CULTURAL ASPECTS OF COMPULSIVE BUYING IN EMERGING AND DEVELOPED ECONOMIES: A CROSS CULTURAL STUDY IN COMPULSIVE BUYING

. Although several studies focused on understanding of compulsive buying in developed countries, this phenomenon remains understudied in other parts of the world. This is rather surprising since there is an increasing interest in understanding shopping behavior of consumers in emergent markets due to the growing importance of these markets. The main reason for the limited attention to compulsive buying in emerging countries is the lack of cross-culturally validated scales. In response to these calls, this paper tests measurement invariance of two prominent compulsive buying scales—the Compulsive Buying Scale (CBS) and the Compulsive Buying Index (CBI) in Western (Spain and the Netherlands) and emerging (Russia and Turkey) economies. In case of lack of invariance the reasons in terms of socio-cultural factors and country conditions are explained. The results establish the partial measurement invariance of the CBI but not the CBS. So, to study the antecedents and consequences of compulsive buying in cross-cultural contexts, the CBI is sufficient. The varying credit card ownership and usage, and different gender roles of women across countries appear to be the main reasons for lack of measurement invariance of the CBS. The percentages of compulsive buyers in emerging countries are lower than those in developed countries.


Introduction
Emerging markets have recently witnessed a rapid process of retail marketplace change: increase in the number of shopping centers/malls, more variety in retail shelves, increasing availability of global brands at marketplace. Furthermore, the purchasing ISSN 2029-4581. ORGANIZATIONS AND MARKETS IN EMERGING ECONOMIES, 2013 power of consumers in emerging markets has increased. ese changes might cause an increase in hedonic shopping value in emerging market consumers, which may result in an increase in the prevalence rate of compulsive buying behavior in these countries. Unfortunately, however, hardly any studies have been conducted on this topic in emerging markets. e main reason for the limited a ention to compulsive buying behavior in emerging countries is the lack of cross-culturally validated scales. erefore, in this article we investigate the measurement invarance of compulsive buying scales that have been broadly used in Western countries. Various sociocultural factors, country shoppingscape, country conditions, for instance availability and usability of a credit card, might in uence the validity of scale items that measure compulsive buying. erefore, in case of scale lack of invariance of the scale, we explain its reasons in terms of socio-cultural factors and country conditions. Additionally, we investigate whether and why compulsive buyers may face di erent consequences of their behavior in di erent countries. Finally, we compare the prevalence rate in emerging and developed countries. In our analysis we rely on data from Western (Spain and the Netherlands) and emerging (Russia and Turkey) economies.
Compulsive buying, or "chronic and repetitive purchasing that becomes a primary response to negative events or feelings" (Faber & O'Guinn, 1992, p. 459), is a prevalent, growing and destructive phenomenon that produces harmful outcomes for consumers and society. e importance of researching compulsive buying behavior is re ected in studies indicating a growing number of compulsive buyers worldwide. Compulsive buying behavior a ects an estimated 2-8% of the general adult U.S. population (Black, 2001;Koran, Faber, Aboujaoude, Large, & Serpa, 2006), and its prevalence appears similar in other Western countries, such as Germany (Mueller et al., 2010), France (Lejoyeux, Mathieu, Embouazza, Huet, & Lequen 2007), and Great Britain (Di mar & Drury, 2000). Compulsive buying appears to be a growing problem in Western countries (Neuner, Raab, & Reisch 2005). From 1990 to 2001, the proportion of compulsive buyers jumped from 5% to 8% in West Germany and from 1% to 7% in the former communist East Germany. Given the East and West Germany case, the prevalence of compulsive buying is likely to increase in the next decades in emerging countries as well.
