conducted on data pooled across countries. Each country is then entered as a dummy variable (0,1). If the effect of gender and age on soft drink consumption were to be compared across four countries, a single multiple regression could be run. Soft drink consumption data in the four countries would be the dependent variable; the data on gender and age in each country the independent variables; and the four countries represented by three dummy variables. This would enable examination of whether there were differences between countries; that is, whether the dummy variables were significant. Alternatively, separate models can be estimated for each country and the different equations compared (see Grein et al. (2001) for an example). 12.2.7. Hierarchical Linear Modeling Hierarchical linear modeling (HLM) is a relatively new technique that is ideally suited to handling the type of multilevel data encountered in international marketing research (Bryk and Raudenbush, 1992). The principal advantage of the technique is that it allows for the simultaneous estimation of the relationships of variables at multiple levels. In doing so it allows for the testing of hypotheses on variables measured at the country level and an assessment of how these variables affect relations at the individual or within-country level. More importantly, the use of HLM helps resolve some of the problems inherent in using both country-level and individual-level variables in the same study. Use of HLM deals with the problems of dependency, random effects, hierarchical nesting and cross-level interactions (Hox and Kreft, 1994). The problem of dependency discussed earlier in this chapter with respect to multiple t-test and familywise error rates can be dealt with through HLM, as well as other multivariate techniques. HLM allows specification of a multilevel analysis model that incorporates random effects. An important feature is that it can handle unequal sample sizes as well as different time periods for repeated measures. Also, HLM allows for the effect of cross-level interactions to be estimated directly. This is particularly important where there are hypotheses about how country-level variables interact with individual-level variables. HLM was applied in a multicountry setting by Steenkamp et al. (1999). Data for the study were collected from over 3000 consumers in 11 different European countries. In addition to individual-level data on variables such as attitude toward the past, consumer ethnocentrism and self-enhancement, the researchers used three measures of national culture (individualism, uncertainty avoidance and masculinity). In the analysis they assessed the impact of four individual-level variables and the three national culture-level variables on consumer innovativeness. The analysis also incorporated three interaction terms as well as three covariates (Table 11.6). Both individual-level variables and national culture-level variables were found to be significant, indicating that variables at both levels influence the degree of consumer innovativeness. The individual-level variables explained 12% of the variance while the country-level variables explained 56% of the variance. Further, the interaction between two of the national culture variables with the individual variables was significant. Fu et al. (2004) used HLM to assess the impact of societal and individual beliefs on the effectiveness of different managerial influence strategies in 12 different countries. The four individual beliefs – social cynicism, reward for application, religiosity and fate control – predicted respondents’ perceived effectiveness of the three different influence strategies (persuasive, assertive or relationship based). The measures of national culture were found to moderate the strength of the relationship between individual beliefs and the perceived effectiveness of the influence strategies. 12.2.8. Multiple Discriminant Analysis Discriminant analysis is similar in principle to regression except that the dependent variable is categorical rather than continuous. Discriminant analysis attempts to predict group membership based on a number of independent variables. It also indicates which variables are significant in discriminating between groups. It might, for example, be used to predict the purchase of private versus manufacturer brands, based on variables such as age, income, price sensitivity or purchase volume. The reader is referred to Hair et al. (1998) and Lattin et al. (2003) for more detailed discussion of the underlying assumptions and statistical formulation. In international marketing research, discriminant analysis can be used to identify which variables are significantly different between two or more national samples. Each national sample is thus treated as a categorical variable or group, and differences in relevant independent variables examined. Similarly, it can be applied to test for differences between subgroups or segments within a country, and the results compared across countries. Husted et al. (1996) used discriminant analysis to test whether Mexican, Spanish and US MBAs differed in their moral reasoning, as indicated by responses to the Inventory of Questionable Practice and the Defining Issues Test (DIT). Discriminant analysis was used to determine whether the responses differed between pairs of countries. Only the MBAs from the US and Mexico were found to be different and only the DIT test discriminated between the two groups. Table 11.6 Effects on consumer innovativeness Independent variables Unstandardized coefficient Relative effect size Intercept (γ00) 2.6975* Main effects: individual level Resultant conservation (γ10) -0.838a 0.114 Resultant self-enhancement (γ20) -0.0100 0.031 Consumer ethnocentrism (γ30) -0.17893 0.122 Attitude toward the past (γ40) -0.1210a 0.114 Main effects: national culture level Individualism (γ01) 0.0073a 0.100 Uncertainty avoidance (γ02) -0.0074a 0.101 Masculinity (γ03) 0.0067b 0.075 Cross-level interactions Uncertainty avoidance Ч Res. conservation (γ11) -0.0013b 0.069 Masculinity Ч Res. self-enhancement (γ21) 0.0009 0.042 Individualism Ч Consumer ethnocentrism (γ31) 0.0022b 0.068 Covariates Income (Ч100 ECU) (γ50) -0.0011 0.028 Age (γ60) -0.0060a 0.116 Level of education (γ70) -0.0094 0.020 Explained variance (%) Individual-level 12.3 Country-level 56.2 Note: Test of significance is based on one-tailed test. a P < 0.01. b P < 0.05. Source: Steenkamp et al., 1999. 12.2.9. Conjoint Analysis Conjoint analysis is a technique that allows the researcher to determine how individuals value different components of a product or service offering. It assumes that an individual’s overall evaluation or judgment of an object or product can be broken down into part-worth judgments about different product or object attributes. The analysis provides estimates of the value or utility of different attributes and the different levels of an attribute. Collection of data for conjoint analysis involves obtaining a number of overall evaluations of different combinations of levels on the different attributes. In a single-country study, consumers would be presented with a number of multi-attribute profiles and asked to evaluate each profile. Depending on the number of profiles, they would be asked either to rank order or to rate each of the profiles. These evaluations are then decomposed to assess the utilities assigned to different levels of attributes. This is discussed in detail in Green and Wind (1973). (See also Green and Wind, 1975; Green and Srinivasan, 1978, 1990.) In multicountry research, conjoint analysis can be used to examine buyers’ preferences for, and evaluations of, products or other objects with regard to certain product attributes in different countries. These preferences can then be compared across countries to assess differences and similarities. The following example will help to illustrate the use of conjoint analysis in multicountry research. Consider the situation facing a Japanese manufacturer of medical imaging equipment. The company is developing the next generation of medical imaging equipment intended for use in hospitals throughout the world. To determine the best design features for the new device, a conjoint analysis study was planned for the US that would subsequently be modified for other countries. The study focused on determining the ‘optimal level’ of five product attributes: (1) image quality, six times better than industry standard, three times better than industry standard and equivalent to industry standard; (2) length of time required to produce an image, 1 s, 5 s and 10 s; (3) color image versus black-and-white image; (4) cost, $100 000, $300 000 and $500 000; and (5) manufacturer, GE, Siemens and Toshiba. The levels of each attribute would be combined into the appropriate number of full profiles. For example, profile 1 might be: image quality, three times better than industry standard; length of time to produce an image, 5 s; color image; cost, $300 000; and manufacturer, Siemens. The last attribute is not a product design element, but provides some indication of whether there are differences in the perceptions of different manufacturers. The different medical specialties that use medical imaging equipment might value different attributes differently and there are at least two different constituencies for the study, physicians and hospital administrators. Not surprisingly, one would expect hospital administrators to be more concerned about the cost of the equipment than would physicians. In applying the basic study beyond the US, one would first have to express the cost in local currency units. One would also have to calibrate the quality of image attribute in terms of the prevailing industry standard in that country. This would be less of a problem in Western Europe but more of a problem in some of the developing countries where the current standard is very low. Alternatively, it might be possible to express the image quality in terms of dots per square inch or pixels per square centimeter. However, the researcher would have to make certain that the more technical expression of image quality is understood by everyone who is filling out the questionnaire. In each country that the study was conducted, the company would have two sets of utilities for the five different design features, one for physicians and one for administrators. If the study were conducted in 20 different countries, this would result in 40 different sets. The company is then faced with sorting through all 40 to find groups of countries that value similar design elements. This is used as input into the final design of the imaging device to make it appeal to the broadest possible audiences. Conjoint analysis was used in a study conducted in Nigeria to examine the effect of where a product was manufactured on consumers’ perceptions of the product and their likelihood of purchase (Okechuku and Onyemah, 1999). The central thrust of the study was to determine the effect that ‘Made in Nigeria’ had on respondents’ perception of a product. Two product categories (cars and televisions) and four different countries (Nigeria, South Korea, Japan and either Germany (cars) or the Netherlands (televisions)) were used in the study. Sixteen conjoint profiles were developed for each product based on five different attributes. A main-effects only orthogonal design was created from the 4 Ч 4 Ч 3 Ч 2 Ч 2 overall full factorial design. Data were gathered from respondents in nine major Nigerian cities. Rural populations were not included as they have much lower incomes and would be unlikely to purchase either televisions or cars. The final sample of 1721 had an average income of almost twice the Nigerian average, which was appropriate as these individuals were much more likely to purchase the products being studied. The key variable, country of manufacture (COM), was found to be more important than price, brand name, reliability, safety (cars) and picture quality (televisions). Figure 11.2 shows the part-worth utilities obtained from the conjoint analysis. For both cars and televisions, respondents exhibited significantly lower preferences for products identified as being manufactured in Nigeria. In an early study, conjoint analysis was used to assess buyers’ perceived needs and preferences for car models in Britain, France, Germany and Sweden (Colvin et al., 1980). A sample of consumers in each country was asked first to evaluate a set of 27 product attributes. This was done pair-wise and enabled identification of a reduced set of alternatives. Respondents were then required to rate a new car model on these attributes at two levels: a ‘low’ awareness level, where respondents were only shown photographs of the exterior of the car; and a ‘high’ awareness level, where respondents were given a full photographic and verbal briefing on the interior and exterior appearance, features, performance and so on. Both current and new models were evaluated in this way. From this, utility values for the various attributes in each country were obtained. These showed significant differences between countries in technical advancement and external size, as indicated in Figure 11.3. The utilities obtained for each individual in the conjoint analysis were then used to predict purchase intentions for different model types. Figure 11.2 Nigerian consumers’ country of manufacture preference Figure 11.3 Utility values for four countries for three attributes There are a number of commercially available conjoint design and analysis programs that can be used to conduct studies. Packages are available from Sawtooth (www.sawtoothsoftware.com). It offers PC-based ACA (Adaptive Conjoint Analysis) as well as additional software that allows for multimedia presentation of the stimuli. The software allows conjoint data to be collected efficiently and analyzed easily. 12.3. Summary In multicountry research, as in domestic marketing research, data analysis needs to be carefully planned in advance. The existence of multiple countries in the research greatly complicates the data collection as well as the analysis. In addition to deciding what types of data are necessary for the specific analytical technique, the researcher has to decide on the appropriate unit of analysis. Since management is often concerned with determining whether strategy can be developed that transcends a single country, it is important to have an analysis plan that can incorporate different levels of aggregation. A robust analysis plan should allow for data analysis within a country as well as across countries. Further, it should allow for analysis of smaller units within a particular country. Coupled with this is the choice of whether the level of variables or the structure of variables will be compared between two or more countries. In this chapter, techniques that are commonly used to examine the differences in the level of variables were examined. In commercial research, data are typically cross-tabulated and a chi-square statistic computed. The t-test is also frequently used to determine whether the difference between two country means is significant. Analysis of variance can be used in situations where there is an experimental design or theory to suggest the impact of multiple factors on a dependent variable. Analysis of covariance can be used to control for the effect of some contextual variable on the dependent measure. This can be extended to multivariate analysis of covariance. Multiple regression is used to examine differences in the level of variables, but it can also be used to assess differences in the structure of variables. A particularly powerful technique is hierarchical linear modeling, which allows country-level variables to be combined with individual-level variables in the same analysis. Multiple discriminant analysis can be used to determine what variables differentiate between countries. Finally, conjoint analysis is particularly well suited to designing products or services that have the potential to meet the needs of buyers in multiple countries. 13. Assessing Differences In The Structure Of Variables The techniques described in Chapter 11 provide insight into the differences in the level of variables between countries. They allow inferences to be made about whether the value of a particular variable is greater in one country than in another. Certain techniques such as regression analysis also allow inferences to be made about the strength of the relationship and whether it differs between countries. However, the cross-national researcher is often interested in comparing the nature of the relationship among a set of variables across countries. These questions are frequently the most interesting, but also the most complex. Answering these questions allows inferences to be made about the underlying structure of the relationships and whether it is the same between countries. When only two variables are involved, the simplest way to examine structural issues is to compute the correlation between two variables. However, typically more than two variables are examined at once. This requires the use of multivariate techniques such as cluster analysis, multidimensional scaling, factor analysis, confirmatory factor analysis and structural equation modeling. Also, given the number of variables that are examined at once, it is often difficult to draw a simple inference about differences and similarities and the nature of the overall structure. In this chapter, the various techniques that can be used to examine structural relationships between countries are discussed. Correlational analysis is covered first, as it establishes the foundation by considering the association between two variables. The multivariate techniques rely on measures of association or similarity and examine a range of variables simultaneously. This has the effect of examining relationships while controlling for the effect of the other variables. 13.1. Correlation Analysis Sometimes in multicountry research the researcher is interested in the association between two variables and whether this association holds across a number of countries. The researcher might be interested in the correlation between teenage girls’ age and expenditures on clothing. The correlation could be positive, suggesting that as girls got older they spent more. Alternatively it could be negative, suggesting that they spent less as they grew older; or not significantly different from zero, indicating that expenditures do not vary as a function of age. Once this relationship is established in each country, it can be compared across countries. As a first step in analyzing a set of data, the correlation coefficient provides some indication of the strength of association. An advisable first step in any multicountry research study is to calculate an intracountry correlation matrix of all variables to get some idea of the strength of associations within each country. This analysis suggests what variables are redundant or unique. This also helps identify potential problems arising from multicollinearity among variables, if regression analysis is to be conducted subsequently. It is possible to test directly whether two correlation matrices are significantly different from each other using the asymptotic chi-square test (Jennrich, 1970). With more than two countries this approach for testing differences becomes cumbersome. It is also useful to run a pan-country correlation analysis, as this provides an initial indication of the correlations for the combined sample. If there are a sufficient number of countries, then an intercountry analysis can be conducted where the observations are the country means. Correlation analysis can be used to test hypotheses concerning the strength of associations as well as to examine the pattern of associations. In applying and interpreting correlation coefficients there are some important things to remember. First, a significant correlation between two variables does not mean that there is a causal relation between them. Second, if two variables are correlated, there is still the question of interpreting meaning and directionality. If there is a strong correlation between advertising expenditure and sales one might conclude that advertising was responsible for the sales. Alternatively, if one looks more closely at the way the firm sets advertising expenditures and finds that the spending levels were based on forecasted sales, then in reality the anticipated sales drove the advertising expenditure. The third issue is that another underlying variable(s) may be responsible for the observed relationship. These cautions apply to any use of the correlation coefficient, but can be even more troubling in the international environment. In situations where the researcher is unfamiliar with the research context, there is a tendency to rely on the numbers as a representation of what is actually occurring, rather than to probe more deeply for alternative explanations. When using correlation analysis it is important to conduct both intracountry analysis and pan-country analysis. Each type of analysis provides a different kind of insight. A good example of the use of correlation analysis in multicountry research is provided by Glick et al. (2004). In a major research project, they examined attitudes toward men and gender inequality in 16 countries. They used two different scales, AMI (Ambivalence toward Men Inventory) and the ASI (Ambivalence toward Sexism Inventory). While they examine a number of different variables, only two of the variables from the AMI will be used for illustration, HM (Hostility toward Men) and BM (Benevolence toward Men). When they examined the correlation between HM and BM for the 16 countries, the coefficients were all significant except for one. Correlation coefficients ranged from 0.81 to 0.15 for the male respondents and 0.70 to 0.19 for the female respondents. Interestingly, the mean correlation coefficient was identical, 0.46, for both groups. However, only two of the 32 coefficients were exactly 0.46 and less than a third (nine) fell within ± 0.04 of the average. When correlation coefficients were computed for HM and BM for males and females across the entire sample a different picture emerged. For males the correlation between HM and BM was 0.85 and for female respondents it was 0.75. Both of the correlation coefficients were much higher than the means that were computed based on the within-country correlations. The type of context can also influence the nature of the relationship between two variables. Suh et al. (1998) looked at whether the correlation between life satisfaction and affect was greater for different types of societies. Life satisfaction was operationalized as a ‘global cognitive judgment of one’s life’ while affect balance captured ‘relative preponderance of pleasant compared with unpleasant emotional experiences’. Overall, the correlation was 0.46 between life satisfaction and affect, with the range being 0.07 to 0.74. For collectivist societies the correlations ranged around 0.30 or lower, while for many of the individualistic societies they were around 