Addressing the cross-country applicability of the theory of planned behaviour (TPB): A structured review of multi-country TPB studies

Introduction

The theories of reasoned action and planned behaviour (TRA/TPB; Fishbein and Ajzen, 1975; Ajzen, 1985; 1991) have received substantial research interest. The TPB is an expectancy value model that states that behaviour is a consequence of one's behavioural intention (the cognitive representation of a consumer's motivation to enact the behaviour), which is in turn explained by the consumer's attitude (positive or negative evaluation of undertaking the behaviour), subjective norm (perceived peer pressure to enact the behaviour) and perceived behavioural control (perception of the ease or difficulty in performing the behaviour). The TPB is an extension of the TRA (Fishbein and Ajzen, 1975) that does not include perceived behavioural control and thus is not designed to explain behaviours that are outside an individual's volitional control. A large number of reviews and meta-analyses have concluded favourably on the ability of the TRA/TPB to explain intention and behaviour across a wide spectrum of contexts (e.g. Albarracin et al., 2001; Armitage and Conner, 2001; Conner and Armitage, 1998; Godin and Kok, 1996; Hagger et al., 2002; Sheeran, 2002; Sheeran and Taylor, 1999; Sheppard et al., 1988; Trafimow et al., 2002; Webb and Sheeran, 2006). There is a rich and diverse literature on the use of the TRA/TPB within consumer behaviour with researchers using these theories to understand consumption contexts such as bidding in online auctions (Bosnjak et al., 2006), purchase of free-of cosmetics (Hansen et al., 2012), shoplifting (Tonglet, 2002) and exercise participation (Yap and Lee, 2013) among others.

One important area of interest that has not been examined within the TRA/TPB research domain concerns the central question as to whether the TRA/TPB can be considered to apply universally (i.e. etic) or indeed whether the TPB is culture bound (i.e. emic). This is important in order to demonstrate the model's applicability and generalizability across national boundaries. Cross-country research is important as it can ‘test the universality and generality of theories and concepts developed in relation to one country to other societies’ (Watkins, 2010 p. 709). The universal acceptance and application of the TRA/TPB have remained largely unchallenged (Malhotra and McCort, 2001), yet researchers have noted that there is a general lack of valid theories that work across countries (Maheswaran and Shavitt, 2001). The current study is the only study to date that systematically reviews studies that utilized the TRA/TPB model across more than one country. The objectives of this review are to (i) document research studies since 2000 that applied the TRA/TPB across more than one country within consumption related topics; (ii) critique the methods applied to assess differences in the TRA/TPB model relationship across countries; and (iii) determine whether the strength of the TRA/TPB model relationships vary across country samples. The findings of this review result in useful recommendations of best practice for cross-country researchers wishing to examine the universal applicability of not only the TRA/TPB but also other theories. The findings can also provide further insight into the behaviour of consumers across different countries and can potentially inform global marketing managers when designing and framing marketing messages.

Culture and its Role Within the Theory of Reasoned Action/Theory of Planned Behaviour

Culture has been defined by Hofstede (1980, p. 25) as ‘the collective programming of the mind which distinguishes the members of one human group from another’. At the country-level, culture operates through languages, education systems, social structures, religions, legal systems and so on. Cultural differences can be examined at various levels including individual, sub-unit, organizational as well as country levels (Leidner and Kayworth, 2006; Soares et al., 2007). In this review, TRA/TPB studies are examined at the country level. Separate nations are considered because ‘a culture can be validly conceptualized at the national level if there exists some meaningful degree of within-country commonality and between-country differences in culture’ (Steenkamp, 2001 p. 36). The work carried out by researchers such as Hofstede (1991) has evidenced systematic differences between countries. Furthermore, Smith and Schwartz (1997) examined cultural differences within regions of China, Japan and the USA but actually found much larger differences between the three countries. An examination of culture at the country level can therefore provide a better understanding than an analysis of within-country factors as to why individuals from one country (as against others in a different country) behave the way they do. Indeed, Hofstede (1991, p. 12) asserts that nations ‘are the source of considerable amount of common mental programming of their citizens’.

