INTRODUCTION
Throughout his career, Kwok Leung was active and innovative in expanding and developing our understanding of distinctive aspects of cultural difference. This article focuses upon his most recent integrative synthesis, jointly authored with Michael Morris (Leung & Morris, Reference Leung and Morris2015). These authors considered the factors on which we must focus if we are to obtain an adequate understanding of the ways in which cultures are created and sustained. Prior explanations that rest solely on contrasts in espoused personal values have come under increased scrutiny, both in terms of studies showing limited value consensus within national cultures (Fischer & Schwartz, Reference Fischer and Schwartz2011), and of evidence that the ability of these values to predict cultural differences may be restricted to specific domains of behavior (Vauclair & Fischer, Reference Vauclair and Fischer2011; Vauclair et al., Reference Vauclair, Fischer, Ferreira, Guerra, Hossler, Karabati, de Carvalho Filho, Porto, Reyes, Ritkonen and Spiess2015). Furthermore, we have mounting evidence of the utility of alternative understandings of culture that are based on subjective norms (Gelfand & Harrington, Reference Gelfand and Harrington2015; Shteynberg, Gelfand, & Kim, Reference Shteynberg, Gelfand and Kim2009) and on experimental inductions of mindsets (Hong, Morris, Chiu, & Benet-Martínez, Reference Hong, Morris, Chiu and Benet-Martínez2000). Leung and Morris delineate circumstances under which each of these psychological carriers of culture becomes operative in guiding behavior. Their perspective has primary relevance to the individual level of analysis, but given that values, norms, and mindsets vary across nations, it can also have potential in explicating differences among nations. This study addresses differences between national cultures, focusing particularly on the contrast that Leung and Morris make between strong and weak situations.
THEORETICAL BACKGROUND AND HYPOTHESES
The method of studying cultural differences using country means of espoused values gained impetus with the pioneering work of Hofstede (Reference Hofstede2001). Despite subsequent evidence of substantial intra-national diversity, there remains a clear consensus that nations can be arrayed along dimensions such as individualism-collectivism in ways that are replicable across measures and across time, even when controls are made for variations in concomitants, such as wealth, and when more representative samples are accessed (Bond & Lun, Reference Bond and Lun2014; Hofstede, Reference Hofstede2001; House et al., Reference House, Hanges, Javidan, Dorfman, Gupta and associates2004; Inglehart & Oyserman, Reference Inglehart, Oyserman, Vinken, Soeters and Ester2004; Schwartz, Reference Schwartz, Vinken, Soeters and Ester2004; Smith, Dugan, & Trompenaars, Reference Smith, Dugan and Trompenaars1996). The measures employed by these authors have included some tapping personal values and others tapping perceived norms. The presence of continuing contrasts between nations that are reflected by these measures may be attributable to shared legal systems, social institutions, access to natural resources and languages (Schwartz, 2014), despite contemporary tendencies toward globalization. Comparisons of cultures as defined by national boundaries is also facilitated by the greater availability of relevant data sources, and indices related to collectivism have been shown to correlate with a broad range of criterion measures (Taras, Kirkman, & Steel, Reference Taras, Kirkman and Steel2010). In this article, I seek principally to disentangle the confounding of norms and values as predictors of national-level variance in relevant behavioral indices.
Leung and Morris (Reference Leung and Morris2015) conceptualize cultures as dynamic systems with considerably less uniformity and rigidity than that envisaged by Hofstede (Reference Hofstede2001). They see individual members of a cultural group as affected at different times by personal values, prevailing norms, or interpretive schemata. In drawing upon their model, we need to bear in mind the importance of distinguishing levels of analysis. The circumstances eliciting individuals' reliance on values, norms, and mindsets may differ from those found when examining large collectivities. To understand how the Leung-Morris perspective might be relevant to the study of cultural differences, we need first to consider the ways in which individuals' behaviors may be influenced by the values and norms of those around them. At the individual level, the contrast between using one's own values as a guide for action and adhering to perceived norms is self-evident. At the collective level, this distinction is somewhat reduced, because an awareness of the values of those around one can form the basis of a different type of perceived norm. Here, the distinction between descriptive norms and injunctive norms first identified by Cialdini, Kallgren, and Reno (Reference Cialdini, Kallgren, Reno and Zanna1991) becomes more useful. Awareness of the values of those around one is an instance of a descriptive norm, since one could rely on that awareness as a guide to one's actions. Injunctive norms are perceptions of what others approve of – how they think one ought to behave (Morris, Hong, Chiu, & Liu, Reference Morris, Hong, Chiu and Liu2015).
Leung and Morris propose that the influence of cultural norms would be most likely in situations where the right answer is ambiguous and social evaluations are highly salient. In contrast they propose that the influence of cultural values would be most likely in situations where one's actions are private or non-identifiable or seen only by those of low expected future interaction. These arguments draw upon the classical construct of situational strength, first formulated by Mischel (Reference Mischel1973), who noted that some situations constrain behavioral options more strongly than others. The Leung and Morris model implies that situational strength will moderate the role of values and norms in driving individual behavior. In strong situations, norms (and especially injunctive norms) will predominate because surveillance and sanctioning is salient. In weak situations, values will predominate because there are few external social cues about how to interpret the task and respond.
The distinction between strong versus weak situations is related to the group-level contrast between tight and loose cultures first postulated by Pelto (Reference Pelto1968) and Boldt (Reference Boldt1978) in relation to differences between pre-industrial societies. In a tight culture, a variety of hazardous circumstances will cause the emergence over time of a set of relatively rigid injunctive norms as to how one should behave in relation to others. In a loose culture, a more benign environment enables the emergence over time of opportunities for more autonomous behavioral choices dependent on one's individual priorities. Measures tapping this contrast have included archive coding (Carpenter, Reference Carpenter2000) and linguistic analysis (Chan, Gelfand, Triandis, & Tzeng, Reference Chan, Gelfand, Triandis and Tzeng1996), but these have been applied to relatively small numbers of cultural groups. More recently, alternative operationalizations tapping many cultures have been developed by Gelfand et al. (Reference Gelfand, Raver, Nishii, Leslie, Lun, Lim, Duan, Almaliach, Ang, Arnadottir, Aycan, Boehnke, Boski, Cabecinhas, Chan, Chhokar, D'Amato, Ferrer, Fischlmayr, Fischer, Fülöp, Georgas, Kashima, Kashima, Kim, Lempereur, Marquez, Othman, Overlaet, Panagiotopoulou, Peltzer, Perez-Florizno, Ponomarenko, Realo, Schei, Schmitt, Smith, Soomro, Szabo, Taveesin, Toyama, Van de Vliert, Vohra, Ward and Yamaguchi2011) and Uz (Reference Uz2015), and these are employed in the present study.