0.60. Based on this, the researchers concluded that individuals’ emotional experiences in individualistic societies are more strongly related to life satisfaction than they are in collectivist societies. One of the problems in applying the correlation coefficient to multicountry research is that differences exist between correlations obtained at the individual level and at the country level. Country-level correlations can be equal to, stronger than or weaker than those obtained at the individual level. Ostroff (1993, p. 569) asserts that the ‘failure to consider random error and measurement error can result in erroneous interpretations about the strength of relationships among variables at different levels of the analysis’. She maintains that the issue is not so much whether to aggregate data or treat it at the individual level, but the need to understand the consequences of either. Further, when correlations are calculated at the individual level, they should be used to make inferences about individuals. When correlations are made at country level, inferences should be at the country level. To do otherwise can result in the ecological fallacy; that is, incorrectly ascribing country-level characteristics to individuals (Robinson, 1950). The process of aggregating individuals into larger units can increase the size of the correlation coefficient (Hannan, 1971). As smaller groups such as countries are combined into larger groups such as regions, there is increased heterogeneity. This causes the correlation within the larger aggregate to increase. For example, consider a situation where individuals in one country are asked among other things how much they spent to purchase their automobile. In a country such as Sweden, incomes are relatively similar and car choices are as well, where many of the inhabitants purchase either Volvos or Saabs. Thus, there would be a very weak but positive correlation between income and amount spent to purchase an automobile. If this same study were conducted in four additional European countries (Germany, France, Spain and Italy), there would continue to be positive correlations of various magnitudes. If all the countries were aggregated, this would increase the heterogeneity in income as well as the amount paid for automobiles. This would result in a stronger correlation at the aggregate level than was observed at the individual country level. Correlation coefficients can provide a useful first look at the data and provide some idea of bivariate relationships. However, relationships are typically much more complex and influenced by many different variables. For that reason it is often desirable to calculate a partial correlation coefficient. This measures the association between two variables after controlling for the effect of one or more variables. In the example of income and amount spent on an automobile, it might be desirable to control for the age of the respondent. Thus, the partial correlation coefficient would represent the association between income and the amount spent on an automobile, while removing the effect of age. In regression analysis where there is more than one independent variable, the standardized coefficients represent the partial correlation coefficient. In the case of bivariate regression (a dependent variable and one independent variable), the standardized regression coefficient is the same as the correlation coefficient. This is true for the Pearson correlation coefficient, which is the most generally used. Depending on the type of data, a form of the correlation coefficient other than the Pearson correlation may be appropriate. When one of the variables takes on only two values, the point-biserial correlation should be used. When both variables are arranged into dichotomous classes, the tetrachoric correlation should be used, or the polychoric for categorical data with more than two categories. Finally, if one is dealing with ordinal data, then the Spearman rank correlation coefficient is appropriate. For more detail on correlation see Hays (1994) or any other basic statistics book. There are also measures of association that were mentioned in Chapter 11 that can be used when data are cross-tabulated, such as the phi coefficient. 13.2. Means–End Hierarchies A technique with considerable potential in cross-national research is means–end hierarchies. The technique has been applied primarily in single-country research, but has received some application in multicountry research. The technique focuses on the links between the means, which are typically product attributes, and ends, which are the consequences and values associated with the attributes. For example, for perfume, attributes might be fragrance, price and image. Specific consequences of using a particular perfume might be prestige, fantasy or comfort. The values connected with the attributes might include romanticism, sense of beauty and self-satisfaction. Typically these relationships are looked at in the context of a single country. When more than one country is examined, the means–end links can be similar or different. The divergence begins to suggest how these relationships vary across countries. 13.2.1. Collection of Means–End Data Means–end analysis involves a number of different techniques. First, there is the issue of how to collect the data. The basic technique is referred to as laddering and is used to elicit the links between attributes, consequences and values. A thorough but time-consuming way to collect means–end data from consumers is through in-depth interviews. A series of questions and in-depth probes are used to determine the links and associations. This method of data collection is, however, extremely cumbersome, as well as onerous to analyze. Consequently, a number of more structured approaches to data collection have been developed that also simplify data analysis. A simpler approach, proposed by Valette-Florence and Rapacchi (1991), is to use card sorting. Subjects are presented with cards that contain product attributes and asked to sort them into three piles, most important, average and not important. From the most important pile, subjects are then asked to select the most important attributes. This process is repeated at both the consequence level and the value level. To allow for the fact that some attributes, values or consequences were not captured in the original list, subjects can write in their own response and include it in the sorting task. Figure 12.1 Paper and pencil version of laddering interview Both these techniques are time consuming and involve one-on-one interaction with respondents. To increase the efficiency of the data collection procedure a pencil and paper approach can be used (Botschen and Hemetsberger, 1998). This greatly reduces the amount of time required to collect laddering data from each respondent and facilitates collection of multicountry data. Respondents are presented with a series of boxes connected by arrows that facilitate the collection of the laddering data (Figure 12.1). They are asked to list up to four attributes that are important and then indicate why a specific attribute is important to them. Up to three reasons can be given for each attribute. Another pencil and paper approach to measuring means–end chains is the associative pattern technique (APT; ter Hofstede et al., 1998). Respondents are presented with two matrices and instructed to indicate when the intersection of a row and column is relevant. There is an attribute–consequence matrix (AC matrix) that portrays consequences and attributes as the rows and columns, as well as a consequences–value matrix (CV matrix) that has the consequences and values as the rows and columns. For the AC matrix, respondents indicate what consequences an attribute leads to. Similarly, for the CV matrix, respondents indicate the values that result from the consequences. The data collection technique is highly structured and leads to straightforward analysis. A good example of the use of the APT approach is found in ter Hofstede et al. (1999), where it was used to collect data from 3000 consumers in 11 EU countries. A pencil and paper approach is efficient and convenient. Since the respondents are responding freely, it is more driven by their own cognitive structure rather than the researchers’ and avoids interviewer bias (Grunert and Grunert, 1995). Further, respondents can stop the laddering process at any time and do not feel any pressure to complete all levels. This can also be a disadvantage, as there is no opportunity to probe more deeply. There is also a problem of how representative the responses to a pencil and paper instrument are when faced with low response rates. With one-on-one depth interviews, if the sample that is recruited is representative of the population, then the means–end hierarchies will reflect those of the overall population. When using a mail survey, there is always the nonresponse problem. For example, Botschen and Hemetsberger (1998) sent out 10 000 questionnaires and received 1081 completed questionnaires back. The relatively low response rate raises the possibility that those who responded are somehow systematically different from those who did not. 13.2.2. Analysis of Means–End Data A variety of approaches can be adopted to analyze the laddering data. The first step is to have independent judges content analyze the responses for meaning. In the Botschen and Hemetsberger (1998) study the judges identified 33 meaning categories for a brand of clothing manufactured in Austria. The responses for each of the three countries were constructed into 33 Ч 33 asymmetric implication matrices (Reynolds and Gutman, 1988). Indices of abstractness and centrality were calculated and appropriate cut-off levels determined. The next step was to draw hierarchical value maps for each of the three countries. Figure 12.2 shows the hierarchical value map for Austrian customers. The maps for Italy and Germany were similar in many respects. The product attributes ‘natural materials,’ ‘warmth’ and ‘appearance’ were associated in all three countries. ‘Quality’ was an association in all three countries. In Austria ‘Austrian product’ and ‘national pride’ were also associated with ‘quality’, while in the other two countries ‘quality’, was associated with ‘durability’ and did not have the additional associations. Another approach is suggested by Valette-Florence et al. (1997). They used a card sort approach to collect data in France and Denmark on fish consumption. Responses were pre-coded and the ladders were converted into a sequence of numbers. Data were initially analyzed using nonlinear generalized canonical analysis. This attempts to find the best representation of the original variables. The authors liken it to a categorical discriminant analysis. Figure 12.2 Hierarchical value map of Austrian customers Once the stimuli coordinates are obtained a cluster analysis can be performed to determine the number of groups of means–end solutions. The analysis revealed seven predominant chains of attributes, consequences and values. Three chains applied to over 80% of the respondents, taste (30.4%), variety seeking (26.5%) and lack of experience/health (23.2%). Some of the chains were quite complex with as many as five attributes and four consequences. The freshness/nature chain represented 1.1% of respondents and was relatively straightforward. It had two attributes, ‘is fresh’ and ‘the fish has lived a free, natural life’, one consequence, ‘enjoy family meal’, and one value, ‘inner harmony’. The differences between countries were quite interesting. The dominant solution for Denmark, ‘taste’, fit 52.9% of the respondents, but was appropriate for only 10.4% of the French respondents. The dominant solution for French respondents, ‘lack of experience/health’ (43.8%), did not show up for the Danish respondents. Means–end analysis is only beginning to be applied in cross-national research. It has considerable potential to provide insights into the different cognitive structures in different countries. As such, it can suggest how a product ‘fits’ into a particular culture and whether a firm can pursue the same positioning strategy in different countries. It can also suggest segmentation strategies as well as providing insights into marketing mix development. For more detail on means–end analysis, a special issue of the International Journal of Research in Marketing on means–end chains is a particularly helpful source, particularly the article by Grunert and Grunert (1995). 13.3. Cluster Analysis One of the more fundamental issues in examining the structure of variables concerns the similarity and differences between objects or constructs. In international marketing research, clustering procedures are most often applied to countries in an attempt to determine those that are similar to each other. Cluster analysis actually refers to a family of related techniques that can be used to group variables, objects or individuals into clusters based on commonalities. Two major types of cluster analysis can be identified: (1) hierarchical; and (2) centroid. Hierarchical clustering starts by linking together the two most similar objects or variables, based on a measure of distance or similarity, the next two most similar and so on, until all variables or objects are in a single cluster (Johnson, 1967). In centroid clustering, objects or variables are divided first into two groups or clusters, such that, overall, a member of a cluster is more similar to a typical group member than to members of another group. Objects and variables are then divided into three, four or as many groupings as are desired, based on the same principle. Cluster analysis can be used in multicountry research to identify different subgroups within countries, and to assess whether these are similar across countries. It can also be and is commonly used to identify groups of countries that appear to be similar in terms of some relevant factors. For a more complete discussion of cluster analysis, see Everitt (1993) and Aldenderfer and Bashfield (1984). The most critical decision the researcher faces in applying cluster analysis is what variables to select as the basis for the clustering technique. While this decision is often influenced by what variables are available, even within that set there are choices to be made. If clusters of countries are formed based primarily on macroeconomic data such as GDP, exports and so on, then countries with high standards of living such as the US, Japan and many of the Western European countries will cluster together. If the variables chosen reflect population and the land area of countries, then another set of clusters will emerge, with China, India, Russia and the US being grouped together. In both cases the US appears in clusters with ‘similar’ countries. However, depending on the variables used in these cases, the clustering technique places the US in two dissimilar clusters where the other members of one share little in common with the members of the other. In addition to selecting the variables on which to base the clustering, there is also the issue of how to express the variables. Consider a situation where one is interested in clustering countries based on GDP. If total GDP is used, then large poor countries will end up being grouped with small rich countries. In this instance it would be better to express GDP on a per capita basis. When only one variable is used, it seems obvious that such an adjustment is desirable. However, when a larger number of variables are used to cluster objects, this may be less evident. A related issue is whether to use the numerical values or to standardize the values. If there is a variable, such as life expectancy, which does not have a very wide range across industrialized countries, then standardizing it will tend to exaggerate the minimal differences between countries. On the other hand, with a variable such as per capita GDP, which varies widely, standardizing it will tend to minimize the differences. In making these decisions about whether to standardize a variable, the important issue is to understand what the implications are and how they affect the conclusions. One approach to dealing with the measurement issue is to use the correlation between two objects as a measure of similarity. If per capita GDPs of two countries are highly correlated, then the two are considered similar. This expresses everything in common units and is not influenced by the relative magnitudes. A series of correlations would be used as input to capture the range of similarities between the different objects. Cluster analysis can also be used to group countries into clusters that are similar in terms of a set of characteristics. Typically, these clusters are based on macroeconomic data such as GDP per capita, degree of urbanization, population, imports and exports, percentage of the population in agriculture, level of literacy, energy consumption and so on. This assumes that these factors imply similar marketing environments and, hence, lead to meaningful groupings for planning international marketing strategies. Countries can also be clustered based on behavioral and preference data. Cluster analysis has, for example, been used to group regions based on food habits and patterns (Askegaard and Madsen, 1995). In a pan-European lifestyle survey of 20 000 respondents in 16 European countries, respondents were asked 138 questions relating to food preferences and behavior, such as preferences for types of flavors or consistencies, and general ways of cooking, grilled, fried and so on. A factor analysis was used to reduce these questions to 41 latent or underlying factors reflecting general food behavior/attitudes, product-related behavior such as nibbling and drinking habits, and health-related habits. Rather than treat the countries as the units of analysis, the researchers divided the countries into 79 regions. These regions were then clustered based on these 41 factors. This resulted in the identification of 12 clusters. Seven of the clusters were nation states, Denmark, Norway, Sweden, Portugal, Spain, Italy and Greece, and five were transnational clusters, including a Germanic cluster consisting of Germany, Austria and Switzerland, the British Isles, the Netherlands and Flanders, France and French-speaking Switzerland, and Brussels, Wallonia and Luxembourg (Figure 12.4). One of the difficulties in using cluster analysis is the interpretation of the clusters. Typically, once clusters are formed the researcher has to examine the clusters and construct a rationale for the observed clusters. There is also the dilemma of how many clusters to include in the final solution. Often objects will move around as an additional cluster is added to or deleted from the solution. The example used in Chapter 4 to illustrate demand estimation is also a good example of a sophisticated use of cluster analysis (Helsen et al., 1993). First, the researchers factor analyzed 23 country trait variables, mainly macro-level data such as GDP, telephones, air cargo, life expectancy, hospitals and so on, for 10 European countries, the US and Japan. Five factors explained 88% of the variance (Table 12.1). The researchers then calculated factor scores for the 12 countries on each factor. The five factors reflected: (1) MOBIL, overall mobility; (2) HEALTH, the country’s health situation; (3) TRADE, foreign trade activity; (4) LIFE, standard of living; and (5) COSMO, cosmopolitanism. The factor scores were then used in a standard K-means clustering algorithm (PROC FASTCLUS in SAS). Table 4.2 (p. 114) shows the results of the two- and three-segment solutions. In the two-segment solution, the first consists of most of the European countries, while Japan, Sweden and the US form the second cluster. In the three-segment solution, the US by itself forms segment 3 and Japan and Sweden form a segment with the Netherlands and the UK. Looking at the centroids of each segment, it appears that MOBIL is driving the solution, with HEALTH and COSMO playing a lesser role. Figure 12.4 Geographical presentation of the 12-cluster solution Table 12.1 Factor loadings of macro-level country characteristics (after Varimax Rotation) Item Factor 1 Factor 2 Factor 3 Factor 4 Factor 5 Factor 6 (MOBIL) (HEALTH) (TRADE) (LIFE) (COSMO) Passengers .969 .050 -.139 .060 -.104 .115 Air Cargo .943 .133 -.192 -.062 -.167 .121 Newspapers .975 .025 -.104 .113 -.071 .036 Population .880 .225 -.330 -.099 -.214 .041 GDP .187 .149 -.142 .841 .180 -.050 Cars .784 -.172 .196 .355 .150 .360 Gas .904 -.029 -.048 .337 .035 .193 Circulation -.121 .510 -.593 -.118 -.240 .059 Phones .175 .071 -.036 .839 -.009 .275 Imports -.349 -.014 .902 -.160 .087 -.031 Exports -.379 -.033 .897 -.127 -.045 -.106 CPI -.166 -.655 -.236 .258 -.433 .424 Life Exp. -.268 .716 -.247 .287 -.070 -.207 Visitors -.110 .046 -.013 -.028 .982 .087 Tourist Exp. -.401 -.072 .330 .470 .607 -.142 Tourist Rec. -.161 -.032 .050 .011 .969 .018 Pol. Stab. .261 .888 .070 -.066 -.001 .193 High Educ. .910 -.325 .058 .114 -.060 -.197 Hospitals .623 -.128 -.147 -.314 -.032 .598 Physicians -.133 .677 -.566 -.025 -.028 .169 Elec. Cons. .048 -.347 -.057 .761 -.181 -.250 Elec. Prod. .957 .091 -.209 .048 -.143 .054 Educ. Gvt. -.399 -.415 .341 -.164 -.197 -.666 Variance Explained 39.9 16.4 13.3 11.1 7.0 4.6 Cronbach a .979 .752 .993 .783 .899 Source: Helsen et al., 1993. Using factor scores deals with the problem of selecting the appropriate units for the variables. Also, as the authors point out, the use of different variables would result in a different clustering. Specifically, they obtained a different result when they used only macroeconomic variables as opposed to all 23 variables. In this and most applications of clustering techniques, there is the implicit assumption that all the variables have equal weight. More advanced clustering algorithms such as SYNCLUS (DeSarbo et al., 1984) or CONCLUS (Helsen and Green, 1991) allow the assignment of different weights to different variables. 