The examination of the TRA/TPB across countries can improve our understanding of the same behaviour across various cultures. In particular, researchers can learn whether and to what extent the model is culturally sensitive. Ajzen (1991) expected that the three components would predict intention equally well across samples and cultures. However, the claim that the TRA/TPB applies universally (i.e. etic) is contested by the extent to which the TRA/TPB components and their interrelationships are dependent on country/cultural characteristics. Researchers (e.g. Kirkman et al., 2006) have asserted that cultural values may moderate relationships such as those present in the TRA/TPB and earlier findings do suggest that the TRA/TPB works differently in different cultures. For example, in Lee and Green's (1991) study, the importance of attitudes and social norms in determining behavioural intentions were substantially different across Korean and American participants with social norms having a stronger impact on intentions in Korea, a collectivistic culture. More recently, Riemer et al. (2014) argued that the roles of attitude and norms on intention and behaviour are affected by cultural factors such as Hofstede's individualism/collectivism dimension. Hofstede's (1980) framework is the most widely used cultural framework in marketing and is useful for formulating hypotheses in cross-cultural studies (Rabl et al., 2014; Soares et al., 2007). Drawing on Hofstede's framework, we outline possible ways in which culture (at the country level) may influence the TPB model relationships. In line with Rabl et al. (2014) and Zaheer et al. (2012), we decided to focus on a small number of the most relevant cultural dimensions to ascertain whether these dimensions moderate the relationship between intention and its antecedents. For the purpose of the current study individualism–collectivism and power distance are chosen as the two Hofstede dimensions that have received the most empirical attention (Taras et al., 2010).

The five propositions given above provide a theoretical viewpoint as to how the TRA/TPB relationships may be influenced by culture, but if Ajzen's (1991) assertion that the TRA/TPB applies universally is correct these propositions will fail to be supported. Understanding these proposed effects is important for consumer behaviour researchers as increasingly the internet provides companies the opportunity to trade across the globe. It is thus important to take account of country-specific factors that would influence how consumer intentions and behaviours are formed across different countries.

Method

To achieve the objectives of this review, past studies were selected for inclusion if they met the following criteria. Firstly, the studies had to be a quantitative application of the TRA or TPB. Further, studies (e.g. Singh et al., 2006) that drew on the TRA/TPB but did not operationalize the TRA/TPB constructs as specified by Fishbein and Ajzen (1975) or Ajzen (1985; 1991) were excluded. For example, Singh et al. (2006) utilized the TPB as an overarching theory but captured the constructs differently (for instance, measuring ease of navigation as a surrogate for perceived behaviour control). Secondly, studies were required to be related to consumer behaviour and as a result must have explored a consumption context. Here, a broad view of consumption was taken to include papers that not only examined consumer purchases but examined individuals' behaviours and decision making in a range of consumption categories such as health promotion/prevention, adopting technology as well as philanthropic consumption acts. Thirdly, studies must have examined data from at least two different countries and reported in sufficient detail on the interrelationships posited by the TRA/TPB for each country so as to meet the requirement for comparison on the TRA/TPB relationships across the countries studied. Fourthly, this review included studies published since 2000, as the number of studies applying the TRA/TPB has intensified since then and researchers have begun questioning whether the TRA/TPB is applicable universally (Malhotra and McCort, 2001). Lastly, this review included only articles published in English language peer reviewed journals and excluded other pertinent contributions from books, theses, online publications and conference publications.

Given the diversity of topics that could be classified as consumption related, the authors did not use any specific terms related to consumption during their literature search. Rather the focus was on identifying all TRA/TPB published research studies, which were undertaken across different countries. The authors began by searching for Ajzen's (1991) article in the Web of Science database and conducted a cited reference search (within Web of Science) that identified all published articles that had cited Ajzen's (1991) article. Within this set of 7,303 articles (as of 28th March 2013) that had cited Ajzen's (1991) article, the authors searched for articles that contained target texts using a mixture of terms: ‘countries’, ‘cross-cultural’, ‘cross-national’ and ‘cross-country’. Examination of the titles and abstracts in the result of this search led to the identification of 350 articles with a cross-country application of the TRA/TPB. Google Scholar was also used to locate Ajzen's (1991) article and then identify subsequent articles that had cited Ajzen's (1991) article (by clicking on the ‘cited by…’ tab underneath the listing). Among this set of articles, similar search terms were applied identifying around 150 articles. To identify additional TRA studies, which might not have cited Ajzen's (1991) work, a search on Google Scholar was conducted using a combination of the search terms: ‘countries’, ‘theory of reasoned action’, ‘TRA’, ‘cross-cultural comparison’, ‘cross-national comparison’, ‘cross-country comparison’ ‘attitude’, ‘subjective norm’ and ‘consumer’. The authors also conducted cited reference searches on three consumer related articles (Bagozzi et al., 1991; Lee and Green, 1991; Malhotra and McCort, 2001), which are frequently cited in studies that apply the TRA across countries. Last, the authors examined the references cited in articles selected for the review to further check for any missed articles. As a result of the above process, approximately 400 articles were identified for further examination. The lead author scrutinized the abstracts of these articles against the inclusion criteria and identified 79 where the full text required detailed examination. Careful scrutiny of the full texts showed 53 articles to be deemed not to fit the inclusion criteria by the lead author. A second author independently conducted the same detailed full text examination of the 79 articles and verified that most of the articles should be excluded but had reservations about the exclusion of three. After discussions between all the authors, these three articles were also included in the review giving the final number of papers to be included in the review as 29.