e prevalence and increase of compulsive buying in di erent cultures suggest that it is not solely a Western problem, though sociocultural, environmental, and contextual factors may a ect its development. Despite some recognition of the potential in uence of contextual and cultural factors on compulsive buying (e.g., Neuner et al., 2005;O'Guinn & Faber, 2005), empirical tests have been hampered by a lack of cross-culturally validated instruments. Existing instruments, such as the Compulsive Buying Scale (CBS; Faber & O'Guinn, 1992) and the Compulsive Buying Index (CBI; Ridgway, Kukar-Kinney & Monroe, 2008), were developed and tested in wealthy Western societies (U.S. or Canada), but their cross-cultural invariance (e.g., Van de Vijver & Leung, 2000) has not yet been investigated in emerging countries. is gap threatens the validity of any cross-national comparisons using these instruments. For example, when scalar invariance is absent, a cut-o point used to indicate compulsive buying in one culture is not valid in another. Lack of measurement invariance of the scale prevents any comparison of the relationships of the scale with other scales. Because the meaning of the construct is not necessarily the same across cultures, any cross-cultural comparison using the scale is invalid. With this study, we therefore examine whether the CBS and CBI, the two most prominent compulsive buying scales, are equivalent across emerging and developed economies. e CBS is by far the most commonly used scale to measure compulsive buying (Ridgway et al., 2008). As a result, the current compulsive buying pro le is mainly based on studies that use the CBS (see, e.g., Black, 1996;Kuzma & Black, 2006); it has been used to measure the prevalence of compulsive buyers, to identify antecedents and consequences of compulsive buying (Scherhorn, Reisch, & Raab, 1990;Faber & O'Guinn, 2008), and to discover psychiatric comorbidities (Black, Repertinger, Ga ney, & Gabel, 1998). Because the CBS was developed to identify severe cases of compulsive buying, its goal is to indicate people who should receive treatment to change their behavior. In contrast, the more recent CBI aims to measure "a consumer's tendency to be preoccupied with buying that is revealed through repetitive buying and lack of impulse control over buying" (Ridgway et al., 2008, p. 461). e CBI has several promising features, since it has been developed to overcome problems with existing compulsive buying scales. In particular, it de nes and measures compulsive buying only in terms of underlying behavioral tendencies, and it allows for an expanded conceptualization of the construct by incorporating both obsessive-compulsive and impulse-control buying dimensions. Unlike other scales measuring compulsive buying behavior, the CBI avoids items re ecting consequences of compulsive buying. Eliminating the consequences of the behavior from measuring it is a crucial step to obtain etic scales (i.e. cross-culturally universal). Since consequences of compulsive buying might be experienced di erently across developed and emerging countries, the CBI is highly likely more cross-cultural equivalent than the CBS scale. e instruments developed in the United States might not be useful in other countries, speci cally in emerging countries, to identify compulsive buying behavior for several reasons. e instruments may contain items on consequences that are not applicable in other cultural contexts and may have been developed with very limited samples (e.g., student samples) that are not re ective of the population, especially in other cultural se ings. If this is the case, the scales may be invalid for use in other cultural se ings and the resultant scores incomparable across cultural groups, even within the same country. In sum, the objectives of the current study are: establishing measurement invariance of two compulsive buying scales; explaining the reasons for lack of invariance, if that is present, in terms of socio-cultural factors and country conditions; and comparing the prevalence rate across in emerging and developed countries.

Culture and Measurement Invariance
While the knowledge of the role of psychological and physiological factors on compulsive buying is extensive due to a great deal of previous research, the understanding of the relationship between socio-cultural factors and compulsive buying are rather limited (O'Guinn & Faber, 2005). Cultures di er in the nature of the boundaries between the person and the group: in some cultures people are more autonomous and in other cultures people are more embedded in groups (Schwartz, 2006). In autonomous cultures, mainly located in Europe and the U.S., people are encouraged to cultivate and express their own preferences and feelings; being unique is important and meaningful. In embedded cultures, mainly located in East Asia, social order and respect for tradition are valued and identifying with the group is important, as are shared goals. Such di erences in values a ect people's behavior. Norms imposed by the in-group are the main motivator for people in embedded societies and people try to emphasize their connectedness to the in-group. In autonomous societies, the self prevails over the group. People in more embedded societies are expected to keep harmonious relationships and may therefore be more motivated to suppress impulses.