13.4. Multidimensional Scaling In some instances the researcher is concerned with examining the relationship between objects, based on either perceptions or objective characteristics. Multidimensional scaling allows the researcher to develop a mapping of objects. When used with individuals it assumes that they evaluate objects, such as product concepts or other stimuli, relative to each other, rather than independently. A measure of similarity between each pair of objects examined is developed, based on perceptual or preference data. This measure is then used to generate a mapping, indicating the relative proximity of the various objects or stimuli to each other. The mapping may be two-, three- or n-dimensional in character, depending on the number of dimensions that appear to underlie similarity judgments. This is evaluated based on a measure of closeness of fit to the original data each time an additional dimension is added (see Kruskal and Wish, 1978, for a more complete discussion). When perceptual data are used, respondents are asked to make similarity judgments about objects or other stimuli. In a multicountry setting, perceptions of the same objects will often differ dramatically. This may indicate lack of equivalence between objects or fundamental differences in the way in which things are perceived. In addition to relying on perceptions, respondents also can be asked to indicate their preferences. Preference data can consist of ratings or rankings of objects or stimuli on different attributes, such as price or taste. Judgments can also be made under different scenarios, as for example, ‘suitable for the family’ or ‘when entertaining guests’. Preference data can then be combined with perceptual data to develop joint-space maps. The perceptual data determine the initial configuration of objects, and the preference data are superimposed on the mapping. If, for example, PREFMAP is used, the data are shown as vectors from the origin, each representing a specific attribute or scenario. The proximity of an object to a vector (as measured by a line drawn horizontally to the vector) thus indicates its rating on that attribute or scenario. The ideal points of individual subjects or of groups can also be projected on the mapping, indicating the position of the subject’s or group’s ideal object. In international marketing research, multidimensional scaling can be used in a variety of ways. As in the case of other multivariate techniques, it can be used to analyze national data or assess similarities and differences across countries. For example, product positioning or concept evaluation studies can be conducted in a number of countries. Consumers in different countries might, for example, be asked to indicate the similarity of different vegetable- or fruit-flavored crisps relative to other snack products such as regular potato chips, fruit, dried fruit mixes, cookies or salty crackers. They might also be asked to rate these on the basis of different attributes such as fattening, nutritious, expensive or liked by the entire family. Joint-space maps of both perception and preferences can be generated for each national sample. Similarities of mappings can then be compared across countries. Alternatively, analysis can be conducted across countries. The initial perceptual configuration of stimuli is based on data pooled across national samples. This indicates the overall positioning of the products or other stimuli. Preference data are then grouped by country and can be shown as ideal points for each national sample or vector ratings on attributes for each national sample. In the preceding example, an overall perceptual configuration of snack products is developed. Preference ratings for attributes could also be collected for each sample and projected on this mapping. For example, there would be a vector for ‘fattening’ for French consumers, another for ‘fattening’ for German consumers and so on. Multidimensional scaling can also be applied to the types of secondary international data discussed in Chapter 3. In this case mappings of countries can be developed based on similarities in macroeconomic characteristics such as political, economic or sociocultural factors rather than using perceptual data. In studying changes in 16 European countries, Japan and the US over a 28-year period, Craig et al. (1992) used multidimensional scaling to examine changes in patterns of similarity among countries over time. Data on 15 variables such as infant mortality, cost of living, per capita income, electricity production, aviation passengers and telephones in use were used to construct a dissimilarities matrix for the 18 countries at four time periods (1960, 1970, 1980, 1988). The dissimilarities matrix was used as input to the ALSCAL multidimensional scaling routine. This analysis resulted in a three-dimensional solution suggesting that the countries are becoming less similar over time, in spite of increased communication and consumer mobility. Figure 12.5 shows the 18 countries for the four time periods on the first two dimensions, standard of living and cost of living. In 1960 and 1970, the countries were relatively close together and in the upper right quadrant. By 1980 and 1988, the pattern had become much more dispersed. The increased dissimilarity evident in the mapping was also confirmed by greater mean dissimilarity across countries over time. Figure 12.5 Multidimensional map of European countries, US and Japan In international research, multidimensional scaling is useful in suggesting how perceptions of products, brands, activities and so on differ between countries. Mappings can be constructed for each country and the results compared. One of the main issues is naming the different dimensions across the different countries and determining whether these are similar. Since the naming process is judgmental, there is often a tendency to ‘anchor’ on one country and interpret subsequent countries in light of this country. A way to deal with the ‘anchoring’ problem is to have equally knowledgeable members of the research team label the various dimensions independently. Different maps and solutions can then be compared and the differences interpreted and reconciled. With a limited number of countries, it is feasible to construct combined maps using all the perceptual data, and then construct individual maps for each country. This approach has the advantage of providing a common starting point for the comparison. An added value of multidimensional scaling is that it provides graphical portrayals of objects (for example brands) that can easily be communicated to managers. 13.5. Factor Analysis Factor analysis is particularly useful in multicountry research in that it allows examination of interrelationships among a set of variables. Essentially, factor analysis groups together variables into factors consisting of intercorrelated variables. While cluster analysis relies on a measure of distance to group objects, factor analysis uses the intercorrelation among variables to develop the factor structure. There are, however, applications of factor analysis that provide results similar to those obtained with cluster analysis. Q-factor analysis shifts the focus from variables to objects and provides groupings of objects, rather than variables. Exploratory or principal components factor analysis can be used for two basic purposes. It can be used to reduce the number of variables to be analyzed and to ensure that there is no multicollinearity between variables. Alternatively, it can be used to identify underlying relationships or structures within the data. (See Harman (1976) or Gorsuch (1983) for a complete reference on factor analysis.) Various types of factor analysis can be identified. Principal components analysis is the most commonly used form. This identifies factors based on the correlation matrix of variables and forms a number of ‘factors’ that effectively summarize the interrelationships among the variables. A reduced number of variables, those that load most heavily on the salient factors, is obtained that can be used in subsequent stages of the analysis. In Q-type factor analysis, on the other hand, rather than grouping variables, the original data matrix is transposed and objects are grouped based on variables. Given a set of data, it is important to determine whether they are appropriate for factor analysis. As a first step the correlation matrix can be subjectively examined to identify groupings of variables that are highly correlated. This begins to suggest that the variables can be grouped into more homogeneous subsets that constitute unique factors. More formally, the Kaiser–Meyer–Olin (KMO; Kaiser, 1970) measure can be used to examine the homogeneity of the variables to be factor analyzed. Kaiser and Rice (1974) suggest that the overall KMO should exceed 0.80, with a value of 0.60 being acceptable. This indicates that the data are suitable for factor analysis. Often the researcher is interested in comparing the results of a factor analysis of a set of variables in one country with a factor analysis of the same set in another country. There are three ways to determine whether the factor structures are similar: (1) visual and judgmental inspection; (2) target rotation; and (3) confirmatory factor analysis. Visual inspection is the weakest approach and can only begin to suggest whether there is any similarity in the factor structure. Typically, an attempt will be made to see whether the same variables load on the same factors. Unless the factor structures are highly similar in all respects, this will be extremely difficult to determine and different researchers may interpret the same structures differently. A researcher looking for similarity will tend to interpret a complex and ambiguous pattern as being more similar than different. This may involve reaching conclusions on the overall similarity, or focusing on the factors that are most similar and disregarding those that are least similar. An analytical approach to assessing factor structure similarity suggested by McDonald (1985) is known as target rotation. Given factor analyses in two countries, one country is selected as the ‘target’ and the other country’s factor analysis is rotated to maximize the agreement between the two. Once the rotation has optimized the relationship between the two factor structures, the similarity can be assessed, factor by factor, using Tucker’s coefficient of agreement (Van de Vijver and Leung, 1997). The coefficient is similar to a correlation coefficient. Alternatively, the identity coefficient can be used (Zegers and Ten Berge, 1985). Identity coefficient values higher than 0.95 indicate similarity of the factor matrices, while values below 0.90 suggest lack of agreement (Van de Vijver and Leung, 1997). While target rotation is superior to visual inspection, neither is as strong an approach as confirmatory factor analysis. This is discussed in the next section of this chapter. In the study cited earlier of food preferences and behavior in Europe, factor analysis was used to reduce the number of variables included in subsequent phases of analysis. It can also be used to identify underlying latent factors. For example, in a cross-cultural study of consumer values, the LOV (List of Values) instrument consisting of nine personal value items was administered to students and their parents in five countries, the US, Japan, France, Denmark and Germany (Grunert et al., 1994). The data for both students and parents were pooled across countries and principal components analyses with oblique rotations performed to identify underlying dimensions. Both two- and three-factor solutions were interpretable for the student samples, and a three-factor solution for the parent sample. However, it is important to note that when the authors used confirmatory factor analysis it showed poor goodness of fit in terms of similarity of mean values, variance–covariance, factor loadings and factor structure across countries, suggesting no cultural comparability. There are three major limitations in the use of exploratory factor analysis in multicountry research. First, by its very nature it is exploratory. There are no a priori hypotheses that can be tested and no way to test them if one wanted to. Typically, the researcher begins with a set of variables related to the topic being investigated and uses factor analysis to reduce their number based on the interrelations of variables. The final set of variables arrived at is totally dependent on the nature of the initial larger set. Theoretical constructs may guide formulation and selection of the initial set of variables, but this does not provide a way of ‘testing’ the theory. Second, while orthogonal rotation of factors is the commonly accepted method of rotation, some researchers apply nonorthogonal methods. A nonorthogonal rotation will tend to provide a unique factor structure that may be difficult to reproduce. The final limitation relates to the difficulty of comparing factor structures between countries, as revealed in the Grunert et al. (1994) study. As indicated above, target rotation can be used, but it is subject to some criticism in the literature (Bijnen et al., 1986; Van de Vijver and Poortinga, 1994). The most serious limitation of factor analysis is, however, its atheoretical nature, which makes it unsuitable for theory testing. 13.6. Confirmatory Factor Analysis In many instances the researcher has sufficient theory to specify relationships and test theory related to differences and similarities across countries. Confirmatory factor analysis is a particularly useful method to test and refine conceptual models across countries. It allows both a test of the overall model and of very specific relationships among variables to be specified and tested. The researcher begins by specifying a theoretically based model that captures the relationships between constructs. Typically, the theory that guides the specification is one that was initially developed in one country. Thus, in most applications confirmatory factor analysis is implicitly used to determine whether relationships are universal. For example, the researcher may be interested in how the construct of ‘involvement’ influences certain types of purchases in more than one country and how similar this relationship is from one country to another. The constructs used in the research may be single variables or groupings of variables that form a particular construct. The relationships between the various constructs are then expressed as a series of causal relationships that portray the hypothesized links between variables. The focus of confirmatory factor analysis is on the measurement model; that is, developing accurate and reliable measures of constructs. Typical applications include determining the dimensionality of a construct in multiple countries (Netemeyer et al., 1991; Hsieh, 2002); the relationship of one set of constructs to another set in multiple countries (Abe et al., 1996); whether a particular consumer behavior model holds in more than one country (Durvasula et al., 1993); the reliability and validity of constructs across different countries (Keillor et al., 2004); and whether a construct is manifested in the same way in a different country or context (Nijssen and Douglas, 2004). It can also be used to examine measure invariance in cross-national consumer research (see Steenkamp and Baumgartner, 1998). As indicated earlier, with exploratory factor analysis the researcher relies on the intercorrelation among variables to find groupings of related variables. These groupings of variables are referred to as factors. The analysis is heavily data driven and may suggest how constructs are defined or operationalized, but it cannot be used to test a theory. Confirmatory factor analysis allows the researcher to specify relationships a priori and to test them explicitly. Confirmatory factor analysis uses either correlation or covariance matrices as input. The appropriate input data depends on the specific application. When the researcher is interested in the pattern of relationships between constructs, the correlation matrix should be used. The variance-covariance matrix should be used when the constructs’ total variance is of interest. The chi-square test and other goodness-of-fit indicators are used to determine whether the same factor solution is appropriate for different samples. Confirmatory factor analysis also allows the researcher to specify a model and test the invariance of different parameters (Marsh and Hocevar, 1985). Once the initial model has been tested, alternative models can be examined to see whether the overall goodness-of-fit measures improve. The technique is also sample-size sensitive. A sample of 150 to 200 is desirable. With much larger samples, relatively small between-group differences become statistically significant. To deal with this problem, Bollen and Long (1993) have developed goodness-of-fit measures that are less sample-size dependent. More information on the technique can be found in Long (1983a), Hair et al. (1998) and Lattin et al. (2003). In multicountry research, confirmatory factor analysis typically involves testing the covariance matrix of measures for all countries to see whether they are the same (Van de Vijver and Leung, 1997). If there are differences, then a set of hierarchically nested models are tested while successively increasing the number of equality constraints across countries. LISREL (Jцreskog and Sцrbom, 1993) and EQS (Bentler, 1995) are the most commonly used computer packages for confirmatory factor analysis. Hair et al. (1998) provide a clear discussion of how to apply confirmatory factor analysis and relate it to LISREL. 13.6.1. Applications of Confirmatory Factor Analysis Confirmatory factor analysis has been used in multicountry research to assess whether attitudinal and behavioral models hold across countries. Durvasula et al. (1993) examined the cross-national applicability of a model of attitude toward advertising based on data from five countries (New Zealand, Denmark, Greece, the US and India). The purpose was to see whether the relationships among the constructs affecting attitudes toward advertising in general were applicable across countries, or were culturally bound. The model initially conceptualized in the US postulated two indirect and two direct antecedents of attitude toward advertising. Four different models of the relationships between the constructs were tested: (1) the null model, where relationships among observable constructs are set to zero (this also serves as the baseline against which the other models are tested); (2) the original model; (3) an alternative model, which added paths from both indirect antecedents to the direct antecedents; and (4) a full model, which added paths from net function thoughts and net practice thoughts to attitudes toward advertising in general. The alternative model was found to provide the best fit. Durvasula et al. (1993) followed a number of steps in their analysis of multicountry data. First, the equivalence of the measures and the relationships among the constructs were examined at three different levels, a national level, a multigroup level and a pooled data level. Analysis at the national level examines measures and relationships in each country separately. The next step is multigroup analysis, which examines whether patterns of measurement and construct relationships are invariant across countries. Finally, pooled analysis is performed. An important element of this analysis is to ‘deculture’ the data first by standardizing responses within each country before examining whether a common core of relationships exists across countries (Bond, 1988). In general, the pattern of these relationships was found to be invariant across cultures, though some differences were observed in their strength. Confirmatory factor analysis can also be used to examine theoretical constructs between countries. Abe et al. (1996) looked at the self-concept in the context of an interdependent culture (Japan) and an independent culture (US). Three separate scales were administered to Japanese and US subjects: (1) a 23-item self-consciousness scale that measured private and public self-consciousness as well as social anxiety (Fenigstein et al., 1975); (2) a 13-item attention to social comparison information scale (Lennox and Wolfe, 1984); and (3) a 20-item action control scale (Kuhl, 1985). Confirmatory factor analysis was used to examine the reliability, the convergent validity and the uniqueness of each of the components. In addition, the researchers tested 12 hypotheses concerning differences and similarities about how the constructs would operate in the two cultures. For example, the researchers found that Japanese experienced higher levels of social anxiety than Americans, but lower levels of private self-consciousness. In both samples, attention to social comparison information was positively correlated with public self-consciousness and social anxiety. These applications suggest that confirmatory factor analysis can be a very powerful tool for examining multicountry data, ideally suited for dealing with the complexity of the data. It allows the researcher to specify within-country relationships and at the same time allows assessment of between-country differences. More fundamentally, confirmatory factor analysis can contribute to the development of theoretically sound constructs that enhance understanding of phenomena occurring in more than one country. 13.6.2. Multitrait Multimethod Confirmatory factor analysis can also be used to examine multitrait multimethod matrices to determine whether there is convergent and discriminant validity. The multitrait multimethod approach was proposed by Campbell and Fiske (1959). They suggest four criteria for evaluating a matrix that has the correlations of different traits measured by different methods. First, correlations of the same trait measured by different methods should be statistically significant and of sufficient magnitude to warrant further use. This suggests convergent validity. Second, these (convergent) correlations should be greater than those obtained between different traits measured by different methods. Third, the convergent correlations should be greater than the correlation between different traits measured by the same method. Finally, a similar pattern of intercorrelations should be found for the heterotrait– monomethod and the heterotrait–heteromethod components of the matrix. However, Jackson (1969) has suggested that there are difficulties in applying these criteria in practice. He suggests that there are some practical issues such as the sheer number of correlation coefficients and the fact that it is often difficult to get an unequivocal answer to issues of convergent and discriminant validity. He also points out four much more methodological issues, such as the fact that ‘Correlations will fluctuate as a function of the sampling error, as well as of error of measurement and of the respective reliabilities of the variables’ (Jackson, 1969, p. 32). As a solution, cross-cultural researchers have suggested applying confirmatory factor analysis to multitrait multimethod matrices to assess method bias (Van de Vijver and Leung, 1997). Watkins (1989) provides an example of the application of confirmatory factor analysis to assess four different traits (self-esteem, extroversion, anxiety and flexibility) measured by two different methods (questionnaire and rating scale). Six different models were tested and the results are shown in Table 12.2. The sixth model treated the two methods as uncorrelated and the four trait factors as correlated. It was the only model that had an acceptable fit level (chi-square/df < 5.00). Table 12.2 Goodness of fit statistics to hypothesized trait/method models (after Watkins and Hattie, 1981) Models X2 df X2/df I (2-method factors) 391.11 20 19.55 II (1 general, 2-method factors) 133.36 12 11.11 III (4 correlated trait factors) 63.82 6 10.64 IV (1 general, 4 correlated trait factors) 20.13 2 10.06 V (2 method, 4 uncorrelated trait factors) 166.44 12 13.87 VI (2 uncorrelated method, 4 correlated trait factors) 16.23 6 2.71 Source: Watkins, 1989. While the data used in this study were from a single country, the technique can be easily extended to a multicountry situation. The analysis is first done within country and then across countries. The extent to which both measures and traits vary across countries can then be examined. Using a confirmatory factor approach to multitrait multimethod analysis requires that careful thought be given to how the data are collected. The added level of complexity is that not only may the traits being measured vary between countries, but also the methods used to collect data across countries may not be comparable. Cadogan et al. (1999) used the multitrait multimethod approach to examine the convergent and discriminant validity for a scale that measures export market orientation. They collected data from a sample of US executives and a sample of Dutch executives. Their results indicate both convergent and discriminant validity as well as consistency between the two samples. 