Challenges in Conducting International Market Research
- 709 Downloads
This chapter explains the need to conduct international market research, identifies the main challenges researchers face when conducting marketing research in more than one country and provides approaches for addressing these challenges. The chapter examines the research process from the conceptual design of the research model to the choice of countries for data collection, the data collection process itself, and the data analysis and interpretation. Challenges identified include differentiating between etic and emic concepts, assembling an adequate research unit, ensuring data collection equivalence, and reducing ethnocentrism of the research team. We draw on the extant literature to determine methods that address these challenges, such as an adapted etic or linked emic approach, to define the concept of the culti-unit, and to identify prominent approaches to cultural dimensions and collaborative and iterative translation and statistical methods for testing equivalence. This chapter provides researchers with the methods and tools necessary to derive meaningful and sound conclusions from research designed to guide international marketing activities.
KeywordsInternational research Cross-cultural research Emic/etic constructs National indicators National culture Data equivalence Culti-unit Ethnocentrism Back-translation
Multinational companies are increasingly finding major opportunities for expansion outside their home markets. The transformation of planned economies into market economies and increasing demand from emerging middle classes in transition countries present new opportunities for firm growth, and rapid advances in technology facilitate access to these markets. As a consequence, many companies generate a large portion of their sales abroad. For example, the US giant Intel generated more than 80% of its overall sales in 2014 outside the US, BMW generated 81% of its sales outside Germany, and Sony generated 72% of its sales outside Japan. In the first quarter of 2012, Porsche sold more cars in China than in its German home market for the first time. Many start-up companies today are even “born global,” generating substantial sales outside their home nations from their founding or soon after (Knight and Cavusgil 2004).
These developments have important implications for marketing science. Practitioners expect advice about whether marketing knowledge and practices that are successful in their home markets (such as how to facilitate a firm’s market orientation, how consumers make purchasing decisions, and how promotion messages work) work in other nations (Katsikeas et al. 2006), as some highly successful companies (e.g., Disney with Disneyland Paris and Walmart in Germany) have experienced problems and even failures in expanding outside their home markets. These blunders have been traced back to failure to understand the new context and to adapt marketing activities to the host country, among other causes (Ghauri and Cateora 2010).
What, then, must marketing science do so it can provide useful recommendations? Single-country studies may be a first step, but they do not permit sound comparisons of phenomena between countries (Craig and Douglas 2005). Instead, multi-country studies that involve data collection in two or more nations are necessary to identify the generalizable similarities and differences in marketing-related insights between countries that marketing managers can use as guidelines for international marketing decisions.
Multi-country studies also contribute to marketing as an academic discipline. Steenkamp (2005) points out that marketing research has traditionally been driven by US researchers, so its constructs, theories, and relationships implicitly reflect the cultural predispositions of these researchers and the respondents in their empirical studies. Steenkamp claims that marketing science has to get out of the “US silo” and either show the cross-national generalizability of marketing phenomena or identify the contingencies that are related to national characteristics. National characteristics are valuable for marketing science since they allow constructs, theories, and relationships to be tested in diverse settings, similar to those that natural scientists create in their experiments (Burgess and Steenkamp 2006). Two nations, such as the US (a highly developed nation) and Cambodia (a developing nation), can provide extreme situations for testing constructs, theories, and relationships that other frequently used external contingency factors, such a firm’s industry, cannot provide. If constructs, theories, or relationships hold in such diverse conditions as those offered in the US and Cambodia, a high level of generalizability can be assumed (Triandis 1994), and differences can be incorporated into theories and research models to make them more complete.
As a result, increasing numbers of multi-country studies have been published in leading marketing journals like Journal of Marketing (e.g., Petersen et al. 2015) and in journals that are dedicated to international marketing topics, such as Journal of International Marketing and International Marketing Review. While growing in number, multi-country marketing studies are often criticized for how they address the challenges that emerge when data is collected from more than one country (Cadogan 2010; Engelen and Brettel 2011; He et al. 2008; Nakata and Huang 2005). A multi-country research project has much in common with a project that focuses on one country (e.g., in terms of choice between primary and secondary data), but additional challenges render multinational studies more complex. From the conceptual design of the research model to the choice of countries for data collection, the actual data collection process, the data analysis and interpretation, pitfalls must be circumvented, and challenges faced in order to avoid limitations that can render an international marketing study all but useless. When differences in constructs, theories, or relationships between nations emerge, marketing researchers must ensure that they reflect the phenomena of interest and are not artifacts of poor research design and execution. The present article presents an overview of the challenges along the steps of the research process and state-of-the-art approaches to addressing these challenges so international marketing studies provide sound recommendations to practitioners and sound conclusions for marketing science.
Challenges in the Research Process
Conceptual Framework (Phase 1)
A typical starting point of a multi-country study in marketing is the particular constructs, theories, or relationships to be tested. For example, a marketing researcher of Western origin might want to investigate whether the relationship between a firm’s market orientation and the firm’s performance, measured as the firm’s profitability, holds across national contexts. Before collecting data for the research model, the researcher should determine whether the constructs and theories that link them are universal across nations, as the theory that guides the research might not be salient in all research contexts, and even if it is, the constructs might not hold the same meaning in one country as they do in another (Douglas and Craig 2006).
If this challenge is ignored at the beginning of a research process, implications drawn from findings in later phases can be misleading. In our example, while firm profitability might be the primary performance measure in the Western nations, Asian cultures put more emphasis on group-oriented harmony than on achievement and consider employee satisfaction an important outcome of a market orientation, maybe even at the expense of some degree of profitability (Braun and Warner 2002). Leaving out this effect of a firm’s market orientation would lead to an incomplete research model from the Asian perspective.