For this review (Table 1), the authors extracted a number of attributes beyond the basic information (authorship, source, publication date, context, population and countries studied) on each article reviewed. Given the distinction between the TRA and the TPB, the authors recorded whether the TRA or the TPB is utilized in each study. To examine the strength of the evidence regarding the emic versus etic nature of the TRA/TPB model, information on the significance of the paths hypothesized in the TPB model for each of the countries studied was extracted. Comments are given on the explanatory power (R²) of the TRA/TPB for both intention, and where possible behaviour, across each country. The review records additional explanatory variables, impacting intention or behaviour, that were included along with the TRA/TPB antecedents so that the reader can meaningfully compare the R² values across the studies reported. The review also documents the assessment of three key methodological considerations. These pertain to firstly the translation method employed in the development of the questionnaire, secondly whether an assessment of measurement invariance was undertaken and the level of invariance established and finally whether statistical differences in the strength of the structural paths in the TRA/TPB model were examined. Translation equivalence and measurement invariance need to be established in order to ensure that the meaning of the TRA/TPB constructs in each study were comparable across country samples. This issue is imperative for researchers to make valid cross-country comparisons (Watkins, 2010). For completeness in our reporting and to allow for comparisons across the studies reported, the analytical approach (e.g. multiple regression) adopted is also recorded. From a theoretical point of view, it is important to note the rationale behind the selection of the chosen countries; thus, information is extracted on the criteria for the selection of the countries used in each study and relatedly the nature of any cross-country hypotheses posited.

To provide further evidence on the emic versus etic nature of the TRA/TPB and to test the propositions developed two methods of analysis were applied to the data collected as part of this review. Firstly, pair-wise tests of difference in the correlations for each of the three TPB relationships reported within a reviewed article were undertaken where possible. Specifically, if an article reported correlation values (and sample sizes) for one of the TPB model relationships across two or more country samples, then a formal assessment based on Fisher's Z-test was undertaken to conclude if the correlation values differed across a pair of country samples. In articles where unstandardized beta values and associated standard errors (or t values) were reported (but not the correlation values), a test of difference in the beta weights across the two country samples was undertaken. The results of these formal tests are reported (see italic results in Table 2) to provide additional aid to the reader in interpreting the cross-country results. However, this piece-meal approach in assessing cross-country variations in the TRA/TPB model relationships does not allow a rigorous examination of the propositions posed in the current research. To overcome this limitation, a second systematic analysis was also undertaken. Thus, multilevel modelling (via hierarchical linear modelling: HLM) was used to assess if the strength of the relationship between intention and its antecedents (attitude, subjective norm, perceived behavioural control) varied across the countries included in the review. The HLM analysis assesses if cross-country variations can be systematically explained by the Hofstede dimensions as proposed. The data for this analysis is based on extracting the correlation value, or when the correlation value is not reported the standardized beta weight adjusted using the formula provided by Peterson and Brown (2005). In order to take into account that a number of correlation values are provided within each article, dummy variables were used to control for within article variations. Articles were included in the HLM analysis if data were obtainable within each country sample for two correlation values (for the TRA) or for all three correlation values (for the TPB). In total 67 level-1 cases (covering 24 articles across 17 countries) were used to examine the impact of the Hofstede country-level dimensions on the two TRA relationships. However, a number of articles did not cover the full TPB model and so only 52 level-1 cases (covering 19 articles across 16 countries) were used to examine the impact of the Hofstede dimensions on the perceived behavioural control–intention relationship. Country-level scores for the two Hofstede dimensions were retrieved from Hofstede's website (http://geert-hofstede.com/countries.html).

Findings of the Review

The discussion of the articles identified in this review is organized as follows. First, an overview of articles identified is presented before moving on to discuss substantial methodological and theoretical issues arising from the findings as presented in Tables 1 and 2.