13.7. Covariance Structure Models Structural equation modeling provides the cross-national researcher with a means to test conceptual models and refine theories. As such, it represents a powerful tool that can help explain relationships between unobserved variables specified by a theory and observed variables measured by the researcher. It has been used extensively by academic marketing researchers and, due to its power and flexibility, it is being used increasingly to analyze multicountry data. As with confirmatory factor analysis, the most common approach to structural equation modeling is to use LISREL (Jцreskog and Sцrbom, 1993) or EQS (Bentler, 1995) to estimate and test the model. For more background on these techniques see Long’s (1983b) monograph on the subject, Hair et al. (1998) and Lattin et al. (2003). Structural equation models and confirmatory models share a number of elements and use the same statistical packages for estimation. It may be easiest to think of confirmatory factor analysis in terms of traditional factor analytical models where there is an attempt to understand the constructs, their underlying dimensionality and the relationship between constructs. Structural equation models, on the other hand, are used in a manner analogous to path models to examine predictive relationships. While this distinction is somewhat arbitrary, as the two purposes may overlap, it is a useful way to distinguish between the two approaches. The Durvasula et al. (1993) and Abe et al. (1996) examples mentioned earlier illustrate the use of LISREL for confirmatory factor analysis. The examples discussed here show how LISREL can be used to establish causal paths and estimate dependencies. Structural equation modeling starts with a theoretically based model. In multicountry research the theory that drives the initial model formulation has frequently been developed in one country and the researcher seeks to answer the question of whether the theory holds in additional countries. The next step is to express the theory in terms of a path diagram that portrays the causal relationships suggested by the theory. The path diagram is then formally structured as a measurement model and a structural model, which incorporate the interdependencies and the dependencies between constructs. The overall goodness of fit of the model is assessed using measures such as the likelihood ratio chi-square, the goodness of fit (GFI) and the root mean square residual (RMSR). Changes in model fit can be assessed by using incremental measures such as the adjusted goodness of fit (AGFI) or the Tucker-Lewis index. Once the fit of the model is deemed acceptable, the researcher is in a position to estimate coefficients and test the fit of competing models. In multicountry research, the challenge is to see whether the same model provides equally good fit across all countries. 13.7.1. Use of Covariance Structure Models Calantone et al. (1996) used confirmatory factor analysis and structural equation modeling to examine factors related to new product success in China and the US with the goal of identifying managerially controllable factors. They gathered data on 142 new product development projects in the US and 470 in China. Baseline data were obtained from an earlier new product development study conducted in Canada on 195 new products (Cooper, 1979). Confirmatory factor analysis was used to refine the number of items in each construct and to ensure the unidimensionality of each construct. Two-group confirmatory factor analysis was then used to assess the measurement models’ equivalence across the US and Chinese samples. The model fit well in both countries with nine of the ten hypothesized relationships being significant. In the US model, the path coefficient for product quality was not significantly related to new product success and in the Chinese sample the path coefficient for technical proficiency was not related to product quality. Figure 12.7 shows the path coefficients for the six variables that are related to new product success for the US and China. Only one of the path coefficients was not significantly related to new product success (product quality for the US sample). In the US sample proficiency of technical activities was most important, while in the Chinese sample competitive and market intelligence was the most important. Figure 12.7 Total effects on new product successes Broderick et al. (1998) used both confirmatory factor analysis and structural equation modeling to examine how involvement in food purchases affects purchase behavior across the five largest European countries (UK, France, Spain, Germany and Italy). The initial step was to factor analyze a 12-item scale measuring respondents’ food involvement. Four factors were identified: (1) normative involvement; (2) situational involvement; (3) enduring involvement; and (4) risk involvement. These four factors were used in a confirmatory factor analysis, first for the pooled sample and then for each national sample. The goodness-of-fit measures suggested that the fit for all six models was acceptable and similar across the six samples. While the factors were the same, the level of involvement with food purchases varied across all five countries, suggesting differences in the level of the variables. A second step in the research was the construction of a structural equation model to examine the mediating effect of involvement on purchase behavior (Figure 12.8). The four types of involvement were shown to mediate the effect of economic status on purchase behavior. Risk involvement had the greatest effect on purchase as well as influence on both enduring and situational involvement. The R2 for the overall impact of the five factors on purchase was 0.66. Further, the findings supported the convergent and discriminant validity of the model and alternative forms of the model did not improve the fit. Figure 12.8 Structural coefficients of path model The virtue of structural equation modeling is that it allows the researcher to find the best model to explain differences. The initial model should be specified based on theory and testing of the theory can proceed while controlling for other factors. When a ‘revised’ model provides a better fit, it is necessary to find additional theory that might explain the new relationship or suggest a revision to the theory. As with confirmatory factor analysis, covariance structure modeling is a technique that is ideally suited to multicountry research. 13.8. Advances in Data Analysis Substantial strides are being made in how data are analyzed. The techniques and examples in this chapter illustrate how multicountry data can be effectively analyzed. In addition, substantial progress is being made along two fronts that are likely to result in even greater understanding and insights into multicountry phenomena. First, more powerful statistical packages have become more accessible and user friendly. Second, new statistical techniques are being applied (or developed) that are well suited to the analysis of multicountry data. These advances mean that once data are collected, there is a greater likelihood that more sophisticated analysis adapted to the complex hierarchical character of multicountry data can be conducted. Extremely powerful versions of SPSS and SAS are available for the PC. This allows data to be uploaded at remote locations and analyzed on a notebook computer. Results and data files can be transferred via the Internet to a central location for integration with data from other countries. In addition to the standard statistical packages, more sophisticated packages such as Mathematica and Gauss are available. Along with the increases in the statistical power and sophistication that are available on the computer, the ease with which the programs can be used has improved as well. In addition, there are a number of relatively new statistical techniques that are beginning to be applied to multicountry data. Hierarchical linear modeling, discussed in Chapter 11, has seen limited use in international marketing research. Its chief virtue is that it allows both individual- and country-level variables to be included in the same analysis. Covariance structure modeling has been and will continue to be used in multicountry data analysis. Its main virtue in multicountry research is that it allows for both within-country and between-country effects to be assessed in the same model. Data warehousing and data mining are two related approaches that have tremendous potential for international marketing research. Data warehousing is a term that refers to the practice of consolidating all relevant data into a unified database that can be used to support decision making (see Banquin and Edelstein, 1996, for more detail). The data acquisition and refinement processes were covered in Chapter 3. The data analysis aspect is referred to as data mining. This covers a variety of techniques that are used to identify patterns in the data. Its use in international markets has been somewhat limited to date, but the potential is enormous. With vast amounts of data on the firm’s marketing activities and performance in multiple countries, data mining techniques can identify common patterns across countries. Many of the analytical techniques that can be used in the data mining stage were discussed in Chapters 11 and this chapter. In addition, there are techniques such as CHAID (Chi Square Interaction Detector) and CART (Classification and Regression Trees). Both techniques sequentially partition a set of data so as to maximize the differences on the dependent variable. The analysis can be run once using country as a variable and once without country as a splitting variable to see which solution provides the best fit. Another data analysis technique that is strongly associated with data mining is neural networks (see Bigus (1996) and Hertz (1991) for more detail). The neural networks approach to data analysis is capable of sifting through large data sets and finding patterns. The most unique feature of neural networks is that they are adaptive. They are able to learn from the initial stages of the analysis and incorporate that learning into the final model. Once a model is arrived at, it can be assessed by how well it predicts or classifies the data. Veiga et al. (2000) used neural network analysis to examine the impact of perceived cultural compatibility on postmerger performance in the UK and France. They were able to predict postmerger performance with 88% accuracy for each sample. Of the 16 items used to make the predictions, five were unique to the French sample, one was unique to the British sample and ten were common. The value of neural network analysis is suggested by the fact that an index constructed from the 16 items was more highly correlated with postmerger performance than one comprised of the 23 original items (0.31 versus 0.23). The potential applications of neural networks in multicountry data analysis are extremely promising. With vast amounts of data being accumulated daily, there is a need for analysis and more importantly incorporation of the results in a decision support system. 13.9. Summary Assessing the differences in the structural relationship of variables between countries is the most complex type of analysis the researcher can undertake. There are a variety of techniques that are well suited to examining whether the structures are similar between countries. Correlation analysis allows the researcher to determine whether the relationship between two variables is of the same strength and in the same direction in two or more countries. Even in more complex analyses, it is often useful to begin with correlational analysis in order to understand the basic patterns. Means-end hierarchies provide an approach to structuring data collection as well as using several related techniques to analyze the data. At the core of the method is a desire to understand how means, typically product attributes, relate to ends, typically the consequences and values associated with the attributes. When the researcher’s objective is to group objects or variables, there are three techniques that can be used: (1) cluster analysis; (2) multidimensional scaling; and (3) factor analysis. Cluster analysis can be used to group objects into clusters based on their similarity. Multidimensional scaling is a very effective way to show the interrelationship of different objects along two or more dimensions. The mapping is graphical in nature and effectively conveys how different objects, such as brands or type of products, relate to each other. Factor analysis can be used to group variables together to help identify composite factors comprised of multiple variables. These groupings may suggest constructs that are common or different across countries. Factor analysis can also be used to reduce a large set of variables to a smaller, more manageable set of variables. One of the limitations of the various clustering techniques is that they cannot be used to test hypotheses. Confirmatory factor analysis is a powerful technique that enables the researcher to explicitly test multicountry hypotheses across countries or contexts. It can also be used to implement multitrait multimethod procedures in multicountry research. A related technique is covariance structure modeling. This examines relationships among variables, with country as one of the explicit factors in the equation. Both techniques make use of LISREL or EQS statistical packages and are quite sophisticated and complex to use. However, both offer considerable potential for providing insights into the complex problems facing international marketing researchers. 14. Challenges Facing International Marketing Research A key theme throughout this book has been the challenges that marketing researchers face when they conduct research in a multicountry environment. On one level they face the challenges of coordinating and controlling research activities spread over vast geographical expanses. On another/ level they face the challenges of achieving comparability and equivalence in sampling, instrument design, data collection procedures, analysis of the data and interpretation of the results. Addressing these challenges adequately ensures that marketing research conducted in a multicountry environment is of the highest possible quality and is adequate to support management decisions or test theory. As markets change and the pace of change quickens, there are some specific challenges that are becoming increasingly salient as research is conducted in the twenty-first century. These added challenges can be viewed within the more general context of what is happening to the businesses that the research community serves worldwide. Businesses are becoming increasingly global in their activities. All firms regardless of their size are beginning to craft strategies in the expanded context of world markets to anticipate, respond and adapt to the changing configuration of these markets. As the ways in which firms conduct their business change, research suppliers need to change if they are to remain relevant and competitive. This chapter is devoted to exploring these challenges and examining how they affect the conduct of marketing research. Emphasis is placed on how developments open up new opportunities and how international marketing researchers might respond to these. The goal is ultimately to conduct research that both enhances understanding of consumers in world markets and guides management decision making. 14.1. The Changing Global Environment Firms attempting to compete effectively in global markets are faced with four interrelated challenges, the challenges of change, complexity, competition and conscience (Craig and Douglas, 1996). The same factors that have an impact on businesses affect the marketing research firms that conduct research for global businesses. Like their clients, they must respond to the challenges if they are to grow and survive. The rapid pace of change implies that the ways in which marketing research is conducted must be monitored continually and adapted to take into account new economic, technological, political and social realities. The interplay of these forces in diverse geographical areas creates a new complexity as market configurations evolve. As firms expand the scope of their operations, their research needs change, both in terms of the types of research they require as well as its geographical extent. Finally, differing standards of ethics in different environments mean that researchers need to establish a consistent code to guide research in diverse contexts. As firms consolidate or enter into new partnerships to provide the geographical coverage required by their clients, new organizational forms emerge, changing the complexion of competition. The most far-reaching challenge is occasioned by the rapid changes in the marketing and mass communications infrastructure, which in turn are the result of and driven by rapid changes in technology. Technological change is pervasive and has profound implications for both marketers and consumers. Change influences not only how and where business is conducted and hence the scope of research needs, but also the way in which research is conducted. Marketing researchers must be able to adapt and incorporate changes in technology into the research process. Many of these changes enhance marketing researchers’ ability to conduct and coordinate diverse research projects. Mastery of the new techniques in turn creates new organizational forms. One of the most dramatic challenges is created by the increased amount of research that will be conducted in emerging market economies. Growth in the industrialized nations is in the low single digits, while countries like China, India and Russia have the potential for rapid double-digit growth. This adds to the complexity of conducting marketing research as the range of research contexts becomes increasingly heterogeneous. As emerging market economies evolve and become more important to firms, there will be an ever-increasing need for sound marketing research to guide decisions in diverse markets. Changes in the competitive environment present direct challenges to marketing research suppliers as they face both a changed and a more intense competitive environment. The increased intensity and accelerated speed of competitive reaction create a dynamic environment for marketing research suppliers. With fewer larger firms competing for multinational clients’ business, a research firm’s ability to respond quickly on multiple fronts is critical. Competitors’ actions also serve to accelerate change and increase the degree of complexity. Figure 14.1 Challenges facing the global marketer Growing awareness and concern with social responsibility and ethical issues require that research firms develop a social conscience and abide by this in conducting marketing research worldwide. The challenge is to conduct marketing research in multiple diverse settings on the highest ethical plane. It is imperative that researchers respond to this challenge, so that the results of research studies are trusted and management has confidence in them as a basis for decision making. This is an overarching challenge that encompasses the other three (Figure 14.1). 14.1.1. Coping with Change: Marketing Infrastructure and Technology Not only is the accelerating pace of change permeating all aspects of life, from daily life patterns and social relationships to changing market boundaries and value delivery systems, but also patterns of change are becoming more fluid and discontinuous. Knowledge and technological obsolescence can dramatically change market configurations and behavior from one day to the next, creating instability and uncertainty, making predictions of future trends and developments difficult if not virtually impossible. At the same time, customers are becoming more mobile and are exposed to new ideas and behavior through global media. They also have instant access to an overwhelming volume of information that is continually changing. Against this backdrop, it becomes ever more difficult for the firm to track consumers’ changing preferences and behaviors and to predict the diffusion of new ideas, products and services. Yet, tracking these changes becomes increasingly imperative for the marketer anxious to respond to rapid change in customer choices and to stay ahead of competition. High-quality marketing research becomes increasingly imperative and at the same time more difficult to design and execute. One of the more visible aspects of change has to do with the changing nature of the marketing infrastructure and the technology that is part of it. 14.1.2. The Changing Market Place Developments in mass communications technology and global and regional media such as CNN, MTV and STAR TV create an environment where certain segments of the population worldwide are developing a common set of expectations, familiarity with a common set of symbols, similar preferences for certain products and services, and an overall desire to improve their standard of living. This globalization of consumers means that it is more difficult to compartmentalize marketing research. The design of research must take into account these new complexities. There is an increased need for global sampling frames, multilanguage versions of questionnaires, well-coordinated research efforts that can be conducted simultaneously, rapid coding and analysis of results from multiple sites, and a means to disseminate research results rapidly. The extent to which the same brands are recognizable worldwide facilitates construction of a ‘universal’ questionnaire. However, while certain brands may be available in most markets, the set of local competitors will vary considerably. Mass media has also helped educate consumers worldwide and given them a consumption vocabulary, as well as educating them on product class attributes and decision criteria for brand evaluations. Differences in per capita GDP will influence their ability to actually consume these products or affect the quantity eventually purchased. The expansion of retailers worldwide is also facilitating marketing research. As chains expand, they incorporate their ‘best practices’ in the new stores. They incorporate POS (point-of-sale) scanner technology, modern merchandising practices and product mixes that both respond to local tastes and reflect the firm’s desire for economies of scale in buying from suppliers. The development of shopping malls where they did not previously exist makes possible mall intercept interviews. Another consequence of the development of the marketing infrastructure is the greater need for marketing research. As firms invest more in the development of the marketing infrastructure, they need more marketing research to guide their decision making. In some countries of the world, notably Africa and parts of Asia, technological developments mean development of the basic infrastructure: roads, electricity, running water, rudimentary transportation and distribution systems. The developments are essential for further development of the marketing infrastructure. Electricity not only powers television sets that carry commercials, but also refrigerators that make possible home storage of perishables. Further, a dependable electrical supply makes possible retail stores with refrigeration for staples and some convenience items, often soft drinks and ice cream. With the arrival of branded products, issues of choice, product attributes and competition become salient and require marketing research. 