The degree to which research models, including their constructs, theories, and relationships, allow for country-specific adaptations is captured in the differentiation between an etic and emic approach (Berry 1989). These terms were coined by the linguistic anthropologist Pike (1967), who draws on an analogy with the terms “phonemic” (referring to language-specific sounds) and “phonetic” (referring to sounds that exist in multiple languages). An etic approach in cross-national research assumes that a research model from one national or cultural background can be applied and replicated in other nations and cultures, so it views the elements of the research model as universally valid. The emic view proposes that theories and constructs are specific to the context in which they are applied and are not universal. By using their own domestic situations as frames of reference that are then, without due reflection, applied to and tested in other nations, researchers often implicitly apply an “imposed-etic” approach . While this approach often leads more easily to comparable results (“comparing apples to apples”), the results might be influenced by a pseudo-etic perspective or bias (Triandis 1972) (“apples only relevant in country A, but not in country B”) that leads to misguided or prejudiced findings. Schaffer and Riordan (2003) find that more than 80% of all published multi-country studies implicitly take such an “imposed-etics” approach.
The “adapted etic model ” assumes that the conceptual framework applies to all nations, with some adaptations to local contexts. As a first step, the conceptual framework, its constructs, theories, and relationships are tested in terms of their applicability and relevance to other national contexts. For example, a market orientation may not be relevant in a planned economy. Next, the relevant constructs and hypotheses are checked with support from local researchers. For example, when a researcher is interested in determining whether a particular kind of corporate culture fosters a market orientation across nations (Deshpandé and Farley 2004), it may be necessary to ask local researchers to identify the values (as elements of a corporate culture) that are particularly relevant in their nations. This approach focuses on the similarities among nations, as even when modifications are made, it is likely that the base nation’s perspective dominates, while the unique specifics of other nations may be ignored.
The “linked emic model” addresses this weakness by starting the process of defining a conceptual research model in multiple countries simultaneously, ideally with the support of a host country-based researcher for each setting. As a first step, the researchers from the various nations agree on the scope of the conceptual framework, which serves as input for the subsequent individual work on a conceptual model for each researcher in his or her national setting. Next, the researchers identify similarities among the locally developed models and factors at the national level that explain differences. Ideally, an overarching conceptual model is derived that covers all identified elements and differentiates between emic and etic elements. Nation-specific factors can be integrated into the model as contingencies to capture the nations’ emic particularities. Assuming a researcher is interested in understanding what drives a firm’s market orientation, collaboration among departments may be more important in collectivistic cultures than in individualistic cultures. This process puts a strong emphasis on the local perspective and is facilitated by effective cooperation among researchers from the nations in which the research takes place.
The efforts required in developing the adapted etic or linked emic model are targeted toward ensuring construct equivalence, a major facet of data collection equivalence that refers to whether constructs, theories, and relationships have the same purpose and meaning in all of the nations under investigation. Construct equivalence has three facets (Bensaou et al. 1999): functional equivalence, conceptual equivalence, and category equivalence. Functional equivalence refers to the degree to which phenomena or objects have the same function across nations. For example, a car provides family transportation in a highly developed country while a motor bike performs the same function in an emerging economy. Conceptual equivalence relates to the degree to which the phenomena are interpreted similarly across nations. For example, in a study of the antecedents of market orientation, local interpretations of the “amount of market-related information transferred between departments” can be tested in exploratory fieldwork. Category equivalence captures the degree to which the same classification scheme can be applied to a phenomenon across nations. For example, a particular market orientation’s primary stakeholder group can be customers in one nation and government institutions in another.
Research Units and Drivers of Differences (Phase 2)
Once the conceptual framework has been established, the next step is to identify a research unit and the drivers of the differences between research units. We investigate the concept of the unit of research (section “Definition of the Unit of Analysis”) and discuss potential drivers of differences between units of research (section “Identifying Drivers of Differences Between Nations”).
Definition of the Unit of Analysis
In cross-national research, a unit of analysis must be established that defines the geographic scope and the group of the people or organizations to be examined within it. A good research unit has a high degree of homogeneity in terms of the members’ behaviors and values among the members of this group, which are heterogeneous to other groups and is as free of influence by other groups as possible (Craig and Douglas 2006; Lytle et al. 1995).
Literature reviews on international marketing research indicate (e.g., Engelen and Brettel 2011; Nakata and Huang 2005) that nations are the primary unit of research. Some of the reasons for a focus on nation as the unit are practical and pragmatic. Nations have clear and defined boundaries, and sampling frames are available at the nation level or for defined geographic areas in countries, such as regions and cities. In addition, multinational firms are often organized on a country basis and are interested in formulating strategies in multiple countries, for which they must assess the similarities and differences among countries. This focus carries through to academic researchers’ interest. For example, empirical cross-national marketing research often focuses on comparing phenomena of interest between the Western nations – often the US – and Asian countries – often China (Sheng et al. 2011). However, whether national borders are the best criterion with which to define the unit of research in international marketing may be questioned.