Context, source and theory applied

The articles included in this review investigated the impact of culture on the TRA/TPB across a range of consumer contexts including food purchases, leisure activities, pro-environment behaviours, technology usage and health choices. The most common consumption issues explored were technology and food consumption. Most of the articles were published in context-related journals (e.g. Appetite) with some published in mainstream marketing journals (e.g. International Marketing Review and Journal of Consumer Psychology). Figure 1 shows that a greater number of the studies reviewed were conducted toward the end of the review period.

The majority (23 out of 29) of the articles reviewed employed the TPB while six employed the TRA. Only seven of these studies captured behaviour at a later time period from intention. In terms of augmenting the TRA/TPB, around a third (11 out of 29) of the studies included additional constructs as antecedents of either intention or behaviour.

Countries and samples studied

The average number of countries studied was three with over half (16 out of 29) examining only two countries. Researchers (e.g. Franke and Richey, 2010) recommend the use of at least seven to ten countries to explore cross-national phenomena. Only two articles, each investigated eight countries, met this criterion (Ruiz de Maya et al., 2011; Saba et al., 2008). Figure 2 provides a breakdown of the frequencies of countries investigated across the articles reviewed. The figure shows that the US, China, South Korea and the UK were the most popular countries chosen. In total, 28 different countries spanning all the continents except Antarctica were examined with ten countries examined only once. Very few studies included countries in Africa or South America.

Regarding the rationale for the choice of countries, the reason most frequently given was cultural differences (e.g. based on Hofstede's dimensions) with some based on context specific reasoning. In some articles, (e.g. Ruiz de Maya et al., 2011; Saba et al., 2008) little or no discussion was provided on the choice of countries selected. The lack of rationale behind the selection of the countries chosen was also evidenced in the nature of country-level hypotheses formulated. Almost half of the articles (14 out of 29) failed to develop country-level hypotheses of any kind. Only seven of the articles formulated specific country-level hypotheses for one or more of the relationships in the TRA/TPB model. A further eight articles offered more general hypotheses stating an expectation of either no or some country/cultural differences.

Both student and mainstream consumer samples were common and sample sizes were mainly satisfactory, with most analysing samples of approximately 200 per country. A small number of studies utilized sample sizes, which were much higher and closer to 1000 per country which make statistical tests more sensitive thus yielding more significant results (e.g. Ruiz de Maya et al., 2011; Tsai and Coleman, 2005). One exception that utilized small sample sizes (of around 50) was Pavlou and Chai (2002), and the results of this study should therefore be interpreted with caution.

Discussion and Recommendations

This review has evidenced cross-country differences in the TRA/TPB model relationships. Principally regarding the relationship between subjective norm and intention whereby countries high in power distance (e.g. China, Saudi Arabia and Iraq) have a stronger association between subjective norm and intention. The systematic assessment of cultural differences using multilevel analysis methods overcame the limitations of the studies reported regarding their ability to provide a testable rationale in explaining the observed variations across the country samples studied. One reason for this failure lies in the small number of countries being assessed in the articles reviewed. This limitation is further compounded in the articles reviewed by the lack of specific and well-argued cross-country hypotheses offered as well as the lack of invariance testing.

Nonetheless, the review findings do highlight some variation in the influence of attitude and perceived behavioural control on intention across countries and contexts. However, few articles made predictions regarding differences in the attitude–intention and perceived behavioural control–intention relationships. Of these, the consensus was that no variation was expected across country samples. This view is partially supported by this review as no systematic variation was evident for either of these relationships. Thus, although differences across countries in these relationships do occur, overall they may not vary consistently or with enough magnitude to conclude that cultural differences played a part. However, the lack of statistical evidence may be attributable to the small sample size at the country level (n = 16 or 17) in the country-level regression within the HLM analysis.