Do tight cultures present people with more strong situations? Leung and Morris (Reference Leung and Morris2015) note that ‘People in tight cultures are chronically aware that their actions are being evaluated, so they are prevention-oriented, cautious, and dutiful; they exhibit more impulse control and more self-monitoring’ (1040). Clearly in both tight and loose cultures, events will occur that require adherence to social norms, but in tight cultures these events are likely to be more frequent and the relevant norms are more likely to be injunctified (Kay et al., Reference Kay, Gaucher, Peach, Laurin, Friesen, Zanna and Spencer2009). In loose cultures, norms will more often be descriptive, providing guidelines as to what values, beliefs, and behaviors are typically found in one's environment, rather than necessarily requiring that one should do likewise.
If tight cultures do indeed present people with strong situations more frequently, the individual-level dynamics postulated by Leung and Morris (Reference Leung and Morris2015) are likely to be also found within group or nation-level data. For instance, the strong situations facing a given individual may also be faced by others around them, and awareness of how others handle such situations may affect one's own actions. To assume that relations will be the same at both levels of analysis is to commit the ecological fallacy (Robinson, Reference Robinson1950), but this does not preclude empirical investigations that test or extend the boundaries within which their model has utility. A full test of the Leung and Morris model would require data in which individual behavior could be predicted from personal values versus perceived injunctive norms across different types of situations in different countries. If results did converge between levels, this would help to build bridges between those who emphasize the fluidity of cultures and those who emphasize their relative continuity. The present study takes an initial step toward such a study by developing and testing nation-level hypotheses. As such, it is a test of nation-level hypotheses stimulated in part by Leung and Morris' model but not a direct test of their model.
Given the available range of nation-level behavioral indices, this study focuses on values and norms rather than schemata. Following Leung and Morris, we distinguish descriptive norms from injunctive norms as guides to one's behavior. We first discuss main effects attributable to values and to norms and then consider the moderating role of tightness-looseness.
Values versus Norms as Predictors
Values
While large numbers of cultural dimensions related to individualism and collectivism have been proposed, it is necessary for present purposes to select those that are unambiguously focused on values or on norms and for whom a maximal number of data points are available. In relation to values, Schwartz (Reference Schwartz and Zanna1992, Reference Schwartz, Vinken, Soeters and Ester2004) has established the reliability and validity of his measures across a broad range of cultures. At the nation level, bipolar dimensions have been identified, of which embeddedness versus autonomy most closely approximates to a conceptualization of collectivism versus individualism.
Morris (Reference Morris2014) has noted the ambiguity as to whether nation-level effects attributed to differences in values such as embeddedness more truly reflects expression of personal values or conformity to descriptive norms. This ambiguity can only be resolved through use of multi-level analyses. In the first of two studies, Vauclair and Fischer (Reference Vauclair and Fischer2011) tested the extent to which nation-level scores for autonomy-embeddedness could predict responses to different types of morally debatable behaviors. They employed Haidt's (Reference Haidt2008) contrast between ‘individualizing’ morality and ‘binding’ moralities. Individualizing morality emphasizes the rights of individuals to make their own choices in relation to issues such as divorce and abortion. Binding morality concerns the justice and fairness to others of actions such as cheating. Vauclair and Fischer analyzed responses to the Morally Debatable Behaviors Scale (Harding & Phillips, Reference Harding and Phillips1986). Data derived from the World Values Survey were available from 56 nations. A prior individual-level analysis yielded one factor for items relating to individualizing morality, and another factor relating to issues of binding morality. As they expected, it was found that nation-level autonomy values predicted endorsement of individualizing morality, but not binding morality.
One can speculate that those endorsing individualizing morality will be more reliant on descriptive norms and those endorsing binding morality will be more reliant on injunctive norms. However, it is preferable to test this relationship empirically. As a first step in this direction, Vauclair et al. (Reference Vauclair, Fischer, Ferreira, Guerra, Hossler, Karabati, de Carvalho Filho, Porto, Reyes, Ritkonen and Spiess2015) compared individual-level and nation-level predictors of morality judgments based on personal values with predictors based on perceived injunctive norms (ratings of what people are ‘. . .expected to strive for in my society’). Across eight nations, the personal measure of autonomy-embeddedness predicted endorsement of individualizing morality at both the individual and the nation level, but the normatively phrased version showed significant effects only at the nation level. Thus, we have evidence that personal values, descriptive and injunctive norms are each predictive under different circumstances.
The distinction between individualizing and binding moralities is also useful within the present study. Among the two dependent measures to be introduced in the next section, helping strangers is more relevant to individualizing morality, whereas emotion regulation has more relevance to binding morality.
Norms
The GLOBE research project (House et al., Reference House, Hanges, Javidan, Dorfman, Gupta and associates2004) endeavored to distinguish different psychological components of cultural dimensions such as individualism-collectivism by surveying managers in 61 nations. Their ‘should be’ questions asked how much members of society should engage in various actions, which subsequent researchers see as tapping not values but perceived injunctive norms (Hofstede, Reference Hofstede2006; Taras, Steel, & Kirkman, Reference Taras, Kirkman and Steel2010). Their ‘as is’ questions asked how much members of their society actually engage in various actions, which is clearly a measure of perceived descriptive norms. The GLOBE researchers did not present individual-level analyses of their data, so it is not possible to determine the degree of within-nation consensus concerning either of these sets of norms. However, using a similar measure of perceived descriptive norms across 11 nations, Fischer et al. (Reference Fischer, Ferreira, Assmar, Redford, Harb, Glazer, Cheng, Jiang, Wong, Kumar, Kärtner, Hofer and Achoui2009) found an overall intra-class correlation of 0.13, with substantial variation around the mean on items related to individualism. Thus, nation-level variance in descriptive norms is probably of similar magnitude to the nation-level variance in values that has been found in earlier studies (Fischer & Schwartz, Reference Fischer and Schwartz2011). This is an adequate level of within-nation consensus to permit meaningful comparisons of the nation-level predictiveness of embeddedness values, descriptive norms for collectivism and injunctive norms for collectivism.