Nations and cultures are increasingly influenced by other nations and cultures. A primary mechanism is the global flows identified by Appadurai (1990). Five primary global flows blur the borders between nations: mediascapes (flow of images and communication, such as by screening US-made films in other countries), ethnoscapes (flows of tourism, student exchanges, and migrants), ideoscapes (flows of political impacts and ideologies, such as democratization and views of equality), technoscapes (flows of technology and expertise), and finanscapes (flows of capital and money). These global flows, which will only grow because of the increasing ease and decreasing cost of data transfers and travel, lead to multicultural marketplaces (Demangeot et al. 2015), so the view of cultures as localized and determined only by national boundaries has lost much of its validity. These flows can cause changes in a country’s cultural elements through cultural contamination (adopting foreign cultural elements), pluralism (individuals in one culture exhibiting features of multiple cultures), and hybridization (fusion of two cultural elements into one new) (Craig and Douglas 2006). For example, the US culture is presented to customers in most other countries via products that are seen as typical of the American lifestyle (e.g., McDonalds, Levis, Marlboro), and it exerts cultural influence via a globally dominant movie industry. Consequently, at least part of many countries’ populations adopt these cultural elements and values, leading to the “Americanization” of other nations’ cultures.
Further, many nations and their cultures are not homogeneous units but contain subgroups (Cheung and Chow 1999) that may be driven by migration, ethnic heritage (e.g., Chinese in Malaysia, the Dutch heritage of South African immigrants), religious beliefs (e.g., US Jews, Chinese Catholics), or the nation’s size (e.g., Russia, China, India). For example, studies find cultural subgroups in Singapore (Chen 2008) and several South American nations (Lenartowicz and Johnson 2002). The frequent presence of such subgroups poses a challenge to international marketing research that seeks specificity in the differences between nations, as findings will depend on the subgroup sampled in such heterogeneous nations.
Given the influences that nations exert on each other and the heterogeneity in most nations, it follows that researchers must examine the homogeneity of the nations or regions they want to analyze. While some countries, such as Belgium, India, and Switzerland, are inherently heterogeneous in terms of behaviors, attitudes, and/or language, some studies provide concrete empirical evidence of this heterogeneity. Based on World Values Survey data, Minkov and Hofstede (2012) identify 299 in-country regions in terms of cultural values in twenty-eight countries that cluster on national borders, even for potentially heterogeneous countries like Malaysia and Indonesia. Intermixtures across borders are rare, even in cases of culturally similar African neighbors, such as Ghana and Burkina Faso. While this study validates empirically that nations can be good approximations for cultures, other studies that focus on a single nation find that there are substantial cultural differences within one nation. For example, Cheung and Chow (1999) use empirical research to find nine distinct subcultures in China, and Dheer et al. (2015) identify nine distinct subcultural regions in India and provide guidance on how these subcultures differ. (For example, the far-eastern and southwestern regions are lower in male dominance than the other parts of India.)
A culti-unit is […] defined in terms of the racial, ethnic, demographic or social-economic characteristics or specific interests (e.g., ecologically concerned consumers) of its members which provide a common bond and establish a common ethnie, a core of shared memories, myths, values and symbols woven together and sustained in popular consciousness.
A commonly shared ethnie distinguishes the members of one culti-unit from others. This ethnie can be a national culture, a shared religion (e.g., Jewish heritage), or a strong interest (e.g., the hacker community of Anonymous). A major merit of the culti-unit concept is that it incorporates the concept of nation when the nation is sufficiently homogeneous, but it can also be applied to other ethnies. Researchers can benefit from the culti-unit construct since it makes ruling out alternative explanations for a theory or relationship easier than does a broader concept like nations. The ethnie core can be revealed by means of qualitative research. By taking the culti-unit as a starting point in defining the unit of research, researchers are forced to define their units of research unit cautiously and not to use national borders without careful reflection.
Sometimes the country or larger region is the appropriate sampling frame and serves as the culti-unit, particularly when culture has relatively little influence on the product or topic being researched. For example, compared to food and clothing, automobiles and consumer electronics do not have a strong cultural component. However, whenever there is likely to be considerable within-country heterogeneity or when the researcher is interested in understanding culture’s influence on a particular outcome, the researcher should either sample from the culti-unit or be able to identify the various cultural or ethnic groups and conduct analysis to determine their affect. For example, Petruzzellis and Craig (2016) examine the concept of Mediterranean identity across three European countries (Spain, France, and Italy) and find elements of an ethnie core related to Mediterranean identity that transcends national borders.
Research on subcultures within a larger culture illustrates the importance of a culti-unit. For example, Vida et al. (2007) conduct research in Bosnia and Herzegovina, where there are three major cultural/ethic groups: Croats, Serbs, and Bosnians. The research analyzes responses by cultural group and finds that ethnic identity influences the dependent variables. Studies that examine a particular subculture face challenges in obtaining a sampling frame, but they can view a homogeneous group of respondents. Within-country differences can also be examined geographically. Lenartowicz et al. (2003) find significant within-country and between-county differences among managers on the Rokeach Value Survey, suggesting that using the country as the unit of analysis would mask important within-country variations.
Ultimately, the selection of the unit of analysis will be a function of practical considerations and the research’s theoretical underpinnings. If a particular theory is being tested in multiple countries, the respondents in each country must reflect the ethnie core of the culture of interest. Ideally, the research would be able to locate appropriate sampling frames to focus on the specific groups, but if such sampling frames are not available, questions should be included that allow for a fine-grained examination so the entire sample can be analyzed and then broken down by specific cultural/ethnic groups. If there are significant differences between groups, the one(s) most relevant to the research can be examined more closely. A related concern is determining what factors account for the observed differences, whether they are contextual factors like the level of economic development or the influence of other cultures. The use of covariates in the analysis will often help in identifying which factors affect the culture.