On the other hand, the impact of subjective norm on intention is found to vary significantly across countries in this review. A number of studies (e.g. Bagozzi et al., 2000; Chan and Lau, 2002; Jin et al., 2012; Pavlou and Chai, 2002) formulated specific hypotheses for the subjective norm–intention relationship drawing on Hofstede's (1980) individualism dimension. The hypotheses in these studies proposed that the role of subjective norm on intentions would be stronger in collectivistic or interdependent cultures; however, the role of power distance as a moderator was not explored. Therefore, the findings from this review agree with Maheswaran and Shavitt's (2001) observation that cross-country research is still too focused on one of Hofstede's cultural dimensions (individualism–collectivism) as our multilevel results show that power distance rather than individualism explains systematic differences in the subjective norm–intention relationship. The results from a meta-analysis (Manning, 2009) examining the role of norms within the TPB found that descriptive norm (perceived prevalence of the behaviour) and injunctive norm (perceived peer approval/disapproval) are conceptually different and as a result both measures should be modelled within the TRA/TPB. Therefore, although the findings of this review highlight differences in the impact of norms across countries, there is a need for future research to capture and model both normative components in order to provide further insight into the cross-country consistency of the effects of these normative elements. Further, researchers need to go beyond individualism–collectivism and explore Hofstede's other cultural dimensions as well as other cultural frameworks such as Schwartz's (2006) cultural values when forming their hypotheses. Drawing together the results of the review, it can be concluded that the TRA/TPB operates differently across countries and as a result researchers need to take into account the country of study when discussing the findings from their studies. Thus, although an etic approach may be adopted in the collection and analysis of cross-country data, an emic approach to the interpretation of the results should take national culture into account and potentially provide a richer theoretical understanding of the role of culture within the TRA/TPB. A caveat to our research, however, is that it cannot be ruled out that the country-level differences identified are not due to factors other than Hofstede's dimensions (e.g. other cultural values, study variations and methods of data collection), and hence, future research should control for a wider range of factors that could impact on how the TPB operates across studies.

Maheswaran and Shavitt (2001) recognize that there is not enough attention paid to societies with a rich cultural heritage such as Latin America, Africa and the Middle East—a view that is evidenced herein. Thus, future studies should consider applying the model in these societies. TRA/TPB studies across a greater diversity of cultures would enable a stronger assessment of whether the TRA/TPB is indeed influenced by cultural factors.

Related to the above is the importance of using a large number of countries in cross-country studies. Franke and Richey (2010) recommend the use of seven to ten countries to explore cross-national phenomena. However, it should be noted that comparing Eastern and Western cultures using one Eastern country and six Western countries would not suffice in order to achieve a balance of cultural influences. Many of the reviewed studies examined only two countries on the basis of differences in Hofstede's individualism/collectivism scores. As influences on people's behaviours in these countries are likely to differ for other cultural (e.g. power distance, egalitarianism versus hierarchy) or contextual reasons (e.g. economic factors, market structures and maturity), one way around this would be to examine two or more highly individualistic countries and two or more countries with low individualism scores. This would provide reassurance that the likelihood of other cultural or contextual factors affecting the validity of the results is reduced. However, to be able to confidently assert that a particular cultural factor has an impact on the TRA/TPB components or interrelationships, the authors recommend that multiple (around 30 or more) countries with a wide cultural variation should be examined with multilevel analysis employed to assess the country-level (cultural) causal effect. The large number (n > 30) of country samples is needed in order to allow a valid examination of the country-level regression model within the multilevel analysis.

A further recommendation is that future studies should include a clear rationale explaining why the particular set of countries has been selected, as this is critical for theory development. The rationale should then lead to specific hypotheses formulated at the country level as results obtained at an individual level may not necessarily be valid at the country-level (Tsui et al., 2007). Lastly, researchers must conclude on the country-level hypotheses in a rigorous manner, namely based on sound statistical reasoning. Qualitative comments contrasting the observed size of scale means, path/regression coefficients or R² values are anecdotal at best and can often be misleading. Thus, formal statistical tests such as chi-square difference test (e.g. Olsen et al., 2008), Chow test, or moderated regression followed by simple slope analysis (e.g. Yun and Park, 2010) are necessary to draw valid conclusions on the specific hypotheses posed. Although our research examines the operalization of the TRA/TPB at the country-level, further research is needed to also examine within-country or subcultural differences, which would also likely impact the application of the TRA/TPB. Finally, our review is limited given the small number of countries (17) examined across the 29 articles reviewed. In particular, multilevel analysis with only 17 countries as level-2 units offers limited opportunity to evidence significant results. This had led to our inability to assess two of our propositions. With the increase in the reporting of cross-country TRA/TPB studies, future reviews should be able to examine more articles and country samples.

In light of increasing global marketing practices, these findings have important practical implications for global marketing managers in understanding drivers of consumer behaviour and when framing marketing messages across countries. Marketing managers should bear in mind that in high power distance countries (e.g. China, Saudi Arabia and Iraq), the impact of subjective norm on behavioural intention is stronger thus marketing messages incorporating references to societal norms in these countries would likely be more effective in persuading consumers to engage in behaviours. Lastly, although the multilevel analysis failed to evidence systematic cross-country variation in the attitude–intention and perceived behavioural control–intention relationships, this review has nevertheless identified some evidence of cross-country variations in these relationships. Thus, cross-country marketing campaigns need to take these variations into account in considering the adoption of a standardized marketing approach.