is doesn't mean that the impulsive tendencies are non-existent: Kacen and Lee (2002) found that although the impulse buying trait was present in Asian consumers, they engaged less in impulsive buying behavior. Gender roles, social expectations, and social norms in a country may also in uence the prevalence of compulsive buying. Changing social dynamics in emerging countries might create a feeling of loneliness, alienation, depression, which might result in a wrong perception, compulsive buying as an escape solution, similar to consumers in developed countries. Before scale scores can be used in cross-cultural comparisons, their cross-cultural invariance (e.g., Van de Vijver & Leung, 2000) in other countries needs to be established. When there is no invariance, correlations between scales, let alone absolute scores on the scales, cannot be compared across countries. Existing instruments, such as the CBS and the CBI, were developed and tested in the U.S. (or Canada), but not in other countries. is gap threatens the validity of any cross-national comparisons using these instruments and the usability of these scales outside North America. e more recent CBI has been developed to overcome some problems with existing scales for compulsive buying behavior such as CBS. One of the problems with the CBS is that it includes the consequent harm within the measure itself, which may limit the potential for invariance across cultures. Harm or adverse consequences should be classi ed as outcomes, rather than as a dimension of the instrument. It is likely that consequences of the behavior will be experienced di erently by individuals in di erent cultural se ings. Furthermore, the CBS is limited to a focus on the more severe aspects of the behavior, whereas the CBI incorporates two dimensions: the early symptoms (i.e., impulse buying), as well as the more severe aspects (i.e., obsessive-compulsive buying). Speci cally, the CBS was developed to identify severe cases of compulsive buying and its aim is to indicate people who should receive treatment to change their behavior. In contrast, the CBI de nes and measures compulsive buying only in terms of underlying behavioral tendencies, and it allows for an expanded conceptualization of the construct by incorporating both obsessive-compulsive and impulse-control dimensions.
In this study, we therefore examine whether the CBS and CBI, the two most prominent compulsive buying scales, are invariant across developed and emerging economies. If we can establish the cross-cultural invariance of the CBS and CBI, we also test the scales' criterion validity in di erent cultures. For example, we may identify whether some known consequences of compulsive buying, such as nancial di culties, hiding, feeling guilty, or family arguments, emerge across cultures. We expect that the correlations of the compulsive scales with consequences and behaviors will be less strong in the more culturally embedded countries. e countries di er on the cultural value orientation autonomy versus embeddedness (Schwartz, 2006), where Russia and Turkey score relatively higher on embeddedness and Spain and the Netherlands on autonomy. e e ects are expected to be strongest in culturally autonomous Western countries (the Netherlands and Spain), which are culturally and economically the most similar to the United States, the country where both the CBS and the CBI were developed.

Participants
We collected data from women shopping at malls in large cities in four countries: Amsterdam (the Netherlands), Madrid (Spain), Moscow (Russia), and Ankara (Turkey), using the commonly employed mall-intercept sampling technique. In each country, we only collected data from women born and raised in that country. We focused on women as previous research has indicated that women are more prone to compulsive buying than are men, such that 80-95% of the compulsive buyers reported are women (Black, 2001). e sample sizes are 104, 167, 150, and 122 respondents in the Netherlands, Spain, Russia and Turkey, respectively. e mean ages of these respondents are 33.81 (SD = 13.36) in the Netherlands, 38.95 (SD = 15.91) in Spain, 37.3 (SD = 20.30) in Russia, and 34.24 (SD = 12.20) in Turkey.

Measures
e CBS consists of seven items; the CBI of six items allocated to two dimensions: impulsive buying and obsessive-compulsive buying. We list the CBS items in Table 1 and the CBI items in Table 2. In addition to measuring the CBS and CBI scales, we collected information about shopping frequency, the consequences of compulsive buying (e.g., nancial di culties), and demographic characteristics (e.g., age, education, income). e questionnaires were rst developed in English and then translated into the native languages of each country by bilingual speakers following the common procedures in cross-cultural research; to ensure translation invariance other bilingual speakers then back-translated the questionnaires into English.