Identifying Drivers of Differences Between Nations
When several nations are compared,1 one of the key questions that arises concerns the underlying drivers between nations that account for the difference and that may even be generalized to explain variations from other nations. Assuming that nations are appropriate units of research for a particular purpose and we find differences (e.g., in the strength of the relationship between market orientation and firm performance between the US and Indonesia). An explanation for these differences can lie in the differing degrees of cultural individualism versus collectivism (Hofstede 2001), but the US and Indonesia also have differences in their economic (e.g., GPD per capita) and development levels (e.g., Human Development Indicator or HDI) , which may be the key drivers of the observed differences.
In their review of empirical cross-national and cross-cultural research , Tsui et al. (2007) find that national culture – typically defined as the values and norms that guide a group’s behavior (Adler 2002) – is the most frequently investigated driver of differences between nations. National culture can be conceptualized along national cultural dimensions that relate to how societies resolve the problems that all societies face (e.g., whether the individual person is more important than group equality and harmony and how much privacy is granted to individuals). Various schemes of cultural dimensions have been proposed, but the four original dimensions from Hofstede (2001) – power distance, individualism versus collectivism, uncertainty avoidance, and masculinity versus femininity – are the most prominent. Later, Hofstede and colleagues added the dimensions of long-term orientation and indulgence versus restraint. The latter, originally proposed by Minkov (2007), has been identified by means of World Value Survey items (Hofstede et al. 2010). Societies that are strong on indulgence allow free gratification of natural human desires, while societies that are strong on restraint prefer strict norms that regulate such gratification.
International marketing research focuses on Hofstede’s dimensions, as the literature review from Engelen and Brettel (2011) indicates. One might argue that the country scores that Hofstede initially developed at the end of the 1960s/beginning of the 1970s are outdated, but Beugelsdijk et al. (2015) show that cultural change is absolute, rather than relative. By replicating Hofstede’s dimensions for two birth cohorts using data from the World Values Survey, Beugelsdijk et al. (2015) find that most countries today score higher on individualism and indulgence and lower on power distance compared to Hofstede’s older data, but cultural differences between country pairs are generally stable. Further, to circumvent the threat of using outdated country data, international marketing researchers can apply the updates provided on Hofstede’s website (http://geert-hofstede.com/). These updated data have been used in some recent cross-national marketing studies, such as Samaha et al. (2014). Other studies, such as the meta-analytical review from Taras et al. (2012), also provide updated country scores for Hofstede’s dimensions.
Several authors criticize Hofstede’s approach in terms of its theoretical foundation and the limited number of cultural dimensions (Sondergaard 1994). Schwartz (1994) and the GLOBE study address some of these criticisms. Siew Imm et al. (2007) find that the cultural dimensions from Schwartz (1994) are broader than those from Hofstede (2001), as Schwartz (1994) covers all of Hofstede’s dimensions and adds the dimensions of egalitarianism and hierarchy. Steenkamp (2001) also highlights Schwartz’ (1994) theoretical foundations, concluding that “given its strong theoretical foundations, [Schwartz’s approach] offers great potential for international marketing research” (p. 33).
Javidan et al. (2006) point out that the GLOBE study adopts a theory-based procedure and formulate a priori dimensions based on Hofstede (2001) dimensions, values that Kluckhohn (1951) and McClelland (1961) described, and the interpersonal communication literature (Sarros and Woodman 1993). In addition to Hofstede’s cultural dimensions of power distance and uncertainty avoidance, the GLOBE study adds performance orientation, assertiveness, future orientation, human orientation, institutional collectivism, in-group collectivism, and gender egalitarianism (House et al. 2001). Some of these novel dimensions are more fine-grained than are Hofstede’s (2001) dimensions. For example, the dimensions of assertiveness and gender egalitarianism reflect two major facets of Hofstede’s masculinity dimension (Hartog 2004). Cross-cultural marketing studies often neglect or even ignore the potential offered by Schwartz (1994) and the GLOBE study. International marketing researchers should be sure to justify their choices of national cultural dimensions as the most appropriate for their purposes.
A marketing researcher who needs to choose one approach should consider the following thoughts: Hofstede’s and GLOBE’s dimensions and country scores have been derived theoretically and/or empirically in the workplace setting, so organizational marketing topics might rather build on their dimensions. Schwartz’ dimensions have their theoretical origin in psychological research on individual values and have been empirically analyzed by Schwartz in a cross-national sample of teachers and students. Therefore, these dimensions are rather appropriate when investigating the decisions of private persons across cultures (such as in international consumer studies).
Further, the targeted nations in an international marketing study can lead to the use of the one or other approach. Schwartz generated data in some regions which have not been covered to the same extent in Hofstede’s and the GLOBE survey (e.g., some former Eastern European bloc countries and some countries in the Middle East). There are also some countries (e.g., some African countries) which have been covered by GLOBE and not the other approaches. So, the individually targeted countries in a research project may determine the choice of dimensions.
In their literature review on cross-national and cross-cultural research, Tsui et al. (2007) conclude that extant research has focused too much on national cultural dimensions while neglecting other drivers of the differences between nations. As a result, the findings of multination studies that focus only on national cultural dimensions may be misleading. Tsui et al. (2007) and researchers like Glinow et al. (2004) call for a polycontextualization of international research in order to accommodate the complexity of the context and avoid misleading conclusions about what drives the differences between nations. Beyond national culture, the physical context (e.g., climate, typology), the historic context (e.g., sovereignty, colonization), the political context (e.g., the political and legal systems), the social context (e.g., religion, family structure), and the economic context (e.g., economic system, technology) may be the reason for differences (Saeed et al. 2014). Sound international marketing research must not neglect these contextual drivers (see Douglas and Craig (2011) for a discussion of the role of contextual factors).
