1 Introduction

What marks what we do as cultural? Surely it is that human behaviour, whether innate reflex or practiced action, is identified and understood according to the customs of the society in which it occurs. That is, the ability to count a slice of behaviour as an instance of this or that folk category—and to understand its meaning according to the shared concerns, beliefs, and commitments that coordinate social activity—involves a good deal of customary knowledge. Seen in this light, the study of culture is the “science of custom” (Benedict 1934). There are, of course, other ways to define and understand culture, but the customs of a society will be central to any account.

In the case of action, or elective, reasoned behaviour, customary meaning involves ethical discriminations. Actions are often praised or censured as evidence of virtue or vice, and of the character of the actor. More broadly, our actions and those of others are intelligible to us only within a “moral topography” (Taylor 1989) of qualitative distinctions whereby we orient to what we believe is good, proper, or desirable. The consideration and planning that is constitutive of human agency across time occurs against the individual’s horizon of ultimate concerns and commitments (MacIntyre 1981). However unarticulated and intuitive, this horizon is a structural requirement for coherent action and self-interpretation. It makes possible what Taylor (1985) has called “strong evaluation,” the weighing of one good against others in deciding what to do in a given situation. All of this suggests an inescapable moral framework in which we chart the course of our lives, define who and what we are, and strive to understand the actions and identities of others. One axis of this framework concerns what makes life worth living, what is should be about, and how best to pursue this conception of the good in one’s own case. The fact that societies tend to differ normatively in this regard has been discussed in Western literature since Herodotus’s ethnographic digressions in the Histories. Perhaps the most wide-ranging survey of cultural diversity in conceptions of the good was offered by Westermarck (1912–1917), who argued against any objective basis for prescriptive ethics. According to him, “There can be no moral truth in the sense in which this term is generally understood. The ultimate reason for this is that the moral concepts are based upon emotions, and that the contents of an emotion fall entirely outside the category of truth.” (p. 17). The causes of cultural differences in moral judgment, according to Westermarck, lay in the economic conditions, physical and political challenges, demographics, and forms of communal organization that distinguish one society from another.

More recently, comparative research on values has sought to distinguish what is universal from what is culturally particular in the qualities that people deem important or take an interest in (Hofstede 2001; Inglehart 2006; Morris 1956; Rokeach 1973; Schwartz 1992, 2006). The investigation of “basic” values has contributed much to our understanding of social diversity and grown into a leading orientation in cross-cultural psychology. As Leung and Bond (2004; see also Bond et al. 2004) have noted, however, knowing an individual’s allegiance to generalized abstractions such as “tradition” or “equality” is insufficient for understanding the direction of their lives. Minimally, we must also know their abiding beliefs about how the world, especially the social world, operates. Behaving in consistency with a value requires translating a commitment, desire, or preference into a course of action that is founded on awareness of conditions, costs, and consequences (Dewey 1939). As individual actors, however, our knowledge of this constitutive context is often incomplete and uncertain. As a result, we tend to rely heavily upon core convictions about the enduring properties and dispositions of people, institutions, and things, the principles which for us define “the way things are.”

It is perhaps misleading to suggest that broad value conceptions translate directly into specific actions at all. For example, few members of most societies, if asked why they chose the fashion of their peer group over their own distinctive preference in a particular instance, would claim the prizing of “conformity” as the reason. Nearly all would mention instead concerns that are tied to the content and circumstances of the choice. Their responses would not reflect a lack of sincerity, but the reality that highly abstract values rarely serve as the proximate rational cause—in the form of conscious reasons or justifications—of a particular, concrete action. Rather, we come to recognize and reflect on our values and those of others in the patterning of conduct and emotional responses over time. In other words, generalized values are inferred. Likewise, a society’s values refer to the consistencies we perceive across its customs, institutions, and practices.

The ethical significance of values as guides to action is further complicated by the fact that values are not themselves precepts, or rules of conduct. To say that one values “security” in life says little about how this commitment applies to practical matters of deliberation and choice. How should security be sought? When and where is it of paramount importance and when and where of minor concern? How does one weigh one kind of security against another, and the security of others against one’s own? How should it be reconciled with other prudential concerns? And how does it subserve the realization of higher goods? That two individuals value security on the whole to the same extent (by some common measure) does not require that they value it in the same way. If these individuals belong to different cultural communities with distinct ways of life, there is every reason to expect that they do not. We are referring here to how an abstract value is expressed in the choices, purposes, interpretations, and reactions that constitute a unified life. Knowing how a value is to be lived requires knowing what life one is trying to live. This brings us back to the teleological significance of cultural conceptions of the good life as an axis of moral orientation. Such holistic, orientative conceptions cannot be reconstructed from a constellation of basic values. To the contrary, the manner in which those basic values are manifest over the course of a lifetime will necessarily reflect the ideals against which a life is organized and measured, and toward which it is directed (Freeman and Brockmeier 2001). This recommends the importance of directly examining these ideals, the criteria by which people assess whether their lives are unfolding or have unfolded in the “right” way.