Procedures
To assess the cross-cultural invariance of the CBS and CBI instruments, we use multigroup con rmatory factor analysis with structural equation modeling (LISREL 8.8;Jöreskog & Sörbom, 1999). Considering the ordinal nature of the items, we use polychoric correlations. e CBI has a second-order factor structure with two constructs, so we impose an additional constraint of equal error variances for the two constructs to estimate the model. To estimate the model parameters, we use maximum likelihood.
For assessing cross-cultural measurement invariance, we followed the three nested, sequential steps as outlined in cross-cultural measurement literature (e.g., Van de Vijver & Leung, 2000). First, con gural invariance means that the same construct gets measured in each country. We test for con gural invariance by ing the same hypothesized factor structure simultaneously in all four countries (Model 1). ere is con gural invariance when the same number of factors is found in all countries. Second, metric invariance implies that the same measurement unit applies across countries, however, with di erent origins. For the testing of metric invariance, we set the factor loadings to be equal among countries (Model 2, see Dimitrov, 2010). Metric invariance "ensures that items and latent factors have equal meaning across di erent groups or populations" (Carolla et al., 2012, p.229) and, hence, means that one can compare di erences on the scales between people in di erent countries, though it is not su cient to allow a comparison of absolute scores. ird, the scalar invariance assessment requires constraining the latent means of the scales in each country to be equal (Dimitrov, 2010), as we show in Model 3. Scalar invariance means that the same measurement unit and same origin exist in all countries, so the scale scores can be compared across people within and across countries.
If full metric or scalar invariance cannot be established, partial metric or partial scalar invariance might exist. Such partial invariance implies that for a subset of items invariance Impulsive buying 5. I buy things I did not plan to buy. 6. I consider myself as an impulse purchaser.
exists; for that subset, a valid comparison across cultures is possible. e detection of invariant items can be also educational; it provides a unique perspective into how di erent cultures respond to or perceive constructs (Carrola et al., 2012). We compare the nested models (Models 1-3) by subtracting the respective goodness-of-t values and degrees of freedom of the more restricted model from the less restricted model, using a chi-square di erence test. If not signi cant, this test indicates that the model meets the criterion to reach the next level of invariance. Because its performance depends on non-normality and sample sizes, we use a recently proposed correction (Satorra & Bentler, 2010) to compare the models (Chou, Bentler, & Satorra, 1991;Curran, West, & Finch, 1996), and we rely on additional goodness-of-t indices as recommended by Steenkamp and Baumgartner (1998), who indicate ve alternative t indices: root mean square error of approximation (RMSEA), consistent Akaike information criterion (CAIC), comparative t index (CFI), and Tucker-Lewis index (TLI) or non-normed t index (NNFI). ese indices account for both goodness of t and model parsimony by imposing a penalty on additional parameters. As suggested, in case of non-normality, we use NNFI instead of TLI.

Measurement Invariance of the CBS
We start by estimating the con gural invariance model for the CBS (Model 1), whose t is inadequate. Although the CFI and NNFI values are at least .97 (CFI = .98, NNFI = .97), the chi-square value is signi cant (df = 56, SB 2 = 121.39, p < .05), and the RMSEA value is moderate (.093). Moreover, we nd three problematic items in the CBS. e t-value of the factor loading of item 5 ("Bought myself something in order to make myself feel be er") is not signi cant in the Netherlands. 1 is item also produces low standardized factor loadings (from .20 in the Netherlands to .58 in Turkey) and low construct reliability measures (from .04 in the Netherlands to .33 in Turkey). Furthermore, the factor loadings of item 7 ("Made only the minimum payments on my credit cards") are very low and insigni cant in Russia (.05) and Spain (-.03), but both these loadings are high in Turkey (.78) and the Netherlands (.77). Finally, our respondents generally disagreed or strongly disagreed with item 4 ("Wrote a check when I knew I didn't have enough money in the bank to cover it"). is item's mean thus falls below 2, and its standard deviation is very low in all countries, so this item is not informative. Because we cannot establish even con gural invariance for the CBS, we do not proceed with any further analyses for this instrument.