International Data Collection (Phase 3)
Extant international marketing research is often built on data from only two nations (Cadogan 2010; Engelen and Brettel 2011). However, this approach has serious limitations, particularly since countries typically differ in terms of more than one cultural dimension, as well as in such contextual areas as the macroeconomic development stage or the educational system (Geyskens et al. 2006). As a positive example, Steenkamp et al. (1999) draw on responses from more than 8000 consumers in 23 countries to isolate the effects of the regulatory system, the moral system (national identity), and the cultural system (degree of individualism) on the perceived value of websites. These effects could not have been separated on the national level with a two-country comparison.
To address what is a meaningful number of nations in which data should be collected and how these nations should be chosen, Sivakumar and Nakata (2001) develop an approach to guide researchers in defining the number of nations for data collection. In their approach, when cultural differences are expected to be due to one cultural dimension (e.g., the degree of power distance), two nations that have strong differences in terms of this dimension and few differences in the other cultural dimensions should be chosen, and when differences are expected to be due to two cultural dimensions, four national cultures should be used to represent all four combinations of the two levels (high and low) for each cultural dimension, while the four national cultures are similar in terms of the remaining cultural dimensions.
While this approach can help researchers determine the appropriate number of nations for identifying the role of cultural dimensions, the procedure does not provide guidance on how to deal with rival and confounding drivers at the national level, such as the stage of macroeconomic development (Ralston et al. 1997). In order to exclude rival explanations for differences between nations, even more nations should be included. For example, Tan (2002) creates a hybrid, quasi-experimental design to determine whether national cultural or contextual effects prevail by drawing on three samples from two subcultures and two countries: mainland Chinese, Chinese Americans, and Caucasian Americans.
In order to identify the roles that national cultural or contextual factors at the national level play, the number of nations for data collection must be extended, as long as identified differences can be traced back to either one of the national cultural or contextual factors, while controlling for alternative explanations at the national level.
Once the national settings are defined but before the data collection begins, three equivalence challenges must be addressed in order to generate sound empirical findings: translation equivalence, sampling frame equivalence, and data collection procedure equivalence (Hult et al. 2008). Collecting data in several countries in which more than one language is involved requires ensuring translation equivalence. Simple back-translation is the dominant approach in international marketing studies, where a questionnaire in the researcher’s native language is translated into another language by a bilingual person (Brislin 1980). This translated questionnaire is then back-translated to English by another bilingual person. Only when the researcher compares the original and the back-translated questionnaire and finds no relevant differences can translation equivalence be assumed. While this approach is the most widely applied in international marketing literature, it has some limitations, as it does not necessarily ensure equivalence in meaning in each language (Douglas and Craig 2007). Referring to the “ emic versus etic” debate, assuming that a simple translation from the base language that does not take the particularities of the other language into account (e.g., words or idioms that exist in only one language) is inherently etic or even “imposed-etic.”
The process starts with the translation, where a questionnaire in one language is translated independently to all target languages by at least two translators. Translators should especially pay attention to items that deal with attitudes since linguistic research indicates that the connotations of words like “happiness” and “mourning” can differ from language to language. The translation of mix-worded multi-item measures – that is, measures that contain positive-worded statements and reverse-worded statements – is a major challenge since empirical studies have found problems with the internal consistency and dimensionality of these measures, which are mostly of US origin, when applied cross-nationally. Wong et al. (2003) identify two reasons for these problems: how languages indicate negation differ such that reverse-worded statements may be difficult or even impossible to translate appropriately, and respondents’ cultural predeterminations affect how they respond to reverse-worded statements. When the dominant norm is to be polite and agreeable, as is the case in some Asian cultures (Child and Warner 2003), respondents may tend to agree with any statement, leading to low internal consistency and disruption of the dimensionality of mixed-worded measures. Therefore, Wong et al. (2003) suggest employing only positively worded statements cross-nationally or replacing Likert statements with questions.
The second step of Douglas and Craig’s (2007) approach is a review meeting with the translators and an independent researcher in order to agree on a preliminary version of the questionnaire in each language. In the third step, adjudication, inconsistencies are resolved and whether the questionnaire actually measures the same meaning in each country is determined. In the fourth step, the questionnaire is pretested with native respondents in each language to ensure comprehension, and issues are referred to the team of researchers and translators to start a new round of iterations. Finally, in the fifth step, if more than one round of data collection is planned, such as may be the case with longitudinal data collected yearly, insights gained during the initial data collection are reported to the team of researchers and translators so they can improve the questionnaires for successive rounds of data collection (without compromising year-to-year comparability).
Sampling frame equivalence refers to the extent to which the samples drawn from the various nations parallel one another (Hult et al. 2008). For example, if high school students from the middle class in India are compared to public high school students from all social classes in the US, the discrepancy in social class could distort the findings. Researchers must select equivalent samples among the various research units while allowing variations in the sample on the factors to be analyzed. For example, individuals of similar income, education level, and gender (equality of sample) are selected from nations whose cultural values (variation to be researched) differ. Hult et al. (2008) consider sample equivalence a major prerequisite for sound cross-cultural comparisons and recommend that organization-level studies match samples in terms of potentially confounding factors like company age, size, and industry sector. Although sample equivalence does not guarantee that findings can be generalized to the participating countries, it helps to ensure that comparisons are not confounded by other factors and that differences can be traced back to the cultural dimensions or contexts under study (van Vijver and Leung 1997).