Behavioral Dependent Measures
To compare the predictive abilities of nation-level indices for values and norms, we require relevant behavioral indices. The 87 nation-level main effects for individualism-collectivism that were included in the meta-analysis by Taras, Kirkman, and Steel (Reference Taras, Kirkman and Steel2010) were predominantly focused on economic indices, life satisfaction, and personality traits. Many of these effects referred to differences across a small sample of nations. It is preferable to sample widely and to identify specific behavioral criteria that can be more directly linked with their potential relevance to values and norms. Two such criteria are included within the present study: helping strangers and emotion regulation.
Helping strangers
Factors relating to the circumstances within which individuals provide assistance to others who are previously unknown to them have been extensively investigated, both in short-term emergency settings (Dovidio, Piliavin, Schroeder, & Penner, Reference Dovidio, Piliavin and Penner2006) and on a longer-term basis (Omoto & Snyder, 2010). Longer-term helping is found to depend particularly on aspects of interpersonal and intergroup relations and is thus relevant to the current focus on norms and values. A critical determinant of whom individuals are willing to help is whether the cultural context is individualistic or collectivistic. For instance, collectivists are found to distribute rewards more strongly in favor of in-group members than out-group persons, whereas individualists differentiate less strongly (Leung & Bond, Reference Leung and Bond1984). Here we are concerned with behaviors toward out-group persons.
The incidence of helping strangers across nations has been rarely studied until recently. In an early experimental study, Levine, Norenzayan, and Philbrick (Reference Levine, Norenzayan and Philbrick2001) staged simulated minor emergencies in public places in 23 nations and compared the frequency of helping behaviors from passersby. This is a measure of helping out-group members. Knafo, Schwartz, and Levine (Reference Knafo, Schwartz and Levine2009) found that helping rates in Levine et al.’s study were negatively correlated with Schwartz's embeddedness dimension of collectivist values and positively correlated with national wealth.
By drawing on data from the World Values Survey, Welzel (Reference Welzel2010) measured willingness to volunteer for political causes within 48 nations. Respondents indicated whether they had or would be willing to sign a petition, to join a boycott, and to join a peaceful demonstration. These actions provide instances of pro-social behavior, but it is less clear whether they occur predominantly in in-group or out-group contexts. Nations scoring high on an index of these behaviors were found to be those with high involvement in voluntary groups, high generalized trust, and high endorsement of the individualistic value of self-expression.
Further enlarging the range of sampled nations, Smith (Reference Smith2015a) analyzed the reported incidence of three specific helping behaviors across 135 nations, using data from representative national samples originally collected by the Gallup Organization. Respondents were asked whether in the last month they had helped a stranger or someone they did not know, whether they had volunteered time to an organization, and whether they had donated money to a charity. Across the overall sample, the mean rate of volunteering was 21 percent, mean donating was 28 percent, and mean helping a stranger was 49 percent. Nation-level mean responses to these items yielded a reliable index of pro-social behavior. Consistent with the earlier results, nations scoring high on this index were significantly higher on wealth and generalized trust and lower on embeddedness values. However, the strongest predictors were low in-group favoritism, another marker for collectivism values, as well as low scores for the orthogonal value of uncertainty avoidance.
Emotion regulation
While there is extensive evidence for the universality of basic emotions (Matsumoto & Hwang, Reference Matsumoto and Hwang2012), cultures differ in how much these emotions are overtly expressed in the context of particular kinds of relationships. Members of collectivist cultures differentiate more strongly between in-group and out-group relationships (e.g., Chen, Brockner, & Katz, Reference Chen, Brockner and Katz1996; English & Chen, Reference English and Chen2007) and therefore have stronger reasons to differentiate which emotions are best expressed or suppressed within each type of relationship context.
The most extensive source for data concerning cultural variations in emotional expressiveness is provided by the research of Matsumoto et al. (Reference Matsumoto, Yoo, Fontaine, Anguas-Wong, Arriola, Ataca, Bond, Boratav, Breugelmans, Cabecinhas, Chae, Chin, Comunian, Degere, Djunaidi, Fok, Friedlmeier, Ghosh, Glamcevski, Granskaya, Groenvynck, Harb, Haron, Joshi, Kakai, Kashima, Khan, Kurman, Kwantes, Mahmud, Mandaric, Nizharadze, Odusanya, Ostrosky-Solis, Palaniappan, Papastylianou, Safdar, Setiono, Shigemasu, Singelis, Solcova Iva, Spiess, Sterkowicz, Sunar, Szarota, Vishnivetz, Vohra, Ward, Wong, Wu, Zebian, Zengaya, Altarriba, Bauer, Mogaji, Siddiqui, Fulop, Garcia Bley, Alexandre, Garcia and Grossi2008), who surveyed students in 32 nations. Respondents were asked how they should behave if they felt each of seven emotions in the presence of each of 21 different persons, including family members, friends, and those known less personally. Half the responses referred to emotional expression in public places and half referred to expression in private places. Overall behavioral expressiveness was significantly related to Hofstede's (Reference Hofstede2001) scores for individualism, but the results in relation to each of the seven emotions varied substantially, and the overall effects were explicable largely in terms of expressions of happiness and surprise.
Of particular relevance to the present study are the data provided by Matsumoto et al. concerning the extent to which respondents reported that they should regulate their emotional expression dependent on the target person and setting. Matsumoto et al. did not in fact find that overall emotional regulation was associated with Hofstede scores for collectivism, perhaps because the Hofstede scores that they employed involved values rather than norms. Using a measure that has more specific behavioral implications than Hofstede's broadly defined values measure, Schug, Yoo, and Atreya (Reference Schug, Yoo and Atreya2017) found that, across 32 nations, generalized trust in others correlated at −0.49 (p = 0.004) with Matsumoto et al.’s overall measure of emotional regulation. Where there is a generalized trust in others, there is less need to differentiate one's particularized trust in in-group members from caution toward out-group members. Emotional regulation toward in- and out-group members can be relatively similar. Generalized trust is known to be stronger within individualist nations (Allik & Realo, Reference Allik and Realo2004), so the level of emotion regulation should be lower.
Based on prior studies of individualistic and collectivistic cultures (Hofstede, Reference Hofstede2001; Smith, Reference Smith2015a; Taras, Kirkman, & Steel, Reference Taras, Kirkman and Steel2010), we may thus expect helping behaviors to be more frequent in individualistic cultures and emotion regulation to be more frequent in collectivistic cultures. Hypotheses can be stated for country level dimensions that involve values, descriptive norms and injunctive norms related to individualism and collectivism. More specifically:
Hypothesis 1: Nations high on embeddedness values will show: (a) less frequent helping behaviors, and (b) more emotion regulation.