Measurement Invariance of the CBI e CBI originally appeared as a second-order factor model (Ridgway et al., 2008), which requires two supplementary considerations to perform measurement invariance tests (Chen, Sousa, & West, 2005): (1) metric invariance (equal factor loadings) for both the rst-and second-order factors and (2) scalar invariance (equal intercept) for both the measured variables and rst-order factors. We assess the con gural invariance by imposing the baseline model on all the four country samples simultaneously (see Model 1, Table 3). e t of the baseline model is adequate; the chi-square value is moderately insigni cant (SB 2 = 46.95, p > .01), the RMSEA value is low (.059), and the CFI and NNFI values are greater than .97 (CFI = .996, NNFI = .993). We therefore conclude that the CBI scale exhibits con gural invariance across all the four countries. We then assess metric invariance by comparing a model with equal rst-order loadings (Model 2) against the baseline model (Model 1). According to Table 3, this restriction leads to a signi cant increase in chi-square (∆df = 12, ∆SB 2 = 126.63, p < .05). e RMSEA increases to .15, and the decreases in the other goodness-of-t statistics are larger than .01. Because the null hypothesis of full metric invariance of all rst-order loadings thus is rejected, we continue with partial metric invariance tests. From a series of tests in which we sequentially relax the constraints, we conclude that the t of Model 2.1 (all rst-order factor loadings set to be equal except for items 2 and 3) is not signi cantly worse than the t of Model 1 (∆df = 6, ∆SB 2 = 18.23, p > .01). Its RMSEA is also acceptable (.07), and the decreases in NNFI and CFI are less than .01. e two problematic items that hinder invariance belong to the obsessive-buying dimension; therefore, we suspect that the second-order factor loading of obsessive-buying should di er across countries. We compare Model 2.2 and Model 2.1 and nd that Model 2.2 has a signi cantly worse t than Model 2.1 in terms of both the chi-square (∆df = 6, ∆SB 2 = 91.88, p < .01) and the alternative statistics: the RMSEA increases to .14, and the other goodness-of-t statistics decrease by more than .01. When we only restrain the second-order factor loading of the impulse-buying dimension (Model 2.3), we nd no worse t than Model 2.1 (∆df = 9, ∆SB 2 = 4.99, p > .01; RMSEA = .07, ∆NNFI and ∆CFI < .01). Our results thus show that all rst-order item loadings except those for items 2 and 3 and the second-order impulse buying loading achieve metric invariance across the four countries, in support of the partial metric measurement invariance of the CBI.
Finally, we examine its partial scalar invariance by constraining the intercepts of the equal factor loadings, then comparing Model 3 (equal error variances of the equal item loadings and second-order impulse buying factor) against Model 2.3. We nd a signi cant increase in chi-square (∆df = 6, ∆SB 2 = 50.07, p < .01), and the goodnessof-t measures con rm that partial scalar invariance cannot be established (RMSEA = .108, ∆CFI > .01).
In summary, we con rm con gural and partial metric measurement invariance for the CBI. e items in the impulse-buying dimension are measurement invariant across four distinct countries; we do not nd invariance for the items in the obsessive-buying dimension though. Whereas we could not establish con gural invariance for the CBS, the CBI is partially measurement invariant. With to regard the CBI, we can assess the relationships of the CBI with other constructs in a nomological network and thus explore its criterion validity.