Data collection procedure equivalence combines administrative equivalence (e.g., telephone, face-to-face, email) and time equivalence in terms of the time between data collection in the participating countries (Hult et al. 2008). While a completely standardized and parallel data collection procedure in all units of research is ideal, regulations (e.g., national rules and regulations against telephone interviews), cultural norms (e.g., nonconformity of impersonal surveys with the cultural preferences related to personal interactions), and infrastructure (e.g., availability of high-speed internet lines) often prevent data collection procedures from being perfectly equal (Hult et al. 2008). Even so, researchers must seek equivalent data collection procedures and keep unavoidable differences (e.g., different survey settings and different times) in mind as a possible explanation for findings.
Finally, in the international data collection phase, researchers must decide on which level to measure cultural properties. Cultural properties can either be directly measured in the surveyed sample at the individual level of each respondent (“direct value inference”) or captured by means of typically nation-related secondary data (e.g., the Hofstede country scores) according to the surveyed individuals’ or firms’ national identity (“indirect value inference”). Direct value inference measures individuals’ cultural properties and derives cultural properties for data analyses by aggregating these cultural properties to the relevant group level (e.g., the national level). Researchers can directly measure the surveyed individuals’ individual values or ask about their perceptions of their environment’s cultural values. This approach ensures that the actual culture of the surveyed individuals is measured. However, the questions concerning whether the surveyed individual can assess the cultural properties correctly and whether there are any bias in the assessments remain. Indirect value inference assigns existing values on cultural dimensions (e.g., from the Hofstede data) to surveyed individuals according to their group membership – that is, surveyed individuals from Germany receive the score for Germany on the relevant cultural dimension as reported by Hofstede or other researchers. In this case, however, a measurement error might occur since the researcher assumes that the surveyed sample’s cultural properties comply with the cultural properties of the samples used in the earlier studies that report country scores (Soares et al. 2007). Given the benefits and perils of both approaches, Soares et al. (2007) recommend a multi-method approach that combines the indirect and direct value inference approaches.
Whether direct or indirect value inference is pursued, Brewer and Venaik (2014) recommend that researchers that are determining the right level of analysis ensure a fit between the levels at which the constructs (e.g., cultural properties) are theorized and empirically validated. Conceptualization of theories, measurement of constructs, and data collection should be conducted consistent with the underlying research question. Brewer and Venaik (2014) refer to the danger of an ecological fallacy when researchers assume that group-level relationships automatically apply to the individual level. Some recent studies use cultural dimensions explicitly at the individual level (e.g., individual power orientation) to make clear that cultural properties at the level of the surveyed individual are theorized and measured (e.g., Auh et al. 2015; Kirkman et al. 2009). When only group-level data on cultural properties is available, Brewer and Venaik (2014) recommend that higher-level constructs (e.g., cultural dimensions at the national level) be cross validated with measures at the level at which the construct is theorized (e.g., at the individual level of a consumer).
Data Analysis (Phase 4)
Data analysis starts with analyses which are not specific to international marketing research but have to be done in any data analysis. Such checks and tests include establishing the reliability and validity of the measures and ruling out biases in the survey data like common-method bias, nonresponse bias, and informant bias (Bagozzi et al. 1991; Podsakoff et al. 2003). However, the particularities of international marketing research impose additional challenges related to the data analysis based on how differences in national response styles can affect findings. A response style refers to a person’s tendency to respond systematically to questionnaire items on some basis other than what the items are designed to measure (Jong et al. 2008). For example, in the US school grading systems range from A+ (best) to F (worst) or points up to 100 (best) and down to 0 (worst), while the German grading system is the other way around, ranging from 1 (best) to 5 or 6 (worst). Therefore, Germans who are used to “lower is better” might unwittingly answer incorrectly on US surveys that range from 1 (worst) to 5 (best). While these scale definition issues might be resolved easily, national cultural predeterminations based on deeply rooted values and preferences may have more subtle effects on a participant’s response style (Clarke III 2001).
Two major response styles have been shown to be subject to the respondent’s national culture: An extreme response style (ERS) is the tendency to favor the end points of rating scales, regardless of the item’s content, while an acquiescence response style (ARS) refers to the tendency to agree with all items, regardless of the item’s content. Chen (2008) reports that US respondents are much more inclined to show an ERS than are respondents from China. In cultural terms, respondents from low-ERS cultures may wish to appear modest and nonjudgmental, whereas members of high-ERS cultures may prefer to demonstrate sincerity and conviction. Regarding ARS, Riordan and Vandenberg (1994) report that a response of 3 on a 5-point Likert-type scale means “no opinion” to American respondents but “mild agreement” to Korean respondents, so a Korean’s “3” may be equivalent to an American’s “4,” and a Korean’s “4” may be equivalent to an American’s “5.” A strong response bias is problematic because whether differences are caused by differences in response styles or differences in the factor of interest remains uncertain. If relationships are compared, differences in response styles lead to variances in the dependent and independent variables that result in unintended and confounding differences in correlations.
Differences in response styles belong to a larger group of issues when measurement models are applied in more than one nation. A major threat to sound multination marketing research occurs when respondents do not interpret the constructs that link relevant relationships or that build theoretical frameworks similarly such that identified differences in relationships are actually due to systematically different interpretations of constructs and measurement models (Mullen 1995). While some countermeasures can be taken in the pre-data collection phase, such as ascertaining translation equivalence, quantitative tests on the actual data collected are necessary.