Hypothesis 2: Nations high on descriptive norms for collectivism practices will show (a) less frequent helping behaviors and (b) more emotion regulation.
Hypothesis 3: Nations high on injunctive norms for collectivism practices will show (a) less frequent helping behaviors and (b) more emotion regulation.
Tightness-Looseness
As noted earlier, Leung and Morris (Reference Leung and Morris2015) identified nation-level tightness as a moderator of the role of norms versus values. In a tight culture, we may expect norms to be more prescriptive and the consequences of deviating from those norms to be more hazardous. In a loose culture, individuals have greater choice and thus personal values will be more salient.
Two contrasting approaches to operationalizing tightness-looseness have been introduced recently. Gelfand et al. (Reference Gelfand, Raver, Nishii, Leslie, Lun, Lim, Duan, Almaliach, Ang, Arnadottir, Aycan, Boehnke, Boski, Cabecinhas, Chan, Chhokar, D'Amato, Ferrer, Fischlmayr, Fischer, Fülöp, Georgas, Kashima, Kashima, Kim, Lempereur, Marquez, Othman, Overlaet, Panagiotopoulou, Peltzer, Perez-Florizno, Ponomarenko, Realo, Schei, Schmitt, Smith, Soomro, Szabo, Taveesin, Toyama, Van de Vliert, Vohra, Ward and Yamaguchi2011) first developed a 6-item measure of the perceived tightness of one's national culture. They then asked respondents to rate the appropriateness of various behaviors in different everyday settings. Across 33 nations, they found that the extent to which student respondents perceived social relations to be tightly bounded was associated with specific expectations of how one should behave in each of 15 settings such as a library, restaurant or park. Tightness could therefore be operationalized as perceived situational constraint. These typical patterns of constraint are not just information that sometimes guides people but are also injunctively binding. Gelfand et al. showed that nations scoring high on tightness are subject to higher rates of various kinds of natural hazards, which may have led over time to the institutionalization of situational constraint.
In contrast, Uz (Reference Uz2015) proposed that tightness-looseness can be represented simply by the average variance within a population across a representative set of measures. Where there is less variability of values, beliefs, attitudes, or behaviors, there will be a greater likelihood that descriptive norms will provide a unified template for individual actions. Uz constructed three measures of this type for 68 nations. The first of these the (domain-general) Cultural Tightness-Looseness index (CTL-DG) is based upon responses to 124 items within the World Values Survey. Her second (domain-specific) measure (CTL-DS) is based on variance in responses to five of the ten items of the Morally Debatable Behaviors Scale that are included within the World Values Survey. Her third ‘combination’ measure (CTL-C) is based on 46 items relating to work, family and religion. Uz reports that none of these measures correlates significantly with Gelfand et al.’s (Reference Gelfand, Raver, Nishii, Leslie, Lun, Lim, Duan, Almaliach, Ang, Arnadottir, Aycan, Boehnke, Boski, Cabecinhas, Chan, Chhokar, D'Amato, Ferrer, Fischlmayr, Fischer, Fülöp, Georgas, Kashima, Kashima, Kim, Lempereur, Marquez, Othman, Overlaet, Panagiotopoulou, Peltzer, Perez-Florizno, Ponomarenko, Realo, Schei, Schmitt, Smith, Soomro, Szabo, Taveesin, Toyama, Van de Vliert, Vohra, Ward and Yamaguchi2011) 6-item measure of perceived tightness.
Though these past results suggest divergence between estimates of tightness-looseness based on normative expectations versus those based on descriptive variation, this may be partly attributable to differential sampling. Furthermore, there is actually some convergence between some of the measures. Although there is no relation between Gelfand et al.’s (Reference Gelfand, Raver, Nishii, Leslie, Lun, Lim, Duan, Almaliach, Ang, Arnadottir, Aycan, Boehnke, Boski, Cabecinhas, Chan, Chhokar, D'Amato, Ferrer, Fischlmayr, Fischer, Fülöp, Georgas, Kashima, Kashima, Kim, Lempereur, Marquez, Othman, Overlaet, Panagiotopoulou, Peltzer, Perez-Florizno, Ponomarenko, Realo, Schei, Schmitt, Smith, Soomro, Szabo, Taveesin, Toyama, Van de Vliert, Vohra, Ward and Yamaguchi2011) perceived tightness and Uz's (Reference Uz2015) tightness measures, there is in fact a significant relationship between Uz's CTL-DS tightness measure and Gelfand et al.’s measure of situational constraint. For the present, we employ principally the measures provided by Uz, since her sampling was more extensive, although the hypotheses are also tested using Gelfand et al.’s measure. The differing bases of the two measures also suggest that the results may vary when using the Uz predictor and when using the Gelfand et al. measure. However, we cannot test such differential effects with the present data, because the samples using each measure vary in extent and coverage. As discussed previously, it can be expected that tightness-looseness will interact with the effects foreseen in Hypotheses 1 to 3, because tightness will increase dependence on norms, whereas looseness will reduce their salience:
Hypothesis 4: In contexts of low rather than high tightness, embeddedness values will be (a) more strongly predictive of helping behaviors and (b) less strongly predictive of emotion regulation.
Hypothesis 5: In contexts of high rather than low tightness, collectivism descriptive norms will be (a) less strongly predictive of helping behaviors and (b) more strongly predictive of emotion regulation.
Hypothesis 6: In contexts of high rather than low tightness, collectivism injunctive norms will be (a) less strongly predictive of helping behaviors and (b) more strongly predictive of emotion regulation.
METHOD
Independent variables
The hypotheses were tested through secondary data analyses. Two measures of values were used. Scores for embeddedness versus autonomy are derived from Schwartz Values Survey data from 74 nations provided by Shalom Schwartz. These scores are based on pooling of data from student and teacher respondents. For each nation, the embeddedness score was subtracted from the autonomy score. In order to test for specificity of effects using Schwartz values, a second measure for values was included, namely the Welzel autonomy index, taken from the World Values Survey database (worldvaluessurvey.org). This is based on responses to three survey items tapping preferences that one's child shall be independent and imaginative but not obedient. To render this index comparable with the other collectivism measures, scores were reversed and renamed as an index of obedience. Scores are available from representative samples drawn from 98 nations.