Criterion validity of the CBI
Because the CBI has never been used in countries other than the United States, it is important to assess its criterion validity, in addition to its invariance. To do so, we computed correlations between the consequences of compulsive buying and its subscales to indicate the scale's predictive ability. We selected behaviors that Ridgway et al. (2008) call relevant to compulsive buying: hiding, guilt feelings, family arguments, and frequency of shopping trips. In addition, we included the number of pairs of shoes owned and the degree of nancial di culty su ered (Black, Monahan, & Gabel, 1997). We report the results for the obsessive-compulsive and impulsive buying dimensions of CBI separately. As the consequence variables (see Table 4) are ordinal or binary, we used Spearman's rho to measure their relation to the CBI. We also used a t-test to determine if people who hide their purchases, feel guilty, or face family arguments (all yes/no) because of their shopping behavior score signi cantly higher on the obsessivecompulsive and impulsive buying dimensions than do others. e results for the t-test and Spearman's rho are very similar; we therefore only discuss the la er outcomes.
As shown in Table 4, most variables are signi cantly and highly correlated with the CBI, and in the expected direction. Frequency of shopping trips and nancial di culty correlate positively with both dimensions of the CBI across all the four countries; we nd a consistently stronger relation with the obsessive-compulsive buying dimension though. e correlation between number of pairs of shoes owned and both CBI dimensions are also positive, except in Russia, which shows very low correlations with both dimensions (rho=.12 and rho=.10) that are not signi cantly di erent from 0. e biggest di erences arise with respect to the three consequences: hiding, guilt, and family arguments. Hiding exhibits a positive correlation with the CBI in Spain, Russia, and the Netherlands (on average r = .28 for the obsessive-compulsive buying and r = .27 for the impulsive buying dimension); however, both dimensions of CBI turn out to be not correlated to hiding in Turkey. Guilt is signi cantly correlated with CBI in Spain, Russia, and the Netherlands (on average r = .28 for the obsessive-compulsive buying and r =. 35 for the impulsive buying dimension). Finally, family arguments only show a signi cant relation with CBI in Spain (r =.33 for obsessive-compulsive dimension; r =.40 for impulsive dimensions). An explanation for these di erences may be that the three consequences are strongly in uenced by cultural expectations and norms for shopping. We elaborate on this further in the discussion section.

Discussion
Shopping beyond one's income and living in a state of debt has become normal for many people, not only in developed countries, but also in emergent countries. Increasing numbers of people are unable to control their excessive spending and o en end up in debt. Compulsive buying and its consequences undermine people's personal relationships and well-being. In the United States, debt tops 16 trillion, total household debt currently stands at $11.44 trillion; this amounts to an average of $98.500 per household, whereas an average American owes $47,500 (Federal Reserve Bank of New York, 2012). is debt situation is not an exclusively Western phenomenon; similar debt situations have been reported from non-Western countries. erefore, policy makers need to be aware of danger and should take necessary precautions worldwide. Based on the research ndings, we recommend establishing some programs for young adults to build skills to counter nancial problems; developing school programs for nancial planning education; o ering nancial counseling, free of charge, by banks in emerging countries similar to developed countries.

Measurement of compulsive buying behavior
We have investigated whether two compulsive buying scales (CBI and CBS) exhibit measurement invariance in countries that vary in their level of a uence and di er in culture. Compulsive buying scales were developed and validated in the U.S. and Canada, but no research has determined if these scales also are valid in other cultures. Furthermore, research in di erent countries and cultural contexts can provide novel insights into the phenomenon and help clarify how socio-cultural factors in uence compulsive buying.
is interesting research area unfortunately has been developing at a very slow rate thus far (Neuner et al., 2005;O'Guinn & Faber, 2005).
Cross-cultural comparability requires that the scales are at least partial metric measurement invariant. In our study, we establish partial metric measurement invariance for the CBI, however, the CBS does not even have con gural invariance. One of the problematic items for the CBS concerns credit card usage. Problems with this item also have been reported in a previous study by Kwak, Zinkhan, and Crask (2003) in Korea. It is likely that this item lacks functional invariance (Van Herk, Poortinga, & Verhallen, 2005), as card ownership and usage varies greatly across countries. In many countries, credit cards are much less common and more regulated than in the United States. An average U.S. consumer owns 4.6 credit cards (Board of Governors of the Federal Reserve System, 2007), whereas in Europe, this average is only 1.5 (ECB, 2009). e lack of invariance of the CBS might also re ect the socio-cultural context. Banks and retailers in some countries do not encourage credit card use and social norms might discourage unnecessary spending and as a consequence lower acceptance of buying on credit. Accordingly, the lack of con gural invariance for the CBS likely results from items that are too context speci c and that thus lack functional invariance.