In particular, measurement equivalence , which relates to whether measurement models are understood similarly across nations, must be established. Steenkamp and Baumgartner (1998) provide a multigroup confirmatory factor analysis approach for reflective multi-item measurement models, which approach consists of configural, metric, and scalar equivalence analyses (for applications, see, e.g., Homburg et al. (2009) and Zhou et al. (2002)). Configural equivalence indicates that respondents from all nations under analysis conceptualize and understand the basic structure of measurement models similarly. Metric equivalence indicates that groups of respondents understand scale intervals similarly. Scalar equivalence indicates that the systematic response style among respondents from the nations under study does not differ.
Configural equivalence , which is tested by running a multigroup confirmatory factor analysis that allows all factor loadings to be free across the national samples, is given when the factor loadings are significant in all samples and the model’s fit is satisfactory. Partial metric equivalence is given when at least one item (in addition to a marker item) for each measurement model has equivalent factor loadings across nations. Metric equivalence models must be specified with at least two factor loadings per measurement model that are kept equal across nations while not constraining the remaining factor loadings. Full metric equivalence is given when all factor loadings are equal across groups, although Steenkamp and Baumgartner (1998) indicate that partial metric equivalence is sufficient in most cases. By means of a χ2-difference test, this metric equivalence model is compared with a model in which all factor loadings are free across samples, and metric equivalence is confirmed when the two models do not differ significantly. Finally, scalar equivalence, which is tested by comparing means, ensures that differences in observed and latent means between national samples are comparable. The procedure for testing scalar equivalence is the same as the χ2-difference test for metric equivalence except that item intercepts are constrained across national samples. Steenkamp and Baumgartner (1998) point out that, in most cross-national comparisons, only partial scalar invariance is realistic.
Since establishing measurement equivalence across a high number of countries would require extremely large sample sizes, some studies have created set of countries with similar cultural and economic conditions between which measurement equivalence is established (e.g., Hohenberg and Homburg 2016; Tellis et al. 2009). For example, Hohenberg and Homburg (2016) cluster their 38 surveyed countries into four categories, differentiating among English-speaking countries, European countries, Asian countries, and Latin American countries.
Once measurement equivalence is established, the relationships of interest can be empirically investigated. When national constructs are integrated as moderators in theoretical frameworks, group comparisons (e.g., in structural equation modeling) and interaction term models (e.g., in regression models) can be applied (Engelen and Brettel 2011), although some challenges specific to international marketing projects must be considered. Most multinational studies investigate a particular relationship (e.g., the effect of top management’s attention on a firm’s market orientation) in multiple nations, for which the researchers have a national dependency in mind, such as a particular national cultural dimension (e.g., the degree of national cultural power distance). However, to accommodate the multiplicity of possible drivers at the national level (section “Identifying Drivers of Differences Between Nations”), researchers should add controls for alternative explanations in their models, an approach that is particularly feasible in regression models (Becker 2005). For example, by integrating a broad set of national cultural dimensions into their model, Samaha et al. (2014) show that national culture has a more multifaceted role in relationship marketing than earlier studies that focus on just one national cultural dimension suggest. Adding several cultural dimensions into a regression model is likely to lead to multicollinearity since the cultural dimensions are often correlated. Individualism versus collectivism and power distance are often strongly correlated. Samaha et al. (2014) circumvent this problem by not adding these two dimensions simultaneously in their regression models, leaving out the individualism versus collectivism dimension in their power distance model and leaving power distance out of all other models.
Interpretation (Phase 5)
While interpretation is an important element in all five steps of the research process – before, during, and after data collection – it manifests particularly at the end of the research process, when the actual findings are available. Of course, interpretation is by no means a particularity of international marketing research projects, but one particular challenge emerges with these kinds of studies. A major assumption of cross-national research projects is that drivers at the national level can lead to differences in marketing relationships, constructs, and theories, and since national drivers, especially national culture, affect everyone living in a nation or culture (Hofstede 2001), the researcher himself or herself is also subject to national or cultural predetermination. Thus, the researcher’s cultural values can affect his or her interpretation of the findings (Berry 1980). This bias, called ethnocentrism , occurs when one person’s or group’s frame of reference is applied in interpreting other groups’ responses without adaptation to other national cultures. If, in our example, we find that the attention of top management to market-related issues drives a firm’s market orientation more strongly in Asia than in Western nations, coming from a Western perspective, we could easily assume that power distance, which is particularly strong in Asian cultures, is the driving force. However, Asian researchers might relate this finding to particularities of the Confucian teachings. To exclude such an ethnocentric bias, researchers in international studies should build and use cross-national research teams during the entire research process, but especially in the last step of interpreting the findings (Hofstede and Bond 1988), and document nation- or culture-specific interpretations of findings.
As firms from developed and developing economies continue to expand outside their home markets, marketing research is essential to guide development and execution of marketing strategy. An implicit challenge is for management to appreciate that there are potentially a wide range of differences between their home market and the foreign markets they currently operate in or are planning to enter. Well-constructed research will not only identify differences but also reveal important similarities. Regardless of the specific purpose of the research, it is essential that valid and reliable research be designed and executed. This is the critical challenge and applies whether the research is to guide management decisions or test the applicability of theories and constructs across multiple countries.