14.1.3. Technological Change Changes in technology continue to have a dramatic impact not only on how marketing research is conducted but also on how the results are disseminated and incorporated into management decisions. Advances in computer technology have dramatically reduced the time required to conduct marketing research and present results to the client. The impact of technology is felt in two main areas of marketing research, the collection of data and the dissemination of data. Technology such as scanners, the Internet, CATI (Computer Assisted Telephone Interviewing), and CAPI (Computer Assisted Personal Interviewing) are well established in the developed countries and are beginning to be used elsewhere in the world. They represent faster and more accurate ways of collecting data, but do not dramatically alter the practice of marketing research. Developments with the Internet are beginning to change not only the way research is conducted but also how it is disseminated. 14.1.3.1. Computerized Data Capture Whether it is product purchase data or responses to questions, computer technology makes possible the rapid and accurate collection of data. The widespread use of scanners at point of sale means that information on brand sales and market share can be quickly and accurately captured and tabulated. The limiting factor is both the scanner infrastructure at retail and the demand for the services by marketers in a country. An added difficulty is the consistency of product class definition across countries. A brand may be in one product category in one country and in a completely different product category in another. CATI and CAPI greatly facilitate the collection of primary data from respondents. The development of CATI depends on household penetration of telephones. While CATI is used widely in developed countries, its use will spread only as the telecommunications infrastructure in other countries reaches the point where it is feasible. Not only does the infrastructure need to develop, but respondents in a particular country need to feel comfortable answering questions over the phone. CAPI is not infrastructure dependent, but its use is limited due to the cost of the computer equipment as well as respondents’ lack of familiarity with computers. Even if a pen or touch screen is substituted for the keyboard, it will be difficult for individuals with low levels of literacy to respond in that fashion. Even if the computer interface can be made extremely ‘user friendly’, the novelty of the computer may serve as a distraction. Technology will continue to evolve and create innovative ways to present stimuli and collect data. Multimedia CAPI makes possible the presentation of highly complex stimuli and facilitates getting consumer reactions to video and audio stimuli (Thomae, 1995). Developments in virtual reality CAPI will heighten the realism in stimulus portrayal and expand the range of topics on which marketing research can meaningfully be conducted (Needel, 1995). Developments in Interactive Voice Response Interviewing can eliminate human interviewers. On another front, software to translate questionnaires automatically will facilitate conduct of multicountry research; although given some of the subtleties involved with abstract constructs, human intervention will still be necessary. Related to this are developments in textual analysis that will facilitate analysis of qualitative responses and responses to open-ended questions from large groups of respondents. This will be increasingly feasible as questionnaires are administered over the Internet. Respondents, in addition to providing pre-coded responses, can also type lengthy open-ended responses to questions. These will already be in a machine-readable form and can be analyzed using computer software. 14.1.3.2. Using the Internet for Data Access and Collection The Internet continues to evolve and affect business practice and the everyday lives of millions of people (Craig et al., 2003). One of its major impacts has been the ready access it provides to vast amounts of secondary data. Rather than have to visit a traditional research library, the marketer can have virtually instant access to data from traditional sources as well as sources that are only available on the Internet. Web sites can provide all kinds of useful information, on topics ranging from countries to customers. The use of the Internet to access secondary data serves to change the timing of access to data and expands the range of information sources available. It enhances the timeliness of data as some of the sources are updated daily. Another use of the Internet, which is growing rapidly, is the collection of primary data. Data can be collected in three different ways: (1) from visitors to a web site; (2) through electronic questionnaires sent over the Internet; and (3) via questionnaires posted on the Internet to which individuals are directed. When information is collected from web site visitors, individuals are typically offered some inducement or incentive to provide basic demographic data and answer a few simple questions. This use does not provide a projectable sample, but it can be employed to suggest characteristics of those who visit the web site and how they feel about certain issues related to the products or service being offered. More specifically, it indicates how those who chose to provide the information feel. Data can also be collected without the visitor’s knowledge through the use of ‘cookies’. More insidiously, spyware can monitor an individual’s use of their PC and use this data later for some other purpose. This use is controversial as it encroaches on an individual’s right to privacy. The Internet can be used to collect data in a more systematic fashion, which is closer in character to more traditional marketing research practice. Subject to the availability of suitable Internet sampling frames, questionnaires can be administered directly over the Internet. The individual can respond to the e-mail by filling in the blank space in response to a question. More typically, links to a questionnaire are sent via e-mail to a list drawn from the sampling frame. Respondents visit the site and are led through the questionnaire. The Internet represents a means to conduct a survey over a broad geographical scope. Further, the results are available almost instantaneously as the responses are captured automatically and can be analyzed in real time as they are received. This approach is most suited to surveys among respondent populations that are technology literate. However, as use of the Internet becomes more commonplace, e-mail surveys will begin to replace mail and telephone surveys. Progress is occuring most rapidly in the US and Europe and will take place more slowly in other parts of the world. The limiting factors will be the extent of Internet penetration and the availability of sampling frames that correspond to respondent populations that are of interest to marketers. Factors such as overall response rate and item nonresponse will also continue to be important. One key advantage of obtaining results rapidly is that it allows additional sampling to be conducted with enhanced incentives, to compensate for shortfalls in the initial phases. 14.1.3.3. Linking Information via Intranets Perhaps a more important benefit of the Internet for international marketers is the extent to which it facilitates the establishment of intranets. These are company-specific networks that link individuals within a firm. They help link individuals who perform the same job function in the same country and in different countries. Further, there are also links to different functions within the organization, again in the same country and in different countries. These are particularly important for firms operating on a highly decentralized basis. The intranet facilitates coordination and control of activities across many countries. Often information exists in some part of the organization that would be helpful to someone in another part of it. This information can be obtained informally through the use of electronic bulletin boards or more formally through the establishment of ‘best practices’ compendia. Procedures can be established to collect information systematically on effective marketing practices throughout the world. These are put into a standard format and organized. A manager facing a difficult pricing or promotional decision can search the database to see how others have approached the problem and determine what would be the most effective solution. For the database to be effective there must be good editing and data collection procedures at the front end and it must be continually updated so that the information is current. 14.2. Contending with Complexity: Conducting Research in Emerging Markets Complexity in marketing research can be a function of many factors. It may relate to the research design, the respondent population, the sensitive nature of the topic or the sophistication of the analytical techniques. It may also be a function of the preceding factors as well as the range of different environments in which the research is conducted. When marketing research is conducted in multiple similar environments, the same approach can be used in each environment. However, when diverse markets such as India, China, Indonesia, Brazil, Russia and South Africa are part of the research plan, the overall process becomes exceedingly complex. The areas of the world that offer the greatest growth potential for marketing research are the areas outside of Europe, Japan and the US. Currently, 86% of all marketing research is conducted in these three parts of the world. As firms seek to expand globally, markets beyond the industrial triad are being incorporated into their global marketing plans. As firms expand into emerging markets, they need sound marketing research to guide and shape their expansion. However, conducting research in emerging markets is not the same as conducting research in highly developed markets. Firms encounter a range of problems that may affect the validity and reliability of the results. In some instances, the environments are so different that it is extremely difficult to obtain comparable and meaningful results. Research in emerging markets poses challenges to the marketing researcher for a variety of reasons. First, there are dramatic differences in the context in which the research is conducted. This makes the actual conduct of research not only different but often more difficult. A related issue is the increased difficulty of ensuring comparability of results. Finally, there are issues related to the cost of conducting research in emerging markets. Often, this is not so much the absolute cost, but the cost relative to likely sales in that market. 14.2.1. Contextual Differences When one thinks of the things that make conducting research in the US, Europe and Japan relatively straightforward, they relate to the availability of a well-developed market research infrastructure, a general acceptance and understanding of marketing research, familiarity with the conventions used to collect data, a communications and mail system that facilitate collection of data, common language within defined groups, high levels of literacy and so on. The factors that are taken for granted in developed countries often do not exist in emerging markets. Table 14.1 summarizes some of the key contextual factors that affect the ability (or desirability) of conducting marketing research. First and foremost is the lack of a well-developed marketing research infrastructure. Many of the factors are interrelated. Often the reason that the marketing research infrastructure – specifically organizations that conduct and facilitate marketing research – is not developed is that the standard of living is so low that firms are not interested in assessing market potential or entering the market. In some instances, there may not be a large number of marketing research organizations in a specific country, but an adjacent country with a more developed research infrastructure may be relied on. For example, bilingual marketing researchers from Germany could be used to design studies, train interviewers and so on in nearby Eastern European countries that may lack the same degree of sophistication. This is going to be more difficult in parts of Asia where the countries are not only geographically dispersed, but there may also be greater cultural differences. Table 14.1 Contextual factors that have an impact on the ease of conducting marketing research Development of the marketing research infrastructure (primarily number and size of research suppliers) Proximity to countries with well developed marketing research infrastructures Level of overall economic development Level of communication infrastructure development Level of development of marketing and distribution infrastructure Degree of urbanization Homogeneity of inhabitants and language There are other factors such as the degree of urbanization that facilitate marketing research. Developing countries that have a large percentage of the population in urban areas are easier environments in which to conduct marketing research than countries where the population lives mainly in rural areas. A higher degree of urbanization means that even if there are not good lists for sampling, the respondents can be contacted through other means. The extent to which the population of a particular country is relatively homogeneous and speaks the same language facilitates the collection of marketing research data. A well-developed general communications infrastructure will also facilitate the conduct of marketing research. A high level of telephone ownership and an efficient mail system make marketing research easier. Related to this is the mass communication infrastructure. If print and broadcast media are well developed, then consumers will have knowledge of a range of products and services. This general exposure to information will enhance their ability to respond in a meaningful way to questions on a questionnaire. Finally, the development of the overall marketing infrastructure influences the ability to conduct marketing research. It is difficult to conduct mall intercept interviews without the existence of shopping malls or their equivalent. Further, if scanners are not widely used it is difficult to provide the types of data and analysis that Nielsen and IRI routinely provide to their clients. 14.2.2. Comparability In addition to contextual factors, there are a series of issues that relate to the difficulty of achieving comparable results in emerging market economies. Throughout the book the issues of comparability and equivalence of results have been key themes. Many of the factors that make it difficult to achieve comparable results are tied to some of the contextual factors identified above. However, they also relate to fundamental cultural and developmental differences that exist between highly developed countries and emerging market economies. There are major differences in the level of literacy between countries. Most, although not all, of the differences relate to the level of economic development. More highly developed economies tend to have higher levels of literacy, while developing countries tend to have much lower levels of literacy. As indicated in Chapter 1, in some of the poorest countries in the world less than 50% of the population is literate. Consequently, when conducting research in countries with low levels of literacy, it will not be possible to use a mail or self-administered questionnaire. Where the research is also being conducted in developed markets, this will necessitate development of a new instrument that relies on pictorial or other types of stimuli. Basic familiarity with the stimuli can influence the ability of respondents to provide meaningful data. For example, Serpell (1979) administered a pattern-copying task to children in Zambia and the UK. Their ability to copy a pattern was assessed in two ways: (1) using a pencil drawing; and (2) using wire to model the pattern. The children from the UK performed better when using a pencil to copy the pattern, while the Zambian children performed better when using wire to model the pattern. Wire modeling is a popular pastime among Zambian boys and suggests why their performance was superior. The impact of stimulus familiarity is also illustrated in a study by Deregowski and Serpell (1971). They gave Scottish and Zambian children two tasks: (1) sorting miniature models of animals and motor vehicles; and (2) sorting photographs of animals and motor vehicles. There were no differences between the two groups when the task was sorting the actual models. However, the Scottish children performed better when the stimuli were photographs. The greater familiarity of the Scottish children with the two-dimensional portrayal of objects accounts for the difference in performance. These types of differences exist in different forms in many developing countries and require special attention when designing instruments for data collection. 14.3. Cost The cost of conducting marketing research in emerging markets can vary dramatically. The absolute cost of conducting marketing research can be very high or very low. When there is no local capability to conduct marketing research, all the design has to be done on site but with foreign nationals who are temporarily stationed in the emerging market. In addition to salary costs, all living expenses would be part of the cost of conducting the research. Further, once the research is designed, research supervisors have to be brought to the country and considerable time spent training interviewers so that they can do the fieldwork. At the other extreme there are emerging market economies, such as India, that have a well-developed marketing research industry that is capable of conducting high-quality research at a low cost. The information presented in Chapter 2 suggests that on average marketing research can be conducted in India or Bulgaria at about a third of the average cost of conducting research. The absolute cost of conducting research is only one half of the equation. The other component is the value of research in terms of the types of decisions it will facilitate and the sales over which the cost of the research will be spread. A usage and attitude study may cost $60 000 in two different countries. In one country the brand may have sales of $300 million and in another country sales of only $30 million. In the first country, the research expenditure is spread over ten times the sales revenue as in the second country, and is only 0.02% of sales. In the second country the research amounts to 0.2% of the brand’s sales. From the manager’s perspective, the research cost ten times as much. Since this would not be the only research expenditure, the manager may look for ways to limit expenditures and possibly conduct less research than might be necessary. 14.4. Confronting Competition: Marketing Research Services in a Global Environment One of the greatest challenges facing marketing research firms is to develop an effective strategy to remain competitive in the changing global environment. The multinational firms that are their clients are dealing with the rapid changes and increased complexity of global markets. Research firms must be able to meet the changing needs of these firms as they expand globally. At the same time, there are also major changes that have been occurring in the marketing research industry. The industry is consolidating, with the top 25 research organizations accounting for around 65% of the total market. Despite this concentration, with the exception of Research International, Millward Brown International and Taylor Nelson Sofres, the major firms do not have significant global presence (Barnard, 1997). Most are organized on a national or regional basis and have operations in the US and the major countries of Western Europe. While the research industry is somewhat concentrated, it remains highly fragmented, with over 3000 serious research companies worldwide (Barnard, 1997). The large number of firms has resulted in many of them competing on price for ad hoc research studies, which has eroded profit margins. Competition is taking place on a variety of fronts. Nielsen and IRI continue to battle it out in the market for data on market share of consumer packaged goods in the US and Europe. Specialized firms such as IMS continue to expand their service for their clients in the pharmaceutical industry. Smaller specialized research firms continue to focus their efforts on the needs of companies in specific industries. There is also considerable growth of marketing research for firms other than consumer packaged goods. Marketing research is becoming increasingly important for business-to-business marketing and for the financial services industry. One of the more important aspects of competitive change, particularly from the standpoint of this book, is that increasingly competition is taking place on a regional or global scale, rather than being confined to one country. 