Three normative measures of collectivism were also included. The measures for descriptive and injunctive norms respectively are House et al.’s (Reference House, Hanges, Javidan, Dorfman, Gupta and associates2004) societal scores for in-group collectivism ‘as is’ and ‘should be’ for 61 nations. House et al.’s in-group collectivism dimension was selected for use because the items defining this dimension refer more directly to individual behaviors than do the items defining their institutional collectivism dimension, and thus come closer to the focus of the dependent measures in the present study. These scores are based on responses by business managers.
Dependent measures
The measures for helping behaviors are drawn from the World Giving Index 2015, a publicly available data source provided by the Charities Aid Foundation, based in the UK (www.cafonline.org/publications/2015-publications/world-giving-index-2015.aspx). These data are drawn from the 2014 wave of the surveys of nationally representative samples within up to 160 nations that are conducted annually by the Gallup Organization. The specific survey question of interest is, ‘During the past month have you done any of the following? – donated money to a charity? – volunteered your time to an organization? – helped a stranger or someone you didn't know who needed help? Respondents answer simply yes or no to each of these items. The single-item nature of these questions yields no further information as to the various possible types of donation, of volunteering, or of helping that respondents have in mind when answering. The score for each nation is the percentage of affirmative responses. Nation-level Cronbach alpha for the three helping behavior measures was 0.65. A single helping behavior index based on standardized scores was therefore constructed.
The dependent measure for emotional regulation is taken from the listing of intrapersonal variability norms for 32 nations provided by Matsumoto et al. (Reference Matsumoto, Yoo, Fontaine, Anguas-Wong, Arriola, Ataca, Bond, Boratav, Breugelmans, Cabecinhas, Chae, Chin, Comunian, Degere, Djunaidi, Fok, Friedlmeier, Ghosh, Glamcevski, Granskaya, Groenvynck, Harb, Haron, Joshi, Kakai, Kashima, Khan, Kurman, Kwantes, Mahmud, Mandaric, Nizharadze, Odusanya, Ostrosky-Solis, Palaniappan, Papastylianou, Safdar, Setiono, Shigemasu, Singelis, Solcova Iva, Spiess, Sterkowicz, Sunar, Szarota, Vishnivetz, Vohra, Ward, Wong, Wu, Zebian, Zengaya, Altarriba, Bauer, Mogaji, Siddiqui, Fulop, Garcia Bley, Alexandre, Garcia and Grossi2008). Students were asked what they should do when they experienced each of seven emotions in the presence of each of 21 persons including age- and gender-defined family members, friends, acquaintances, classmates and professors, in either a private or a public setting. The response scale specified seven options, ranging from ‘show more than you feel’ to ‘neutralize’. The published scores are for the standard deviation in emotional expressiveness across all emotions, persons and settings.
Moderators
The principal tightness-looseness measures are those provided by Uz (Reference Uz2015). The data are based upon responses by representative samples in 68 nations to the 2000 wave of the World Values Survey. Her CTL-DS measure based on within-nation variance on the five MDBS items is preferable because it is available for slightly more nations, because the content of the items defining it are more precisely specified, and especially because they are phrased injunctively. However, her CTL-C measure has the alternative virtue of being based on a much broader range of items. The hypotheses are tested twice, using these two alternative indices. Her three measures correlate at between 0.74 and 0.85 with one another across 65 nations. It should be noted that although Uz (Reference Uz2015) describes her measure as tapping tightness-looseness, it is looseness that scores high on her scales. For the present analysis these scores were reversed, so that tightness scores high.
The measure of situational constraint is described in the supplementary materials to the paper of Gelfand et al. (Reference Gelfand, Raver, Nishii, Leslie, Lun, Lim, Duan, Almaliach, Ang, Arnadottir, Aycan, Boehnke, Boski, Cabecinhas, Chan, Chhokar, D'Amato, Ferrer, Fischlmayr, Fischer, Fülöp, Georgas, Kashima, Kashima, Kim, Lempereur, Marquez, Othman, Overlaet, Panagiotopoulou, Peltzer, Perez-Florizno, Ponomarenko, Realo, Schei, Schmitt, Smith, Soomro, Szabo, Taveesin, Toyama, Van de Vliert, Vohra, Ward and Yamaguchi2011). Scores for 33 nations were provided by Michele Gelfand. Scores are highest for nations that are low on situational constraint, so these scores were also reversed, yielding high scores for tightness.
Controls
Data for nation-level wealth per capita based on purchasing power parity for the year 2015 were downloaded from www.worldbank.org. To discount the skewed distribution of means a log transform was employed.
RESULTS
Means and standard deviations for all variables as well as their correlations with one another are shown in Table 1. There is a strong correlation between the embeddedness and obedience values measures, as well as a positive correlation with collectivism descriptive norms. Furthermore, in most cases these measures also correlate substantially with the three tightness-looseness indicators. The scores for injunctive collectivism norms are independent of all other measures.
Table 1. Means, standard deviations, and correlations for all variables
Notes: n = 31 (Emotion Regulation); n = 22–31 for Tightness (Gelfand); n = 34–50 for Tightness (Uz); n = 50–74 for all other variables. *p < 0.05; **p < 0.01; ***p < 0.001.
The hypotheses were tested through a series of hierarchical regressions, with main effects entered in Model 1 and interaction terms added in Model 2. In all analyses, national wealth based on purchasing power parity was controlled when assessing the predictiveness of values or norms. Interaction terms were computed using standard scores, and their effect is evaluated in terms of R2 change. The upper part of Table 2 shows the results for embeddedness values in relation to Uz's (Reference Uz2015) two tightness measures. Contrary to hypotheses 1(a) and 1(b), there are no main effect relations between Schwartz's embeddedness and the dependent measures. However, supporting hypothesis 4(a), low tightness significantly affects the relation of embeddedness with helping when CTL-DS is used as the moderator. This effect is shown as Figure 1. Simple slope analyses at one standard deviation above and below the mean show a significant negative slope for low tightness (b = −0.60; p = 0.02) and a non-significant positive slope for high tightness (b = 0.27; p = 0.55).
Table 2. Hierarchical regressions showing values as predictors of helping and emotion regulation, with Uz's tightness measures as moderators
Figure 1. Interaction between Schwartz embeddedness values and tightness (CTL-DS) as predictors of helping
Furthermore, supporting hypothesis 4(b), tightness does significantly affect the relation of embeddedness with emotion regulation. This effect is shown as Figure 2. Simple slope analyses show a non-significant negative slope (b = −0.87; p = 0.07) for high tightness and a non-significant positive slope for low tightness (b = 0.40; p = 0.10), with the contrast between the two slopes significant. With CTL-C as the looseness measure, only the effect for helping behavior is found.