However, international marketing research is more complex and time consuming than single country research. Advances in technology, particularly ready access to internet samples in multiple countries has greatly facilitated rapid collection of multi-country data. However, unless careful attention is paid to the design and execution of the research to achieve equivalence on all dimensions across the units studies, the results may be misleading or meaningless. Careful attention to all the steps outlined in this chapter is essential to ensure that the results of international marketing research are reliable and valid and can be used to make meaningful inferences and advance the state of our knowledge about markets outside our own.
For the sake of simplicity, we will subsequently refer to nations as the unit of research, acknowledging that other culti-units may be more appropriate as outlined in section “Conceptual Framework (Phase 1)”.
- Adler, N. (2002). International dimensions of organizational behavior. Cincinnati: South-Western College Publishing.Google Scholar
- Auh, S., Menguc, B., Spyropoulou, S., Wang, F. (2015). Service employee burnout and engagement: The moderating role of power distance orientation. Journal of the Academy of Marketing Science, 1–20. 10.1007/s11747-015-0463-4.Google Scholar
- Berry, J. (1980). Introduction to methodology. In H. Triandis & J. Berry (Eds.), Handbook of cross-cultural psychology (pp. 1–28). Boston: Allyn & Bacon.Google Scholar
- Braun, W., & Warner, M. (2002). The “Culture-Free” versus “Culture-Specific” management debate. In M. Warner & P. Joynt (Eds.), Managing across cultures: Issues and perspectives (pp. 13–25). London: Thomson Learning.Google Scholar
- Brislin, R. (1980). Translation and content analysis of oral and written materials. In H. Triandis & J. Berry (Eds.), Handbook of cross-cultural psychology (pp. 389–444). Boston: Allyn and Bacon.Google Scholar
- Child, J., & Warner, M. (2003). Culture and management in China. In M. Warner (Ed.), Culture and management in Asia (pp. 24–47). London: Routledge.Google Scholar
- Craig, C. S., & Douglas, S. P. (2005). International marketing research (3rd ed.). Chichester: Wiley.Google Scholar
- Featherstone, M. (1990). Global culture: Nationalism, globalism and modernism. London: Sage.Google Scholar
- Ghauri, P., & Cateora, P. (2010). International marketing (3rd ed.). New York: McGraw-Hill.Google Scholar
- von Glinow, M. A., Shapiro, D. L., & Brett, J. M. (2004). Can we talk, and should we? Managing emotional conflict in multicultural teams. Academy of Management Review, 29(4), 578–592.Google Scholar
- Hartog, D. (2004). Assertiveness. In R. House, P. Hanges, M. Javidan, P. Dorfman, & V. Gupta (Eds.), Culture, leadership, and organizations: The GLOBE study of 62 societies (pp. 395–436). Thousand Oaks: Sage.Google Scholar
- Hofstede, G. (2001). Culture’s consequences: Comparing values, behaviors, institutions, and organizations across nations. Thousand Oaks: Sage.Google Scholar
- Hofstede, G., Hofstede, J., & Minkov, M. (2010). Cultures and organizations – software of the mind: Intercultural cooperation and its importance for survival. New York: Mcgraw-Hill.Google Scholar
- Kluckhohn, C. (1951). The study of culture. In D. Lerner & H. Lasswell (Eds.), The policy standard (pp. 393–404). Stanford: Stanford University Press.Google Scholar
- Lenartowicz, T., & Johnson, J. P. (2002). Comparing managerial values in twelve Latin American countries: An exploratory study. Management International Review (MIR), 42(3), 279–307.Google Scholar
- Lytle, A., Brett, J., Barsness, Z., Tinsley, C., & Janssens, M. (1995). A paradigm for confirmatory cross-cultural research in organizational behavior. Research in Organizational Behavior, 17, 167–214.Google Scholar
- Minkov, M. (2007). What makes us different and similar: A new interpretation of the world values survey and other cross-cultural data. Sofia: Klasika y Stil Publishing.Google Scholar
- Naroll, R. (1970). The culture-bearing unit in cross-cultural surveys. In R. Naroll & R. Cohen (Eds.), A handbook of methods in cultural anthropology (pp. 721–765). New York: Natural History Press.Google Scholar
- Schwartz, S. (1994). Beyond individualism/collectivism: New cultural dimensions of values. In U. Kim, H. Triandis, C. Kagitcibasi, S. Choi, & G. Yoon (Eds.), Individualism and collectivism: Theory, methods and applications (pp. 85–119). Thousand Oaks: Sage.Google Scholar
- Steenkamp, J. (2005). Moving out of the U.S. Silo: A call to arms for conducting international marketing research. Journal of Marketing, 69(4), 6–8.Google Scholar
- Tan, J. (2002). Culture, nation, and entrepreneurial strategic orientations: Implications for an emerging economy. Entrepreneurship Theory and Practice, 26(4), 95–111.Google Scholar
- Triandis, H. (1972). The analysis of subjective culture. New York: Wiley.Google Scholar
- Triandis, H. (1994). Culture and social behavior. New York: McGraw-Hill.Google Scholar
- van Vijver, F., & Leung, K. (1997). Methods and data analysis for cross-cultural research. Thousand Oaks: Sage.Google Scholar
- Waldman, D. A., de Luque, M. S., Washburn, N., House, R. J., Adetoun, B., Barrasa, A., et al. (2006). Cultural and leadership predictors of corporate social responsibility values of top management: A GLOBE study of 15 countries. Journal of International Business Studies, 37(6), 823–837.CrossRefGoogle Scholar