14.4.1. Organizational Options Business organizations are restructuring themselves, making acquisitions and entering into strategic partnerships. One consequence of the consolidations on the client side is that before a merger there may have been a need for two marketing research organizations to service the two firms. After the merger, there may only be a need for one. On a more positive note, as firms make cross-border acquisitions, they have a requirement for a research supplier that has the capability of doing research in the new countries. Further, since many of the firms are highly decentralized, each division or product group will have its own set of preferred research suppliers. To remain competitive, research organizations must find ways to expand the geographical scope of their operations to meet the ever-changing needs of their clients. The critical issues are: (1) market presence and access – having or appearing to have the capability to conduct marketing research in country X, as well as access to the resources necessary to conduct marketing research in that country; (2) market knowledge – sufficient knowledge of a country to be able to plan a research project that takes into account local nuances and idiosyncrasies; and (3) local capability – the ability to execute a marketing research project in that country, including all the fieldwork. While these three components are not independent of one another, all three must be present for the firm to be credible. Each of these capabilities can be achieved in a number of ways. If the research firm does not have offices in some of the new countries, it can either form strategic partnerships with local research organizations, acquire an equity position in a local supplier or establish an office. Each of these approaches has to be looked at in terms of what it provides and how it enhances the firm’s ability to supply high-quality market research. As an initial step, firms will typically enter into strategic partnerships with local suppliers. This can be done on a project basis or may entail some more formal relationship that involves sharing of resources and an understanding that the relationship will involve multiple projects. These relations can also be reciprocal, where two firms agree to provide research capability for each other in their respective home markets. Another advantage of strategic partnering is that it allows research to be conducted immediately and provides considerable knowledge of the local environment. However, it does not give as much control over local operations and may not offer as sound a basis for future growth. Based on a favorable first experience with research in a particular country market and the growing importance of a particular market, a research firm may decide to take an equity position in its partner. Sometimes this is necessitated by increased capital requirements as the local partner attempts to expand and at other times simply by the growth ambitions of the larger firm. At the extreme, a firm may acquire a local firm. Negotiating an acquisition may take time and there is also some uncertainty as to how effective the acquired firm will be after the acquisition. Often there will be multiyear performance goals for the acquired firm, with the full price being contingent on meeting the goals. Starting a greenfield operation in a country takes time and it is likely to be difficult to find qualified personnel. If the country is going to be very important for the future, typically a current employee who speaks the language can be charged with setting up the new office. Recruiting and training of additional employees can then occur at the local level. This process takes time and often it is easier to acquire or take an equity position in a strong local firm. The local partnership also removes a potential competitor from the local market and creates an important ally. 14.5. Conforming to Conscience: Ethics in International Marketing Research So far the discussion has focused on ways to ensure that the research carried out in international markets is of the highest possible quality. Having data that are comparable, equivalent reliable and valid allows conclusions to be drawn that form the basis for sound business decisions or build on existing theory. An overarching issue concerning the conduct of research, whether by academics or commercial firms, is that the research process adheres to the highest possible ethical standards. When research is conducted in multiple countries, there may be more than one set of ethical standards. Further, what is considered ethical in one country may be considered highly unethical in another country. This section will not attempt to resolve the inherent differences in different religious, moral and ethical traditions. It will look at the virtue of conducting research according to high ethical standards so as to ensure the integrity of the research results. Much has been written on the subject of ethics in marketing (Schlegelmilch, 1998; Smith and Quelch, 1993), as unethical or illegal behavior is often manifested in the market place. Most typically, the unethical behavior involves some aspect of the marketing mix, deceptive advertising, price fixing and so on. Unethical behavior can result in negative publicity in the media, loss of consumer confidence, and in some cases legal action. Ethics are also extremely important when it comes to conducting marketing research (see Kimmel, 2001; Kimmel and Smith, 2001). Just as unethical market-place behavior can undermine trust and confidence in the firm and its products, unethical marketing research practices can undermine trust and confidence in the research and the research process. If research is not conducted on the highest ethical plane, then those who are asked to participate will decline to do so, those who supply research will find that their results are called into question or ignored, and those who need sound research on which to base decisions will not trust the results, or increasingly not commission research in the first place. Ethics in marketing research can be viewed in terms of those who are involved in the research process. The specific issues relate to the four main parties to the research process: (1) the respondent; (2) the interviewer; (3) the research supplier; and (4) the client. Each party has a specific role in the research process and a different stake in the outcome. Respondents play a relatively passive role in the process and have limited interest or knowledge in the outcome of the research. Ultimately, they may benefit in the aggregate through, for example, products that better meet their needs or enhanced services. However, their participation is voluntary and critical to effective research. The interviewer and research supplier gain economically from the conduct of research and are concerned about repeat business from the clients that hire them. This depends on producing research that can be relied on and is valued by the client. In assessing value most clients consider the quality of the research, but there are some that may also be swayed by the outcome; that is, whether it supports a planned course of action. The client needs to know that the research that is conducted will be valid, reliable and provide a sound basis for important business decisions. The client is also concerned that the research is worth the expenditure that has been made. Figure 14.2 Multicountry research. Solid lines represent direct transfer of information. Dotted lines represent influence on the process. Arrows represent the direction of the flow while two lines indicate exchange of information at different points in time. The flows of information are influenced by expectations, knowledge, standards and cultural backgrounds While these issues are far from straightforward when conducting research in one country, they take on an added complexity when the research is conducted in multiple countries (Figure 14.2). In multicountry research, there may be multiple clients, multiple research suppliers, certainly multiple types of interviewers, and respondents who differ on a variety of dimensions. Each of these parties will have different expectations, cultural background, standards and so on. Further, each of these components will interact with other components. For example, a research supplier from one country may have more difficulty in understanding the research requirements of a client from another country. The types of research being conducted can range from large, multicountry studies to those involving a few contiguous countries. Typically, one entity within the firm will commission the research to help solve a particular problem facing it. If the issue is developing a global brand, corporate headquarters will be the client and very actively coordinate the research. However, there may be heavy involvement from the country managers where the research is actually being conducted. This participation will help ensure ‘buy-in’ at the local level. Other marketing research projects may be organized regionally and conducted on a more decentralized basis. When the research project is more decentralized, it is more difficult to ensure that the same ethical standards are applied throughout. ESOMAR has established a set of guidelines for the conduct of research to help ensure that it is conducted on the highest ethical plane. The first code was published in 1948 and has been revised periodically. The most recent version was prepared in 1994 and is available on the Internet at www.esomar.org. In 2001 an amendment to the code was added to cover the data protection principles established by the EU Data Protection Directive. In addition to the code of conduct, ESOMAR has established guidelines for different types of research. These are available at its web site and cover areas such as the distinction between marketing research and direct marketing, pharmaceutical marketing research, and conducting marketing research on the Internet. To enforce the code of conduct, in 2000 ESOMAR established disciplinary procedures. These consist of a Professional Standards Committee (PSC) and a Disciplinary Committee (DC). Both committees are charged with examining alleged infringements of the ESOMAR Code of Marketing and Social Research Practice at the international level. The PSC can issue a warning, a reprimand, or refer the matter to the DC, which can impose more stringent penalties, including suspension, expulsion or notification of appropriate authorities. Details of the procedures can be found at the ESOMAR web site. The ESOMAR code covers the rights of respondents, and by implication the role of the interviewer, the professional responsibilities of researchers, and the mutual rights of researchers and clients. It also specifies that research studies must comply with national and international legislation applicable in the various countries. The next three sections are based on the ESOMAR code. 14.5.1. Respondents and Interviewers The rights of the respondent are the most fundamental to the marketing research process. Without respondents, there would be no research. Further, the primary means of obtaining information from respondents is through interviewers; although this is changing somewhat as data can be collected through the mail or directly over the Internet without the intervention of an interviewer. However, the general ethical concerns are relevant, whether or not an interviewer interacts directly with the respondent. To begin with, respondents’ participation in the research project must be entirely voluntary and it is unethical to mislead them in an attempt to gain their cooperation. Respondents must also be assured that their responses will be held in strict confidence and that their right to privacy is ensured. A related aspect is respondent anonymity. If data are to be passed on, subject to respondent approval, then steps must be taken to ensure that the identity of the respondent is not revealed. Respondents must not be harmed in any way by the research process and they must be informed if observation techniques or recordings are to be made of their responses. Mechanisms must be put in place to allow respondents to verify the identity of the researcher and that the researcher is connected with a legitimate research organization. Special care must be exercised when children or young people are interviewed. In most cases this will involve obtaining parental consent. There is a separate set of guidelines for interviewing children and young people, available at the ESOMAR web site. 14.5.2. Researchers The researcher has certain obligations to the client and more broadly to the research profession. Researchers have to conduct themselves so as not to discredit the marketing research profession or do something that would lead to the loss of public confidence in the research profession. Researchers have to be truthful in making claims about their skills and experiences so as not to mislead clients. This can be problematic in multicountry research as it may be more difficult to interpret or verify claims made by research organizations. Researchers must always endeavor to design and conduct research in the most cost-effective manner. Given the different cost structures in different countries discussed in Chapter 2, this may be more difficult than it appears. Fieldwork may be conducted in a certain country because of the need to understand that particular market. However, the tabulation and analysis can be conducted in another location, which may be more costly. The issue is whether this is being done for the convenience of the researcher or because it provides the best approach for the client. Security of the data is a key issue. If the right of respondents to privacy is to be preserved, then the researchers must take steps to ensure the security of all data that they have collected. Also, researchers have an obligation to see that any research findings that are disseminated are adequately supported by the data. The interesting dilemma is where results obtained in one country, for example country A, may be disseminated in country B, as being indicative of responses in country B. However, this may be less of an issue for the researcher and more one of how the client uses the research. There are a number of issues that relate to the conduct of the researcher, independent of any direct interaction with the client. For example, the ESOMAR code specifies that ‘Researchers must not unjustifiably criticize or disparage other researchers’. In addition, research firms need to establish a clear delineation between their research activities and any nonresearch activities undertaken. This can be a problem if a research firm has a large database and also engages in direct marketing and promotional activities. As long as the research information about individuals is not used in the nonresearch activities, there is no ethical breach. 14.5.3. Clients and Researchers The client enters into a specific contractual agreement for a particular research project. The contract spells out all the terms and conditions covering the specific research engagement. All provisions of the contract should be consistent with ethical guidelines established by ESOMAR, the American Marketing Association or another professional body. The researcher must inform the client if the collection of the data is to be done in combination with the data collection for another client. The researcher must inform the client if any part of the research process is to be subcontracted to others. Both of these provisions fulfill the need for full disclosure so that the client is fully aware of how the research is being conducted. In addition, documents that the client provides to the researcher remain the property of the client and should not be disclosed to others. Typically information provided to the client, such as research proposals and cost information, remains the property of the researcher and should not be disclosed to others. The research organization should maintain records of the research for a period of time after the research is conducted. This is likely to present more of a problem when the research project is highly decentralized and spans a number of countries. Further, the researcher should not disclose the identity of the client for whom the research is being conducted. The above guidelines and obligations of the various parties establish the minimum standards for the conduct of ethical research. Specific countries may have more stringent laws that prohibit certain practices, for example pertaining to the privacy of individuals. Problems of misunderstanding, particularly between the client and the researcher, are most likely to arise outside the Western European countries, Japan and the US. The research infrastructures are less sophisticated and the amount of research being conducted is far less in less-developed countries. Consequently, things that are taken for granted in the major markets may require special arrangements. 14.6. Summary Change is occurring in virtually all aspects of business and personal life. Businesses are both being buffeted by change and acting as change agents to bring about change in markets throughout the world. Consumers face a much more complex consumption environment and have more choices than ever before. These changes are being played out at different rates in different parts of the world. Against this backdrop, marketing research firms are being challenged to conduct research that is of the highest possible quality, as quickly as possible, in multiple diverse settings. The issues that marketing researchers face are multifaceted and relate to where and how research will be conducted, who the respondents will be, and the tools and techniques that will be used. To prosper and grow, marketing researchers must find creative ways to harness the new technologies to facilitate the conduct of research and enhance its value to clients. At the same time, research organizations must begin to develop the capability to conduct marketing research simultaneously in the developed and the developing world. Increasingly, multinational marketers are designing and selling international brands and need research to guide their decision making across a diverse and disparate world. To accomplish this, marketing research organizations need to acquire the capability to meet their clients’ needs through strategic partnerships and acquisitions. Finally, marketing researchers must strive to ensure that the research they conduct adheres to the highest ethical standards, so that it is trusted and relied on by their clients. 15. Future Directions In International Marketing Research With the accelerating pace of market globalization, communication and movement of firms, people and goods across national boundaries, the need for information on international markets continues to grow. This has to be timely, reliable and accurate as well as to furnish an adequate basis for making complex decisions in today’s fast-paced markets. Yet, it is not enough simply to keep pace with the latest technological developments in collecting and delivering information: rather more fundamental issues relating to the design and comparability of information collected in multiple and diverse environments have to be addressed. In particular, issues relating to the comparability and equivalence of theoretical constructs or research questions, developing the research design and measurement instruments, and selecting appropriate analytical methodologies become more pressing and pertinent. As an increasing volume of research is conducted in cross-cultural psychology, comparative sociology, political science and other social sciences, as well as in consumer behavior and marketing, more knowledge is accumulated about variations in behavior in a broader range of environmental and cultural contexts. Frequently these are, however, directed toward examining the universality of theories and constructs developed in a single sociocultural context or anchored in a specific research philosophy or paradigm. More attention is needed to identify issues or constructs specific to a given context or situation, and to assess the cultural embeddedness and dependency of constructs. Such concerns imply a need for greater rigor in the design of international marketing research. This entails explicitly building in procedures to examine and evaluate the comparability and equivalence of constructs and phenomena in different contexts, as well as to ‘decenter’ the impact of a dominant culture or research philosophy on research outcomes. In addition, theories and constructs need to be studied in a broader range of environmental contexts to examine variation under different conditions (Triandis et al., 1972). At the same time, attention needs to be paid to clearly defining the relevant unit of analysis so as to isolate the impact of contextual and other influences on the phenomena studied. Finally, more emphasis needs to be placed on developing more rigorous and better calibrated measurement instruments and methods of analysis. This entails greater attention to pre-testing and calibrating measurement instruments. In addition, multiple methods of analysis should be adopted in examining the patterning of the phenomena studied in order to assess and eliminate potential method bias. Each of these issues is next further probed in more depth and some directions for future research suggested. While the discussion is not intended to provide an exhaustive list, it is nonetheless intended to establish important priorities for future research. These need to be addressed if further progress is to be made in the development of more rigorous, reliable and useful research on international markets. 15.1. Comparability and Equivalence Revisited Comparability and equivalence issues have been a central theme throughout this book. These have traditionally been and remain a fundamental concern of cross-cultural and comparative researchers throughout the social sciences. It should be noted that considerable ambiguity exists with regard to the terminology used in relation to equivalence, comparability and bias. In this book, we have relied primarily on terminology used in cross-cultural psychology and notably that used by Van de Vijver and Leung (1997). They distinguish between equivalence and bias. Equivalence relates to comparability in the level at which different cross-cultural groups are compared. Three levels are identified: construct equivalence, measurement unit equivalence and scalar equivalence. Bias, on the other hand, is a more general term that relates to any factors that may jeopardize the validity of a cross-cultural comparison and is a function of the measurement model and procedures, rather than the nature of the comparison. Bias and equivalence are related, in that the presence of bias will lower equivalence. Absence of bias does not, however, imply equivalence. Consequently, establishment of equivalence and the absence of bias are essential to any meaningful comparison. Considerable progress has been made toward developing procedures to identify and deal with problems arising from measurement and instrument equivalence and use of scales in different cultural environments and settings. Yet, it is clear that much remains to be done in the earlier stages of research in identifying and examining issues related to conceptual and construct equivalence, in ‘decentering’ theories and constructs, and in examining the reliability and validity of constructs and operational measures of these in different contextual settings. 