Figure 2. Interaction between Schwartz embeddedness values and tightness (CTL-DS) as predictors of emotion regulation
The lower part of Table 2 shows a second set of tests of the same four hypotheses, now using the Welzel (Reference Welzel2010) obedience index as the predictor. With this index, there are again no significant main effects. Tightness again negatively moderates helping, supporting hypothesis 4(a), but only when CTL-DS is used as the moderator. The moderation of emotion regulation is not replicated.
Table 3 shows the results for the measures of descriptive and injunctive norms derived from the GLOBE project (House et al., Reference House, Hanges, Javidan, Dorfman, Gupta and associates2004). Descriptive norms for collectivism show significantly lower helping and higher emotional regulation, supporting hypotheses 2(a) and 2(b), but no significant interactions, thereby rejecting hypotheses 5(a) and 5(b). In contrast, injunctive norms show no main effects, rejecting hypotheses 3(a) and 3(b), but tightness significantly affects the relation of injunctive norms to emotion regulation, strongly with CTL-DS as predictor and marginally with CTL-C as predictor, supporting hypothesis 6(b). The effect when using CTL-DS as predictor is shown as Figure 3. Simple slope analyses show a significant positive slope (b = 0.89; p = 0.004) for high tightness and a non-significant negative slope for low tightness (b = −0.39; p = 0.39).
Table 3. Descriptive and injunctive norms as predictors of helping and emotion regulation, with Uz's tightness measures as moderators
Figure 3. Interaction between collectivism injunctive norms and tightness- (CTL-DS) as predictors of emotion regulation
Within the more restricted range of nations for which data are available, the hypotheses are tested for a third time, using Gelfand et al.’s (Reference Gelfand, Raver, Nishii, Leslie, Lun, Lim, Duan, Almaliach, Ang, Arnadottir, Aycan, Boehnke, Boski, Cabecinhas, Chan, Chhokar, D'Amato, Ferrer, Fischlmayr, Fischer, Fülöp, Georgas, Kashima, Kashima, Kim, Lempereur, Marquez, Othman, Overlaet, Panagiotopoulou, Peltzer, Perez-Florizno, Ponomarenko, Realo, Schei, Schmitt, Smith, Soomro, Szabo, Taveesin, Toyama, Van de Vliert, Vohra, Ward and Yamaguchi2011) tightness measure as the moderator. Table 4 shows no significant main effects for any of the predictors. However, tightness significantly moderates the relation between embeddedness and helping, supporting hypothesis 4(a), as well as the relation between collectivism descriptive norms and helping, supporting hypothesis 5(a). Tightness also shows marginally significant moderations of the relation of the same two predictors with emotional regulation, supporting hypotheses 4(b) and 5(b).
Table 4. Collectivism-related values as predictors of helping and emotion regulation, using Gelfand's measure of tightness as moderator
DISCUSSION
The major emphasis of this study has been upon the importance of distinguishing values from norms as predictors of cultural differences. Nation-level measures related to variations in individualism-collectivism and similar concepts have often been treated as generic, without due attention to the precise ways in which they have been measured. This conflation is understandable, given the convergence between the scores on different measures, as illustrated by the correlation of 0.65 between embeddedness values and collectivism descriptive norms in the present analysis. Despite this convergence, the results indicate that distinctive effects are obtained when precisely defined dependent measures are employed across an adequate range of samples.
The values-based measures have relevance to both dependent measures, but are significantly predictive only when moderated by tightness-looseness. Using Uz's measure, embeddedness values predict low levels of helping in the context of loose cultures, and high emotion regulation in the context of tight cultures. These effects are strongest when using CTL-DS as the looseness measure, and when using Schwartz's measure rather than Welzel's measure of obedience. In the key interaction shown as Figure 1, it is the slope for low tightness that achieves significance. This is consistent with the likelihood that helping of strangers occurs in unconstrained and ambiguous contexts where values can guide one's actions.
In contrast, in Figure 2 there are trends for both high and low tightness, which together achieve significance. Emotion regulation is maximised both by high tightness and by high embeddedness. This suggests two possible bases for high emotion regulation: contexts in which there is the predicted simple effect of high mean endorsement of embeddedness values, and an additional simple effect of contexts in which there is low variability of relevant values.
The collectivism descriptive norms measure shows main effects for both kinds of dependent measure, but these effects are only moderated by tightness when operationalized as Gelfand et al.’s situational constraint measure. This raises the question as to why interactions with tightness are obtained with values but not with descriptive norms. One possibility is that the items used by House et al. (Reference House, Hanges, Javidan, Dorfman, Gupta and associates2004) to define collectivism descriptive norms focused on family relationships and may already implicitly convey tightness rather than looseness. The items in question refer to ‘whether children take pride in the individual accomplishments of their parents and vice versa [these two items reversed], whether aging parents live at home with their children, and whether children live at home with their parents until they get married’ (463). This might explain why tightness did not moderate the effect of collectivism descriptive norms.
Finally, the injunctive norms measure is unrelated to helping, but predicts stronger emotional regulation in tight cultures. This effect is marked when using CTL-DS as the moderator, but there is also a trend in the same direction with CTL-C. The significant slope in Figure 3 is for high tightness, suggesting that tightness and injunctive norms together augment emotion regulation. This contrasts with the suggestion in the preceding paragraph that tightness may provide a basis for emotion regulation independently of prevailing values.
In evaluating the validity of the overall results, it is first necessary to consider the extent to which they may have been affected by multicollinearity between the predictors. The key effects presented as Figures 1 and 2 concern embeddedness and CTL-DS, and these variables are shown in Table 1 to correlate strongly with one another and with nation-level wealth. Following normal practice for nation-level analyses, the effects of wealth were controlled in the regression analyses, and this may have served to reduce the multicollinearity of embeddedness and CTL-DS. The partial correlation between these variables with wealth controlled is 0.44, substantially reduced from the figure of 0.64 shown in Table 1. The correlation between injunctive norms and CTL-DS is much lower, so the result shown in Figure 3 is less ambiguous. There is no clear consensus as to the level at which variance inflation factors become problematic in interpreting effects. The mean VIF levels shown in the three key effects illustrated in the figures were between 1.7 and 2.0, suggesting that there is a modest rather than a severe effect of multicollinearity.