Although the CBI is con gural invariant, only the items in the impulse-buying dimension exhibit metric invariance across the four countries. We had to remove items such as being a "shopaholic" or a "life centered around shopping" to obtain metric invariance. e rst one (others might consider me as a "shopaholic") was hard to translate into other languages, so its lack of invariance might be due to a translation bias. e second item ('much of my life centers around buying things') might re ect the gender role of women in these four countries, as we discuss subsequently.

Cultural di erences a ecting compulsive buying behavior
We included three main consequences of compulsive buying in our study: family arguments, hiding, and guilt. e family arguments item relates positively and signi cantly to compulsive buying only in Spain. In neither Russia nor Turkey family arguments turned out to be related to compulsive buying. is result might be explained in two ways. First, the la er two countries score higher on embeddedness (Schwartz, 2006) and, as argued by Kacen and Lee (2002), the compulsive buying tendency might result in less compulsive buying behavior in such countries. e second explanation may be the prevailing gender roles in these countries: the gender equality index (UNDP, 2010) indicates more traditional gender roles in Russia (.44) and Turkey (.62), compared with Spain (.28) and the Netherlands (.17), and, therefore, (extensive) shopping might be a more accepted activity for women within Russian and Turkish society. Moreover, a recent study by Ergin (2010) on compulsive buying in Turkey reveals that traditional gender roles are a primary reason for women in Turkey to engage in shopping activities. is study also nds that Turkish women prefer to shop as a key reaction to negative feelings such as boredom, stress, low self-esteem, or even depression, because shopping is accepted as a woman's role. us, when female shopping behavior has high social acceptability, compulsive buying might not provoke family arguments (e.g., Rook & Fisher, 1995).
Another interesting nding is that hiding and guilt relate positively to CBI in the Netherlands, Spain, and Russia, but not in Turkey. is might be explained by cultural di erences in facing feelings of guilt and shame (Creighton, 1990). In more autonomous countries (e.g., Faber & O'Guinn, 2008), a positive relationship between compulsive buying and guilt might be expected. In such countries, people develop their own standards of conduct and are less in uenced by tight social controls; being unique is valued. When engaging in behaviors, people in autonomous countries compare their actions against their own internalized standards; if they violate these standards, they feel guilt (Bierbrauer, 1992). For instance, the idea of collective guilt is deeply rooted in Protestant (Calvinist) cultures (Baskerville, 1996), such as the United States or the Netherlands. In these counties guilt correlated signi cantly and positively with CBI. Fewer guilt feelings are expected to arise in embedded societies (such as Turkey, which is the most embedded of the 4 countries), which feature tight social controls and where people compare their actions against social norms. If people in an embedded society do not comply, they are criticized by close others and thus feel shame, not guilt.
To sum up, CBI is a con gural invariant and its items in the impulse-buying dimension exhibit metric measurement invariance across the four countries we investigated. We nd evidence of criterion validity and also observe that the way a culture views both impulsive behavior and gender-speci c behavior directly in uences the consequences compulsive buyers may face.

Limitations and further research
We acknowledge some limitations of this study. First, we used female respondents, to ensure comparable samples across countries. We recommend the inclusion of both genders, more and more diverse countries, as well as more individual-level variables, such as cultural values (Schwartz, 1992), personality (Allik & McCrea, 2004), or selfconstrual (Markus & Kitayama, 1991). Multilevel analyses (Van de Vijver, Van Hemert, & Poortinga, 2008), including country as well as individual characteristics might provide a more holistic view of compulsive buying.