15.1.1. Decentering Theories and Constructs A key issue closely related to that of assessing conceptual equivalence is the ‘decentering’ of theories and constructs, or removing the influence of a dominant culture or philosophy. While this has been extensively discussed in relation to measurement and particularly with regard to translation (Werner and Campbell, 1970), it is also pertinent in relation to many of the theories and constructs used in international marketing research. Often these have been developed and their reliability and validity tested in relation to a single environmental context, frequently the US. Such constructs typically reflect the specific characteristics, values, philosophy and organizational structure of a given economic and sociocultural setting. For example, the two-step flow of influence model, hypothesizing that the impact of mass media is filtered by opinion leaders, was initially developed by Katz and Lazarsfeld (1955) based on research conducted in the US. In examining this theory in Sweden, Cerha (1985) found the flow of influence to be horizontal, across interest groups, rather than vertical, reflecting the flatter societal structure in Sweden. Similarly, research approaches tend to reflect the dominant research paradigm or philosophy of a given culture. Again, this is closely related to the self-referent criterion or bias whereby researchers tend to perceive and interpret stimuli and other phenomena in terms of their own cultural background (Lee, 1966). In some cases, this may result in a lop-sided emphasis on issues and concerns that are of primary interest in the dominant culture but not in others. For example, a study might examine issues relating to consumerism in a culture where a consumer culture has not yet emerged. At worst, it may result in a framing or forced fitting of research questions in terms of those that are relevant in the researcher’s own culture. Decentering of theories and research design typically requires the participation of researchers from different cultural backgrounds and research paradigms. For example, in conducting research in another country or environment, it may be helpful to enlist the collaboration of colleagues in the country studied as well as from other countries or backgrounds. Here, two different approaches can be adopted to design a culturally balanced study, which is not dominated by a single culture (Van de Vijver and Leung, 1997). A decentered approach can be adopted in which researchers from different cultures participate in the design of the study, develop the research instruments, and add culture-specific measures and/or concepts to a common core. Alternatively, a convergence approach can be followed, in which a researcher from each culture designs his or her own instrument. These are then combined and administered in each culture. Similarity of findings across instruments provides strong evidence of validity, while discrepancies may highlight sources of bias. The latter approach is likely to prove time consuming and cumbersome, as well as being difficult to organize and coordinate. 15.1.2. Examining Construct Equivalence A widely recognized and yet persistent issue in international marketing research and in cross-national comparative research in the social sciences is that of ‘construct’ equivalence. As discussed in Chapter 6, this is concerned with whether the underlying construct or concepts studied have the same meaning or salience in another environmental context or setting, and whether, if so, they are best expressed in the same way, for example attitudes, behavior and so on. The issue is not whether the same measurement instrument can be used effectively in another and different context, but rather whether the underlying construct that the instrument is designed to tap has the same function or meaning in a different societal or cultural context. While this issue has been widely discussed in surveys of international marketing research methodology (Cavusgil and Das, 1997; Sekaran, 1983), in practice it is rarely considered in empirical studies. Typically, the researcher focuses on examining the reliability and validity of a specific measurement instrument in another social setting or context, without questioning the appropriateness or relevance of the underlying theoretical construct (Douglas and Craig, 2004). For example, the reliability and validity of a specific innovativeness scale are examined in different countries or sociocultural settings, without investigating whether the concept of ‘innovativeness’ per se is expressed in the same way in each sociocultural context. More attention is needed to conducting emic or culture-specific studies relating to the particular construct or phenomena of interest. This will help to identify relevant construct dimensions and to determine how best to construe the concept in a given context. Conduct of studies specifically assessing the conceptual equivalence of constructs in different settings is, however, extremely rare. 15.1.3. Greater Reliance on Unstructured Approaches Use of an unstructured approach to research design can prove helpful in further probing the contextual embedding of attitudes and behavior and identifying culture-specific concepts or contextual influences. Using this approach, culture and other environmental factors are viewed as providing the context within which other attitudinal and behavioral processes occur or develop. This is in contrast to a structured design, which views culture as an independent variable with a direct influence on the object or behavior studied (Lonner and Adampoulos, 1997). In international marketing research a primary advantage of qualitative methods is that they do not require the imposition of a pre-specified conceptual model or structure to the research. This reduces the likelihood of cultural bias and at the same time increases the probability of identifying new constructs and concepts of relevance to the study. Observational techniques may shed further light on the situational context and its role in shaping processes and behavior. Pro-jective techniques or depth interviews may bring to light new attitudinal dimensions or aspects of behavior, as well as providing deeper understanding and fresh explanations of motivational factors and relationships between variables. These can then be further probed and may provide input to develop and further refine research hypotheses in subsequent phases of research. Technological developments also allow the use of richer stimuli and the collection of data online or through computerized techniques, facilitating data analysis (Pawle and Cooper, 2002). Qualitative data collection techniques can be combined with quantitative techniques to develop a ‘qualiquant’ approach. 15.2. Developing the Research Design Developing a rigorous research design that enables the researcher to focus on examining the phenomena and constructs of interest, while isolating the impact of other confounding influences and eliminating plausible rival hypotheses, is critical (Harkness, Van der Vijver and Mohler, 2003). This is invariably an important issue in research design throughout the social sciences. However, cross-cultural research poses particular problems in this regard due to the multiplicity of levels in the research design and their hierarchical character. In addition, the increased intermingling and flow of people, ideas and information across markets imply that it is increasingly difficult to isolate the impact of diverse influences on the attitudes and behavior of any particular group or set of respondents. This is further compounded by the rapid pace of economic and social change, implying a growing need for attention to tracking changes in the environmental context and monitoring the underlying dynamics of behavioral change. 15.2.1. Extending the Range of Contexts Conduct of studies in a broad and diverse range of sociocultural contexts constitutes another important priority. This is particularly critical where the purpose of the study is to examine how concepts and theories are manifested in different societal contexts; or to identify the impact of societal or contextual factors on specific attitudinal and behavioral phenomena. Systematic examination of both constructs and the relationships among constructs in a broader range of sociocultural settings may help to provide further insights into the nature of these relationships and to broaden understanding of these constructs. Where studies ‘replicate’ a theoretical framework developed or used in a study conducted in a ‘base’ country or context, they may also provide further evidence relating to the universality of this framework. In this case it is important to allow for the addition of emic or culture-specific measures of theoretical constructs, in order to glean further insights into how these are expressed and vary in different cultural contexts. Where the purpose of the study is to examine the impact of the cultural context on attitudinal and behavioral phenomena, the specific types of cultural contexts or elements making up a cultural context need to be clearly identified and categorized. This may include specification of relevant elements of the sociocultural setting at the macro-, meso- and microenvironmental levels, as well as in terms of the specific behavioral, consumption and purchase situations. Insofar as the latter vary from one society or macro context to another, it is often preferable to build up the range of cultural contexts to be studied based on different scenarios or sociocultural situations in structuring the research design. 15.2.2. Establishing Geospatial Boundaries Despite widespread recognition of its limitations, the country remains the dominant geospatial unit used in international marketing research. This stems in large measure from its historical role as the dominant organizational and political unit. The market research information structure is typically organized around country units, and most secondary data, particularly at the macroeconomic level, are collected or available on a country-by-country basis. With the growing integration of markets and flows of communications across national boundaries, the country is no longer the focal unit for the organization of marketing activity. Firms are increasingly targeting global or regional market segments, such as teenagers or affluent consumers, which span national boundaries and are present in markets worldwide. Even where the target market is broader in scope and not segmented, firms are developing and organizing marketing plans relative to geographical regions or world markets as these become more integrated. Equally, sampling frames are becoming available on a world or regional basis, for example lists of world or regional trade organizations, consumer organizations and Internet groups. The changing spatial configuration of markets suggests that research units should be defined on a market-specific basis rather than with reference to geopolitical entities. For example, research should be conducted relative to natural market units or segments with common consumption or behavioral patterns, for example heavy users of cooking oil or small and medium-sized exporters of high-tech goods. 15.2.3. Isolating Confounding Influences A second problem closely related to that of defining the geospatial boundaries of the research design is how to isolate confounding influences on behavior within the unit or units studied. Since international marketing research typically focuses on comparing or understanding differences and similarities between different cultural or social entities in different spatial locations, communication and interaction between units will plague the researcher and contaminate the research results. These problems become particularly acute insofar as respondents are not only exposed to direct and indirect influences from other units, but also move from one unit to another. As Galton noted in his remarks following Tylor’s 1889 presentation of his classic cross-cultural paper, it is typically impossible to obtain cross-cultural sampling units that are independent of each other (Tylor, 1889; Naroll, 1965). Supposedly culturally distinctive traits have often spread between neighboring or historically related regions through historical fusion, diffusion or migration of people. This problem, apparent over 100 years ago, has been further accentuated by increased consumer mobility and the spreading influence of mass media and the Internet. In recent years massive waves of migration (Hispanic and Asian peoples to North America, North African Arabs to France, Russians to Israel and Germany, Turks to Switzerland and Germany) have added further complexity to changing customer patterns. Where once migrants from other cultural backgrounds would gradually become absorbed into the host culture, today a more complex pattern of cultural interpenetration and cultural pluralism is taking place. Migrants are intermarrying and adopting certain facets of the host culture, but also retaining distinct features or traits of their own ethnic or cultural identity. As a result, new cultural entities are emerging and the boundaries of cultural groupings are becoming more intertwined and more fluid. Consumers may belong to multiple groupings, crossing boundaries. In any given situation, the relevant identity of an individual or the dominant social influences may vary depending on the situational context, or the people with whom he or she is interacting. Consequently, isolating the impact of these diverse influences and particularly the nature of their interaction and influence in the formation of preferences and response patterns has become even more problematic. 15.2.4. Extending the Time Dimension The rapid pace of change in consumption patterns in emerging markets, as well as the emergence of new market segments and behavioral and purchasing modes, implies an increasing need to track change in world markets. At the same time, the nature of the market infrastructure, how consumers obtain information as well as systems for the delivery of goods and services are all changing, altering the forces underlying purchase, consumption and disposal behavior. Such trends mean that increased attention is needed to the time dimension of research. In the first place, studies tracking attitudinal and behavioral patterns among specific consumer groups are required. Second, longitudinal studies of how consumption patterns and behavior are changing over time in different situations and contexts will help to shed light on the forces underlying change. Of particular interest here is how behavior and consumption changes as a result of movement to a new geographical location or interaction with a new societal grouping, as also is how newcomers in turn influence the behavior of others. Further complexity is thus added in attempting to distinguish the impact of these structural changes from those of natural evolution or those associated with a specific cohort. 15.3. Improving Analysis of Cross-cultural Data A final issue is the selection of analytical procedures that provide unambiguous means of testing the research hypotheses developed earlier. Here, of primary concern is the development of more rigorous and better calibrated measurement tools. Closely related is the selection of analytical procedures that fit the nature of the design and are capable of capturing and testing the hypothesized patterns of relationships, as well as being free of method bias. Use of multiple methods or procedures, while time consuming, provides greater confidence in the robustness of the research results and enables analysis of the data from different angles and perspectives. 15.3.1. Developing More Rigorous and Better Calibrated Measures Greater emphasis on developing more rigorous and better calibrated measures of constructs is a key concern. This includes not only examining the reliability and validity of existing measures in different societal contexts, and in a range of different environments, but also developing improved versions of instruments and better adapting these to specific research contexts. All too frequently, largely due to the use of ‘borrowed’ constructs and theories, measure reliability and validity are at best only cursorily examined in other research contexts. Often tests, where conducted, are limited to examining internal reliability (for example, Cronbach’s alpha). This in itself is not a particularly stringent test of internal reliability (Rossiter, 2002). Frequently attention is centered on assessing whether or not the measure works; that is, whether there is limited measurement error and adequate respondent comprehension. Rarely is effort devoted to examining whether the instrument in fact measures what it purports to measure and whether measuring the construct in a given culture is at all meaningful. More extensive pre-testing of measures and concepts prior to their inclusion in a study is an essential first step to developing more rigorous measures. Specifically, this might begin with testing alternative formulations of measures on a broader and more diverse subject pool. Often measures, where pre-tested, are administered to small student samples. However, students are unlikely to be representative of a broader spectrum of the population in terms of age, education and other key factors underlying differences in response. In particular, problems relating to comprehension of wording or familiarity with stimuli in the broader population are likely to go undetected. In addition, testing and measure validation typically focus on examining existing measures and constructs. As noted earlier, much research focuses on applying measures and constructs borrowed from prior studies. Relatively little attention has been paid to developing and testing new or different formulations of existing constructs and measures that are better adapted to another cultural context. This is essential if progress is to be made in developing better measures and measurement procedures, and in further understanding cross-cultural concepts. 15.3.2. Triangulation A procedure widely recommended in cross-cultural research in the social sciences to assess method bias is triangulation, or the use of monotrait multiple methods (Marsh and Byrne, 1993). This approach uses multiple diverse methods to examine the same phenomenon or construct. Where responses are consistent across methods, greater confidence can be placed in the results. Low consistency in response, however, suggests the existence of method bias. Wherever feasible, the methods used should be as diverse as possible. For example, results obtained through use of the experimental method could be compared with telephone interviewing. Equally, results obtained using different procedures for recruiting or questioning respondents can be compared. Convergence of results using different methodologies and research approaches provides greater confidence in the validity and reliability of results and in the absence of bias due to use of a particular approach. Triangulation is particularly helpful in situations where other procedures for assessing the reliability of results cannot be applied. For example, examination of results from qualitative approaches can be contrasted with statistical approaches. This approach is also useful where attention is focused on single-item measures or contextually dependent measures, or where replication is not feasible. 15.3.3. Fitting Analytical Models Another related issue concerns the fit of analytical models with the structure of the research design. As noted earlier, the design of cross-cultural research is frequently complex and hierarchical in nature. Attention is centered on examining and comparing relationships between a number of variables in different sociocultural settings and contexts. This implies not only examining differences in the configuration or patterns of relationships among these variables, but also in comparing the impact of the sociocultural context or scenario on these patterns. Consequently, analysis occurs at two levels, at the level of the country or sociocultural setting or scenario, and at the within-country level. An added element of complexity is that some variables are measured at the country level and some at the individual level. To incorporate both types of variables adequately into the same analytical framework requires the use of sophisticated analytical techniques. Application of covariance structure models and confirmatory factor analysis (LISREL and EQS) means that multicountry individual-level data can be analyzed at different levels. Data can be analyzed at the country level to examine the measures and relationships within that country; at the multigroup level to determine whether patterns of measurement and constructs are invariant across countries; or pooled to examine the entire sample. Recent developments in hierarchical analytical techniques provide an appropriate set of techniques to examine the impact of country-level and individual-level variables. Hierarchical linear models can be used to examine variables, such as Hofstede’s measures of national culture, in the same model with individual-level variables. While as yet limited in application, use of hierarchical linear models offers considerable promise in examining the impact of different types of variables as well as the interaction between variables at both levels. As more researchers become familiar with these techniques and apply them to multicountry data, knowledge and understanding will advance dramatically. 