Values, Norms, and Tightness
I next discuss in turn the substantive issues addressed in this article and then the significance and limitations of the results for the dependent measures. The contrasting results for the different predictors challenge us to further clarify what we mean when we define large entities such as nations in terms of values and norms aggregated from individual-level data. Does an awareness of which values are shared within one's group provide a basis that could be considered to be a descriptive norm? Does an awareness of normal behaviors within one's group imply that there is also an implicit sharing of values that are relevant to those behaviors? Schwartz (Reference Schwartz2011) argues that cultural values are ‘the normative value emphases that underlie and justify the functioning of societal institutions’ (314). He further states that his ‘. . . conception of cultural values as normative value emphases that underlie societal functioning does not assume substantial sharing of values by individuals. Societal means obtained by aggregating individual responses point indirectly to the underlying cultural value emphases’ (314).
Schwartz's conceptualisation of the dimension of autonomy versus embeddedness also renders it distinctively relevant, since his items defining embeddedness imply reliance on a specific set of norms, whereas those defining autonomy imply a degree of choice as to how much one relies on any of a range of available norms. Based on the preceding discussion of different types of norms, it could be better to say that the permanence and centrality of one's group membership that defines embeddedness implies greater reliance on injunctive norms, whereas the relative impermanence defining autonomy implies greater reliance on descriptive norms. Nonetheless, the present results indicate that effects associated with aggregated personal values cannot be equated with effects associated with descriptive norms, at least as they were measured by House et al. (Reference House, Hanges, Javidan, Dorfman, Gupta and associates2004). In terms of main effects, descriptive norms yield more consistent results.
However, we must also consider the relevance of tightness-looseness. Uz's (Reference Uz2015) CTL-DS index is based on values items, whereas her CTL-C index is based on items referring to a broad mix of values, attitudes and behaviors. Conversely, Gelfand et al.’s (Reference Gelfand, Raver, Nishii, Leslie, Lun, Lim, Duan, Almaliach, Ang, Arnadottir, Aycan, Boehnke, Boski, Cabecinhas, Chan, Chhokar, D'Amato, Ferrer, Fischlmayr, Fischer, Fülöp, Georgas, Kashima, Kashima, Kim, Lempereur, Marquez, Othman, Overlaet, Panagiotopoulou, Peltzer, Perez-Florizno, Ponomarenko, Realo, Schei, Schmitt, Smith, Soomro, Szabo, Taveesin, Toyama, Van de Vliert, Vohra, Ward and Yamaguchi2011) tightness measure is based on items specifying the appropriateness of behaviors. Despite their different bases, all three measures correlated with one another and with many of the predictors.
Uz (Reference Uz2015) discusses threats to the validity of measures of looseness-tightness due to cultural variations in response style. While response styles are typically monitored as intrapersonal variations in acquiescence or extremity across items, looseness-tightness indices are based on variability across persons in response to the items. Since response styles are known to vary with some consistency between national cultures (Smith, Reference Smith2004, Reference Smith2011), it is likely that tight cultures are also acquiescent cultures. However, the finding of differences between the results using predictors based on values and on norms, as well as the effects found with interaction terms provide some assurance that we are not simply studying variations in response style.
The various tightness indices also yielded a similar pattern of effects, namely moderation effects with embeddedness and fewer moderations with collectivism descriptive norms. The results are stronger with the Uz than the Gelfand measure, most likely because of the larger sample of nations for which data are available. Thus, contrasting tight versus loose cultures does, as Leung and Morris postulated, strengthen our understanding of the contexts in which values guide behavior. However, the contrast appears less useful in relation to descriptive norms. The significance of descriptive norms may vary, dependent on whether they are interpreted in a context that is tight or loose. With increasing tightness, descriptive norms would more likely be interpreted as injunctive. Some support for this view was found in the analyses using Gelfand's measure, which is the one with the more strongly injunctive emphasis.
The results obtained are dependent upon the precision and validity with which the predictors and dependent measures have been defined. While the cross-cultural measurement of values has been extensively explored (Schwartz, Reference Schwartz, Vinken, Soeters and Ester2004, Reference Schwartz2011), delineation of norms has hitherto received less detailed attention (Gelfand et al., Reference Gelfand, Raver, Nishii, Leslie, Lun, Lim, Duan, Almaliach, Ang, Arnadottir, Aycan, Boehnke, Boski, Cabecinhas, Chan, Chhokar, D'Amato, Ferrer, Fischlmayr, Fischer, Fülöp, Georgas, Kashima, Kashima, Kim, Lempereur, Marquez, Othman, Overlaet, Panagiotopoulou, Peltzer, Perez-Florizno, Ponomarenko, Realo, Schei, Schmitt, Smith, Soomro, Szabo, Taveesin, Toyama, Van de Vliert, Vohra, Ward and Yamaguchi2011; Gelfand & Harrington, Reference Gelfand and Harrington2015; Shteynberg, Gelfand, & Kim, Reference Shteynberg, Gelfand and Kim2009; Smith, Reference Smith2015b). Moderations may also only be found where tightness and collectivism have differential implications. The high variance inflation factor scores found in the analyses for collectivism descriptive norms no doubt reflect the strong correlations in the present sample shown in Table 1 between the tightness measures and collectivism descriptive norms. Gelfand et al. (Reference Gelfand, Raver, Nishii, Leslie, Lun, Lim, Duan, Almaliach, Ang, Arnadottir, Aycan, Boehnke, Boski, Cabecinhas, Chan, Chhokar, D'Amato, Ferrer, Fischlmayr, Fischer, Fülöp, Georgas, Kashima, Kashima, Kim, Lempereur, Marquez, Othman, Overlaet, Panagiotopoulou, Peltzer, Perez-Florizno, Ponomarenko, Realo, Schei, Schmitt, Smith, Soomro, Szabo, Taveesin, Toyama, Van de Vliert, Vohra, Ward and Yamaguchi2011) found collectivism and their perceived tightness measure to correlate at just 0.49, so there would be some locations in which collectivism norms and tightness would not be coterminous.
The broader range of tightness measures included in the present study enables a fuller examination of the extent to which tightness and collectivism are associated with one another. The measures of tightness and injunctive norms are clearly independent, but there are strong associations of tightness with embeddedness and in most cases with descriptive norms, but not with obedience. The substantially lower correlation of 0.38 between the more specifically focused CTL-DS measure and collectivism descriptive norms suggests that there may be aspects of values and behavior in which tightness and collectivism are more clearly differentiated. Conversely, where values and norms favor conformity and conventionality this may also imply tightness.