Second, although we used a set of known consequences of the compulsive buying behavior, we did not investigate closely why there are di erences in the relationship between the CBI and the consequences across cultures. erefore, further research is warranted on why the strength of the relationship between the compulsive buying trait and compulsive buying consequences and behavior di er among cultures. e factors moderating the trait-behavior relationship need to be uncovered. ese factors can be (a) normative beliefs, (b) self-construal, and (c) peer presence. Rook and Fisher (1995) nd that in the non-embedded U.S. culture, normative beliefs moderate the relationship between the impulse buying trait and buying behaviors, but not between the compulsive buying trait and buying behaviors. In their study on impulse buying across countries, Kacen and Lee (2002) nd that individuals' self-construal (interdependent or independent) a ects the strength of the relationship between the impulsiveness trait and the impulsive behavior. One interesting avenue for future research is to uncover the extent to which within-culture di erences in normative beliefs or self-construal can explain compulsive buying tendencies.
Another interesting avenue for further research is to uncover the role of other people (e.g., peers) in engaging in compulsive behavior. A study by Zhang and Shrum (2009), using student samples in the U.S., showed that individuals with an interdependent selfconstrual tend to display less impulsive tendencies than students with an independent self-construal; peer presence enhanced the e ect for the independents, whereas it depressed the e ect for the interdependents. What still needs to be uncovered is whether the e ects are the same in a consumer se ing, what the role of the cultural context is, and to what extent the peers' characteristics (and gender in particular) ma er. us, further research across cultural contexts on the factors a ecting the strength of the compulsive buying trait-behavior relationship is warranted.

Implications for research
As we failed to establish con gural invariance of the CBS, we recommend that researchers interested in understanding cross-cultural problems, such as how cultural aspects in uence the development and prevalence of the compulsive buying disorder in a culture, should refrain from relying on this measure. We nd that the CBI is be er suited for this purpose. Of the two dimensions of the CBI, the dimension on compulsive buying is a con gural invariant and the dimension on impulsive buying is a metric invariant. is implies that people in culturally di erent countries conceptualize both dimensions of the CBI the same way. Additionally, for the impulsive buying dimension, variances, co-variances, and regression coe cients relating factors can be compared across cultural groups. So, the impulsive buying dimension of the CBI is appropriate to investigate the antecedents and consequences of compulsive buying in cross cultural contexts. Comparisons between the absolute scores on the scale between people from di erent countries, however, cannot be made nor can universally determined cut-o points separate compulsive buyers from non-compulsive buyers.
Scales are o en developed in one country without consideration for crosscultural comparability. ese scales are o en routinely exported from the country in which they were developed to other places without alterations and without checking whether the scale is applicable in the other country. At the same time, international researchers in the social sciences have been focusing on how to develop cross-culturally comparable measures. Ever since Berry's (1969) seminal article, they have argued that country-speci c ("emic") items might have to be added to or replace cross-nationally standardized ("etic") items when developing valid instruments. Recent advances in scale development provide models that allow for emic items while retaining crossnational comparability (see De Jong, Steenkamp, & Veldkamp, 2009). Such scales might be useful to assess compulsive buying tendencies, not only in international contexts, but also in the various ethnically heterogeneous populations within Western countries. We, therefore, recommend researchers in the area of compulsive buying to rely on these models when developing cross-culturally comparable valid scales for compulsive buying.

Implications for the society
Compulsive buying scales have a broad user base, including not just consumer researchers but also psychologists, therapists, and public policy makers. e CBI can be used for this purpose because it turns out to be a partially metric invariant, hence, its items have the same meaning in the di erent cultures/societies and possess the same measurement unit across countries. e scale can be especially useful to check people's compulsive buying tendencies in di erent countries. Our results illustrate that employing frequently used CBS in new cultural groups is inappropriate and may lead to biased results and de cient classi cations, which waste time and resources. To measure compulsive buying tendencies, the CBI can be validly used within a culturally di erent context.