15.4. The Growth of Internet Research The increase in Internet penetration in countries around the world has had a substantial impact on international marketing research. Individuals with Internet access can now be reached at relatively low cost no matter where they are located. In numerous cases, in order to ensure adequate response, companies have established large Internet panels such as the Harris panel or American Consumer Opinion, which has over 3.5 million members in the US, Canada, Latin America and Asia. While these panels are not necessarily fully representative of a national population, they are likely to tap innovators, the segment of greatest interest to marketers who are launching new products. In addition, they typically cover a broad spectrum of different interests, activities and product ownership. Internet surveys offer the advantage of enabling the researcher to conduct research rapidly and at relatively low cost. Surveys can be sent out worldwide, responses received and a report delivered within a week, enabling research to be done at a speed that was previously impossible. The cost of conducting an Internet survey is also less than that of a comparable telephone or face-to-face survey. According to one estimate, an Internet survey can be completed, analyzed and presented to the client for approximately 60% of the cost of a traditional survey. Use of an Internet survey also enables the targeting of specialized market segments who could only be reached at much higher costs, if at all, by conventional methods. Richer visual stimuli can also be used on the Internet, for example advertising stimuli or product concepts, in order to provide a more realistic effect. Where respondents have broadband access, the survey can be transmitted in streaming video, closely paralleling the live impact. Respondents can also be sent products and instructed on how to use them, and their reactions can be probed interactively. A subtle problem is that conduct of research on the Internet also tends to reduce the researcher’s exposure to the local research environment. Traditionally, members of the research team would visit a site where research was to be conducted to organize and oversee the research. With the Internet, research is organized centrally, eliminating the need for travel to the site. As a result the researcher may lack understanding and familiarity with the local context and may have a limited basis on which to interpret and understand research results. In brief, the spread of the Internet in many industrialized countries and among certain segments in other countries provides a relatively inexpensive approach to tapping such international populations, as well as a means of obtaining rapid responses. While undoubtedly this medium will grow rapidly in the future with the expansion of Internet access, its use needs to be tempered with a certain caution. There is an inherent danger that researchers will become too enamored of its speedy, low-cost features and fail to look closely at the inherent biases and limitations, as well as at its suitability for collecting specific types of data. 16. Report On Research После того, как исследование проведено и получены результаты, требуется должным образом их оформить. Только тогда выполненная работа не пропадет даром и принесет пользу. Отчет должен быть подготовлен со всей тщательностью35. Небрежно или неграмотно составленным отчетом можно практически свести на нет все усилия, затраченные на проведение маркетинговых исследований. Чтобы этого не случилось, отчет должен удовлетворять ряду требований, которые можно разделить на три группы: требования к структуре, к содержанию и к оформлению. 16.1. Structure of Marketing Research Report Структура отчета должна быть привычной для читателей, даже если авторы отчета считают, что можно сделать и лучше. Как правило, тот, кто будет читать Ваш отчет, просмотрел уже сотню-другую подобных документов. Непривычная структура отчета сразу создаст мнение, что исполнители не знают, как проводить исследования, а полученные ими результаты не заслуживают доверия. Структура отчета должна следовать определенному образцу, принятому для научных работ36. Структура и правила оформления отчета о маркетинговых исследованиях должны соответствовать ГОСТ Р 7.32-91 «Отчет о научно-исследовательской работе. Структура и правила оформления». Ниже лишь добавлены некоторые комментарии по содержанию отчета именно по маркетинговым исследованиям. Отчет должен содержать следующие элементы (в порядке их расположения). 1. Титульный лист. 2. Список исполнителей. 3. Реферат. 4. Содержание. 5. Перечень сокращений, условных обозначений, символов, единиц и терминов. 6. Введение. 7. Основную часть. 8. Заключение. 9. Список использованных источников. 10. Приложения: 10.1. формы всех анкет, таблиц баз данных; 10.2. детализированные вычисления; 10.3. промежуточные результаты, не включенные в тело отчета; 10.4. библиографию37. Титульный лист – достаточно важный элемент структуры отчета с точки зрения оформления. Его несоответствие Стандарту бросается в глаза в первую очередь. В списке исполнителей указываются фамилии и инициалы, должности, ученые степени, ученые звания руководителя исследований, ответственных исполнителей, исполнителей и соисполнителей38. Реферат по значимости занимает особое место. В настоящее время можно найти очень большое количество отчетов и научных работ, причем их число растет в геометрической прогрессии. Поиск материала по нужной теме осуществляется в первую очередь по названиям. Затем, подобрав подходящие названия, исследователь знакомится с рефератами выбранных работ. И только затем он заказывает полный текст работ. Таким образом, реферат – главная визитная карточка работы39. Реферат представляет собой отчет в миниатюре. В нем должны быть отражены следующие моменты: объект исследования; цель исследования; методы и инструменты исследования; полученные результаты и их новизна; рекомендации по внедрению или итоги внедрения результатов; область применения полученных результатов; экономическая эффективность, значимость проделанной работы; прогнозные предположения о развитии объекта исследования. Непросто уместить все это на одной странице40! В содержании перечисляются в порядке появления в отчете разделы, подразделы и пункты с указанием страниц, на которых они начинаются. Перечень сокращений, условных обозначений, символов, единиц и терминов. Перечень сокращений рекомендуется составлять, если количество принятых сокращений более 10. Слева располагают используемые сокращения, справа – их расшифровку. Специальные термины, повторяемые в отчете менее 3 раз, в перечень не вносят, расшифровывая в тексте при первом упоминании41. Общепринятые сокращения (РФ, ГОСТ и т.п.) не расщифровываются. Хотя использование сокращений делает отчет «солиднее»42, оно усложняет работу автора отчета (надо точно следовать принятой системе сокращений, следать, чтобы не было одинаковых сокращений с разным значением) и читателя (все время приходится искать, что значит то или иное сокращение). Поэтому сокращения используются все меньше и меньше. Воздерживайтесь от использования сокращений, кроме общепринятых. Введение пишется с учетом образования и опыта читателей. Смысл введения – подойти к описанию проблемы и показать актуальность исследования. Во введении к отчету указывается место решаемой проблемы в науке и практике, ее связь с другими областями знания. Вкратце приводится современное состояние теоретических и практических методов, которые применимы в данной области. Показывается необходимость проведения данного исследования. Можно также привести основание для проведения работ (реквизиты соответствующего договора). Основная часть (тело отчета) должна содержать сведения о сущности, методике и основных результатах исследований. Она содержит текст, рисунки и таблицы. Основная часть обычно делится на разделы, подразделы, пункты и подпункты. Каждый такой элемент должен содержать логически законченную информацию. Желательно, чтобы каждый элемент структуры содержал в конце выводы. Изложение начинается с постановки задачи исследования. Как правило, этому отводится целый раздел. Как показывает практика, далеко не все понимают отличие постановки задачи от ее решения. В описываемом разделе показывается, чтó требуется сделать. Не говорится, как будет решаться проблема и тем более не приводится никаких решений. Важность данного момента далеко выходит за рамки грамотной постановки задачи маркетингового исследования на основе проблем управления предприятием. Она обусловлена еще и тем, что, прочитав раздел постановки задачи в любом научном труде, читатель поймет, нужно ли ему читать работу дальше. Очень негативное впечатление производят работы, в которых постановка задачи не выделена, и в которых материал не соответствует поставленной задаче. После того, как задача поставлена, дальнейшее изложение может быть более свободным. Можно, например, следовать предложенной в данной книге схеме этапов выполнения исследования. Однако в любом случае следует осветить следующие моменты. Выбор метода решения поставленной задачи. Следует обосновать, в частности, почему было выбрано анкетирование, почему оно должно проводиться именно методом телефонного опроса. Изложение строится от общего к частному. Вначале выбирается тип исследования (поисковое, описательное, исследование причинности). Показывается, почему для решения поставленной задачи лучше выбрать, например, описательное исследование. В конце раздела становится ясно, как будет проведено исследование, почему выбранный способ наилучший. Рассуждения о методах также не столь просты, как может показаться на первый взгляд. Это все еще не решение задачи, а пока лишь обсуждение того, как она будет решаться. Проведение исследований. Подробно описывается, что было сделано, приводятся образцы анкет, излагается способ сбора данных, меры снижения ошибок. Анализ данных. Приводится выбранный для анализа подход (тоже соотнесенный с поставленной задачей), приводится схема и ход анализа. Основные результаты исследований должны сопровождаться оценкой полноты решения поставленных задач и предложениями по дальнейшим направлениям работ, обсуждением достоверности полученных результатов, сравнением их с аналогичными результатами отечественных и зарубежных работ. Должна быть обоснована либо необходимость продолжения работ, либо, при получении отрицательных результатов, целесообразность прекращения исследований. Лучше, если ограничения на применение результатов дадут сами исследователи. Если ограничение будет обнаружено заказчиком, то это снизит доверие к исполнителю. Полезно также обсудить возможные источники неточностей. Выводы. Из полученных результатов делаются обоснованные выводы. Здесь они приводятся без обсуждений. Как правило, это просто суммирование полученных результатов43. Для маркетинговых исследований часто добавляется раздел рекомендации, в котором достаточно подробно и обоснованно описывается, как, по мнению исследователей, следует применить полученные результаты в маркетинговой деятельности фирмы-заказчика. Заключение включает: общий вывод из всей проделанной работы (показывается, что поставленная задача полностью выполнена); оценку полноты решения поставленных задач; рекомендации и исходные данные по конкретному использованию результатов исследования; оценку технико-экономической эффективности внедрения или народнохозяйственную, научную, социальную значимость работы; оценку научно-технического уровня выполнения исследований в сравнении с лучшими достижениями в данной области; возможные направления дальнейших исследований. В приложения рекомендуется включать материалы, связанные с проведенными исследованиями, но не вошедшие в основную часть отчета. Как правило, это материалы, дополняющие отчет; промежуточные математические доказательства, формулы и расчеты; таблицы вспомогательных цифровых данных; протоколы испытаний; описание аппаратуры и приборов, применяемых при измерениях в ходе исследований; заключения экспертов; инструкции, методики, описания алгоритмов и программ, которые были разработаны в процессе исследований; иллюстрации вспомогательного характера; копия технического задания на исследования, программы работ, договора или другого исходного документа, на основании которого проводятся исследования; акты внедрения результатов исследования. Этот раздел будут читать только заинтересованные и компетентные люди. 16.2. Report Contents Главный критерий, которым следует руководствоваться при написании отчета – понятность для читателя. Руководство фирмы будет скорее продолжать жить со своими проблемами, чем примет рекомендованное исследователями решение, которого оно не понимает. Поскольку одного читателя интересует только сводка результатов, другой хочет получить информацию о методах исследования, а третий ждет рекомендаций к действию, следует вначале ознакомиться с требованиями заказчиков отчета. Иногда приходится писать несколько отчетов, иногда – выделять специальный технический раздел. Важными критериями написания отчета являются также полнота, точность, ясность, связность, краткость. Полнота важна для отражения всей информации в одном отчете. Как правило, объем пояснительной информации, представляемой в отчете, должен быть пропорционален объему информации, который будет использован для принятия решения по данному вопросу. Точность подразумевает достоверные исходные данные, правильную процедуру их обработки, логичные обоснования используемых методов, ясную фразеологию. Часто встречаются следующие неточности: общая доля рассматриваемых категорий респондентов менее 100%. Например: «20% опрошенных положительно относятся к товару, а 60% – отрицательно». Что можно сказать об оставшихся 20%? Это не ответившие на вопрос? Это нейтрально относящиеся к товару? В этом случае обязательно требуются пояснения; общая доля рассматриваемых категорий респондентов более 100%. Например, «50% опрошенных указали в качестве причины покупки приемлемую цену, 40% – привлекательный внешний вид, 30% – высокие потребительские качества товара». Здесь необходимо обязательно указать, что респонденты при опросе могли указать несколько причин; неправильное использование круговых диаграмм. При построении круговой диаграммы берутся несколько значений. При этом, часто неявно, предполагается, что все эти значения образуют в сумме некоторую имеющую смысл величину. Если, например, отображать распределение доходов группы респондентов на различные цели, то для круговой диаграммы следует брать не только расходы на продовольственные и непродовольственные товары, но и на все другие статьи расходов. Если интерес представляют только первые две категории, то диаграмма будет иметь три «дольки»: расходы на продовольственные товары; расходы на непродовольственные товары; прочие расходы. Таким образом, вся круговая диаграмма (полный круг) долженсоответствовать общей сумме расходов; • смешивание процентов и процентных пунктов, например: «ранее 6% населения были знакомы с товаром, теперь – 12%». Количество возросло не на 6%, а ◦ на (12%-6%)/6%=1=100% или ◦ в 12%/6%=2 раза или ◦ на 6 процентных пунктов; • грамматические неточности, например: «доходы уменьшились от 1 до 2 тыс. руб.» Здесь, очевидно, подразумевалось: «доходы уменьшились на величину 1…2 тыс. руб.»; • терминологическая путаница, например: «средний годовой доход на душу населения в России увеличился за рассматриваемый период с 2000 руб. до 200 000 руб., следовательно, покупательная способность возросла в 100 раз». А возросшая стоимость жизни? Наиболее трудно добиться ясности. Согласно поговорке, «если читателю дать малейшую возможность неправильно понять текст, то он обязательно так и сделает». Основные рекомендации заключаются в предварительном составлении плана, корректировке текста предложение за предложением, параграф за параграфом, тщательном подборе слов. Вот еще несколько советов. • При подборе слов отдавайте предпочтение более простым и коротким. Лучше сказать продлить, чем пролонгировать, определение – чем дефиниция. • Избегайте клише. Вместо в основном точно честнее сказать с нарушением инструкций, вместо ограниченный успех – небольшой подъем, вместо жизненно важно – можно порекомендовать. • Избегайте технического жаргона. Заменяйте слова типа субоптимальный на не самый лучший. • Пишите просто и естественно. Вместо слова визитация лучше использовать визит. • Удаляйте лишние слова. План гораздо лучше, чем план на будущее, действия – чем предпринятые действия, консенсус – чем консенсус мнений. Это поможет избежать слов-паразитов44 и неграмотных высказываний. Например, часто по радио и по телевизору можно слышать словосочетание самый оптимальный45. • Не пренебрегайте описанием хорошо известных Вам фактов и принципов. То, что очевидно для Вас, то, с чем Вы работали каждый день при выполнении исследований, может оказаться совершенно неизвестным заказчику, а тем более читателям Вашей научной работы. • Старайтесь говорить об одном вопросе в одном месте отчета. Фразы типа полностью эта проблема будет рассмотрена ниже значительно ухудшают впечатление об отчете, да и обо всей проделанной работе. • Изложение должно быть последовательным. Последующие разделы должны опираться на понятия, решения и результаты, описанные в предыдущих. • И наконец, нельзя не упомянуть о ссылках. Если Вы цитируете какого-либо автора или просто повторяете его мысль своими словами, обязательно приводите ссылку на первоисточник. Таких ссылок довольно много и в данной работе. Когда же Вы будете выписывать важные положения из книги, статьи или сайта, не забывайте записывать и ссылку. Свой первый отчет лучше отдать на прочтение более опытным коллегам и быть готовым к его полной переделке. О краткости можно сказать только то, что повторы не компенсируют плохого изложения. Из хорошего отчета нельзя убрать ни одного слова. Очень полезно полностью переписать готовый отчет, сокращая при этом его объем в два раза. Это позволит критически рассмотреть все написанное, избежать повторного изложения одних и тех же мыслей в разных местах. 16.3. Report Design Правильная форма представления материала также играет большую роль в восприятии отчета. Излишне говорить о том, что тщательность его оформления является визитной карточкой исследователя и исследовательской фирмы, которую он представляет. Правила оформления отчета о научно-исследовательской работе регламентируются ГОСТ Р 7.32-91 «Отчет о научно-иссле­до­ва­тель­ской работе. Структура и правила оформления». Выдержки из этих правил даны в приложении. Для упрощения работы над оформлением отчета можно порекомендовать взять за образец уже выполненный и принятый заказчиком отчет. Существуют определенные общепризнанные правила подачи материала, которые делают его значительно более наглядным и позволяют уменьшить объем отчета и одновременно улучшить его понятность без сокращения объема представляемой информации. 16.3.1. Tables Таблицы, как это ясно из предыдущих разделов книги, – важнейшее средство представления данных. Чтобы они стали более понятными, требуется правильно их составлять. Столбцы и строки лучше упорядочить по убыванию средних значений или итогов. Все таблицы со сходным материалом должны иметь одинаковый формат. Цифры для сравнения лучше располагать в столбик, по убыванию. Значения следует округлять до двух значащих цифр, если нет специальных соображений об использовании повышенной точности. В последнем случае приводят три значащие цифры. Например, вместо 1232323,783 руб. следует писать 1,2 млн. руб. 16.3.2. Graphics Правильно составленный рисунок или график иногда может заменить до тысячи слов, однако использовать его следует только тогда, когда он выполняет свою задачу лучше, чем текст или таблица. Графиком или диаграммой можно наглядно представить цифровые данные (сколько). Блок-схема показывает взаимосвязи между объектами или концепциями (как). Карты используются для отображения положения (где). Круговая диаграмма содержит обычно не более 6 элементов («долек»), начинается с направления строго вверх («на 12 часов»), «читается» по часовой стрелке. Элементы должны быть упорядочены по убыванию (кроме последней «дольки», которая показывает менее важные элементы исследования, рассматриваемые совместно46). Обычно последний элемент называется «прочие», «остаток» и т.п. График полезен для отражения изменений во времени. Время откладывается по оси X, значения определенного параметра – по оси Y. Диаграмма в виде столбиков (брусков) служит для наглядного сравнения величин. Диаграммы, показывающие значение одной переменной для ряда объектов, строятся горизонтально: по оси X откладывается, например, цена, а по оси Y – метки, обозначающие вид товара, расположенные по убыванию приводимого значения (рис.Рис. 6). На диаграммах с двумя переменными по оси X можно отложить, например, время, а по Y – доходы. Тогда эта диаграмма будет напоминать график. Рис. 6. Проценты респондентов, назвавших различные причины покупки *** Важная проблема, которая уже упоминалась выше – соответствие названий рисунков, графиков, диаграмм и таблиц той информации, которая в них представлена. 16.4. Report About Research Иногда приходится делать доклад о проведенном исследовании. Здесь действуют те же правила, что и при проведении других докладов (по дипломному проекту, по кандидатской и докторской диссертации). Вот некоторые основные правила. Время доклада не должно быть более 15 минут47. Если докладчик не укладывается в отведенное время, то это свидетельствует не о большом объеме проделанной работы, а о непрофессионализме докладчика. Обычно слушатели слабо знакомы с проблемой, поэтому не следует жалеть время на обоснование актуальности исследования и описание сущности решаемой проблемы. Если описывается решенная частная задача, то ее надо представлять по схеме: важность задачи; необходимые исходные данные; источник их получения; метод решения; результаты; выводы по результатам. Не следует приводить ход решения, особенно если оно было получено стандартными методами. Плакаты, графики, диаграммы, схемы демонстрируются около одной минуты. При меньшем времени слушатели не успеют понять предложенный материал. Не надо читать то, что написано на рисунках (это все уже прочитали). Следует только давать пояснения. В результате доклада у слушателей должно сложиться мнение о важности проблемы, проделанной работе, полученных результатах и выводах из них. Для маркетинговых исследований важны и практические рекомендации, полученные на основе выводов. Порядок проведения доклада также достаточно традиционен. Вначале слово предоставляется докладчику, и он делает основной доклад. Прерывать докладчика не принято. Затем слушатели имеют возможность задать вопросы. Обсуждение обычно занимает гораздо больше времени, чем сам доклад. После этого заслушиваются мнения тех слушателей, которые пожелают его высказать48. В конце обсуждения докладчику предоставляется заключительное слово. 17. Conclusion International marketing research offers tremendous promise as a means to expand knowledge about consumption and purchase behavior and the impact of the market environment and marketing activities on that behavior in other contexts and cultures. It is, however, also evident that for progress to be made, greater attention is needed to rigor in the conceptualization of research programs, research design and application of research tools. While considerable progress has been made with regard to measurement issues, this needs to be matched by equal attention to ‘decentering’ the research approach and broadening theories and constructs to remove the dominance of a single culture or research paradigm. Greater attention therefore needs to be paid to the earlier stages of research design to examine the issues of conceptual and construct equivalence and the relevance of theories in different sociocul-tural contexts. Ideally, researchers from diverse sociocultural backgrounds and research paradigms should participate in the initial phase of the research design. This provides perspective and input relating to the local research context and sociocultural setting, as well as the manifestation of behavioral phenomena within that setting. While posing complex organizational challenges, international marketing research offers considerable promise for deepening understanding of unique emic or cultural-specific phenomena, and also for explaining how universal or etic constructs and theories are manifested in diverse sociocultural settings. At the same time, it helps to provide insights into how cross-border and cultural influences are changing these patterns and their underlying dynamic. This is imperative in an increasingly global and interrelated world. Such research will provide a more comprehensive and less parochial picture of the phenomena studied. It will help managers and academics to reach beyond a nationally embedded perspective and lead to a better understanding of the complex collage of a constantly changing global environment.