Variation in the association between tightness and collectivism might account for the results of Aktas, Gelfand, and Hanges (Reference Aktas, Gelfand and Hanges2016) who showed across 29 nations that although team-oriented leaders were perceived as significantly more effective in nations high on collectivism descriptive norms, this effect was significantly weaker in tight cultures. In contrast, these authors found no moderation of effects by tightness-looseness in relation to two other styles of leadership for which collectivism descriptive norms were also a significant predictor.
The lesser number of effects for collectivism descriptive norms may also indicate that tightness moderates reliance on descriptive norms less than injunctive norms. Adherence to injunctive norms is where tightness should matter most. Consistent with this, the present results with injunctive norms did show a strong moderation in relation to emotion regulation. Greater attention to the hitherto substantially neglected significance of injunctive norms in future studies would be beneficial (Chen & Hong, Reference Chen and Hong2015; Gelfand & Harrington, Reference Gelfand and Harrington2015).
Overall, it appears that several factors are involved in whether moderation effects are obtained. These include the precision with which predictors have been measured, the relevance of the predictors to the dependent measures, whether an adequate number of nations have been sampled, and the extent to which aggregated values can serve as descriptive norms. However, in support of the propositions advanced by Leung and Morris (Reference Leung and Morris2015), where moderations of values effects are found, they predominantly link more individualistic values with looseness and latitude. Equally, where moderations of norm effects are found they link collectivism with tightness and constraint.
These results leave open the question of how tightness and collectivism interact with one another in the formulation of norms. Uz's (Reference Uz2015) measures of tightness concern only the degree of variability within a population, whereas that of Gelfand et al. (Reference Gelfand, Raver, Nishii, Leslie, Lun, Lim, Duan, Almaliach, Ang, Arnadottir, Aycan, Boehnke, Boski, Cabecinhas, Chan, Chhokar, D'Amato, Ferrer, Fischlmayr, Fischer, Fülöp, Georgas, Kashima, Kashima, Kim, Lempereur, Marquez, Othman, Overlaet, Panagiotopoulou, Peltzer, Perez-Florizno, Ponomarenko, Realo, Schei, Schmitt, Smith, Soomro, Szabo, Taveesin, Toyama, Van de Vliert, Vohra, Ward and Yamaguchi2011) has a more injunctive element, focused on the extent to which it is appropriate to vary one's behavior in a given situation. Members of collectivist cultures vary their behavior more between settings, in order to act appropriately in each context (English & Chen, Reference English and Chen2007). Collectivism thus has elements of context-specific tightness combined with an acceptance that there is no requirement for cross-situational consistency. Descriptive norms are likely to be interpreted injunctively where they are salient to one's commitment to one's in-group, but other contexts will provide opportunities for individual or collective deviance (Chen & Hong, Reference Chen and Hong2015). To explore such issues, we shall require more specific measures of tightness.
Dependent Measures
Helping behaviors
Among the dependent measures, the index of helping has distinctive value due to its precisely-described behavioral content, and its derivation from representative sampling. However, we have no information on the ways in which respondents in different samples understood the meaning of helping, volunteering and donating, nor on the extent of these activities in which they engaged. The present study was based on the 2015 dataset, whereas the study by Smith (Reference Smith2015a) used the 2014 dataset. However, the equivalent indices from these datasets correlate with one another at between 0.84 and 0.89, so there is good evidence for their reliability.
The present results amplify the conclusions of earlier studies of helping behaviors (Smith, Reference Smith2015a; Welzel, Reference Welzel2010). These studies both showed generalized trust to be a significant predictor, but Smith found the strongest predictors to be low in-group favoritism and low uncertainty avoidance. There was no interaction between these two predictors. The measure of in-group favoritism was a composite of several indices, one of which was House et al.’s (Reference House, Hanges, Javidan, Dorfman, Gupta and associates2004) measure of collectivism, so this measure is in common between the earlier study and the present one. The variance attributable to Hofstede's (Reference Hofstede2001) measure of low and high uncertainty avoidance may be better captured by measures of tightness. Tightness can be defined empirically with greater precision and this could account for its being a significant moderator in the present study, particularly of embeddedness values. The contexts enhancing helping behaviors may thus be those in which generalized trust can be interpreted in terms of low embeddedness and looser norms.
Emotion regulation
The measure of emotion regulation derived from Matsumoto et al. (Reference Matsumoto, Yoo, Fontaine, Anguas-Wong, Arriola, Ataca, Bond, Boratav, Breugelmans, Cabecinhas, Chae, Chin, Comunian, Degere, Djunaidi, Fok, Friedlmeier, Ghosh, Glamcevski, Granskaya, Groenvynck, Harb, Haron, Joshi, Kakai, Kashima, Khan, Kurman, Kwantes, Mahmud, Mandaric, Nizharadze, Odusanya, Ostrosky-Solis, Palaniappan, Papastylianou, Safdar, Setiono, Shigemasu, Singelis, Solcova Iva, Spiess, Sterkowicz, Sunar, Szarota, Vishnivetz, Vohra, Ward, Wong, Wu, Zebian, Zengaya, Altarriba, Bauer, Mogaji, Siddiqui, Fulop, Garcia Bley, Alexandre, Garcia and Grossi2008) was focused on all types of emotional expression and on both public and private settings. This may have masked some of the cultural differences of interest, since there is likely to be greater differentiation between cultures in the expression of negative emotions than of positive emotions. The sample of nations for which this index was available was also more restricted. These factors would limit the likelihood of detecting significant effects. Despite this enhancement of possible type 2 errors, a significant interaction between values and tightness was found, and even within the smaller dataset, there was a strong effect for collectivism injunctive norms in relation to tightness.
CONCLUSION
The propositions advanced by Leung and Morris (Reference Leung and Morris2015) specify numerous factors that may enhance the relative predictive abilities of values and norms. These propositions require testing at several different levels of analysis. At the nation level and in relation to the contrast between tight and loose cultures, the two types of dependent variables that were selected show substantial support as regards aggregated values. Low embeddedness values more strongly predict helping behaviors in loose cultures, while high embeddedness values more strongly predict emotional regulation in tight cultures. However, effects are weaker for norms for collective practices and are only unambiguously supportive of the Leung and Morris model in relation to injunctive norms. The extent to which studies of descriptive norms can also guide our understanding of injunctive norms requires fuller investigation. Multi-level analyses featuring measures of values and norms will be required if we are to fully understand the extent to which their effects are moderated by nation-level tightness-looseness.