The foregoing suggests that how a society defines a good or worthy life, as reflected in the acquired ethical commitments of its members, is crucial for understanding the internal meaning of its cultural practices and traditions. This includes, most centrally, what is important or desirable about these practices and traditions, and the human virtues or excellences they foster (MacIntyre 1981; McDowell 1998). Comparing societies on this score has a contemporary press. Our age of globalization has been marked by the increasingly borderless reproduction and integration of political, economic, and cultural forms around the world. This integration has given birth to new kinds of proximity, connection, and commonality, and produced rapid rates of ideological diffusion and synthesis (Held and McGrew 2007). Globalization raises the question of just how similar we have become across once-distant cultures in our moral horizons, at least in regard to what we aim at and hope for in our lives. The answer to this question has obvious implications for the viability of a universal ethics and the possibility of a truly cosmopolitan social democracy (Appiah 2006; Harvey 2009; Nussbaum 1997). The forces of globalization in various cultural spheres do not simply homogenize societies, but also produce distinctive tensions, asymmetries, and “disjunctures” (Appadurai 1996) around the world—generating new differences and well as new similarities. This complexity recommends that we look closely at what individuals in different societies think is important in life, both as a reflection of convergence due to global cultural flows and as distinct, indigenized responses to those flows.

With the above concerns in mind, the present study was conceived as an initial, exploratory step toward investigating variability in conceptions of the good life across societies. At this initial stage, we chose to examine students in Canada, China, India, and Japan. Cross-cultural comparison of individuals from these four nations—two developed and two developing economies—over the past three decades has revealed a wide array of attitudinal differences. Thus, they offer a promising starting point for revealing and interpreting alternative moral horizons. Our aim was to highlight the significance of intraculturally shared commitments, rather than demographic and socioeconomic standing, for explaining any response differences across groups. We therefore focused exclusively on students at competitive national universities, effectively studying the same social class or class fraction across countries. This minimized the confounding influence of literacy, education, wealth, exposure to technology, and other facets of socioeconomic position, which was further controlled for statistically. Of course, studying university students to examine cultural diversity is a bit like exploring ethnic cuisine by dining at different McDonald’s outlets around the world. The strategically indigenized menus will exhibit far more uniformity of content than difference. Similarly, high-achieving university students tend to be highly “globalized” in their mentalities and practices, and are therefore more alike around the world than most other demographic groups. One could even argue that their closely shared way of life and prospects argues against the claim that they occupy clearly distinct “cultural worlds.” Even so, the tradition of comparing student samples from the nations represented in this study has produced no shortage of marked differences in thought, motivation, and behavior over the years. We were guided by this precedent in examining cultural differences here.

It should be noted that the four-way comparison was not conducted to confirm specific pairwise differences. Such hypotheses would have been premature, given that the substantive criteria of a good life—on which quantitative comparisons were to be conducted—could not be defined a priori in an exploratory study designed to reveal them. More generally, however, we did predict that any differences across nations in what accounts for a good life would not be redundant with differences in basic values. This reflects our theoretical claims about the nonderivativeness of narrative conceptions of a good life.

How, then, does one go about studying the teleological background of an individual’s moral intuitions? As we learned from our pilot research, asking university students (excepting philosophy majors, perhaps) to articulate the conception of a good life that gives shape and direction to their everyday aims and actions most often results in a long and embarrassed silence. In making qualitative moral distinctions, most people are guided by tacit and unquestioned commitments, not ideological manifestoes (Bourdieu 1990; Taylor 1989). The former are notoriously difficult to articulate, as explication often invites critical evaluation and the burden of justification. To overcome this problem, we used an indirect, assisted form of elicitation. We first instructed students to imagine themselves near the end of their lives and then asked them what they would point to as (material or immaterial) evidence that they had lived a worthy or good life according to their own beliefs and commitments. The strategy of projection ensured that students would consider life as a unified and complete story, or idealized narrative, rather than focus exclusively on those goods that are of immediate and pressing concern to young, unmarried, childless, and pre-professional adults. The reliance on ostensive description rather than explicit definition ensured that students’ inability to articulate commitments that are largely implicit and tacit would not pose a problem.

2 Method

2.1 Participants

Participants were 109 students (53 women and 56 men) of Western European ethnicity at the University of Toronto in Canada, 108 students (56 women and 52 men) of Chinese ethnicity at Jilin University in China, 108 students (59 women and 49 men) of Indian ethnicity at the University of Mumbai in India, and 106 students (53 women and 53 men) of Japanese ethnicity at Nagoya University in Japan. Age ranged from 17 to 30 years with a mean of 19.39 for Canadians, 20.69 for Chinese, 22.19 for Indians, and 18.60 for Japanese. All pairwise comparisons on age were significant (α = .05) according to Tukey’s studentized range test. To control for the confounding of cultural background with age in comparing responses across nationalities, we included age as a covariate in the regression models reported in the next section.

Father’s highest level of education, measured on a 10-point scale ranging from elementary school to advanced university degree, was used as an indicator of socioeconomic status (SES). Ninety-seven percent of participants provided this information. Those who did not were distributed evenly across nationalities, p = .61 for Fisher’s exact test. Means for father’s education were 7.31 for Canadians, 3.96 for Chinese, 5.97 for Indians, and 7.11 for Japanese. Tukey’s studentized range test revealed that Chinese were lower than all other groups on the SES indicator, and Indians were lower than both Canadians and Japanese, who themselves did not differ significantly. To control for these confounding differences, father’s education was included as a second covariate in the regression models reported in the next section. The few participants who had not provided this information were assigned the mean value for their national sample. This allowed us to retain all observations in estimating the models.

2.2 Procedure

Participants completed questionnaires in small groups and received either course credit or a small gift in exchange for their time. Materials were presented in English to Canadian and Indian participants,Footnote 1 Chinese to Chinese participants, and Japanese to Japanese participants. Considerable care was taken in translation, including back-translation checks and adjustments. Participants completed a questionnaire consisting of several parts, three of which are relevant here.

First, participants were instructed to imagine themselves in the distant future, at 85 years of age and nearing the end of their lives. They were asked by which criteria they would decide whether they had lived worthy lives according to their personal beliefs and commitments. That is, what indicators would they look to as evidence that their lives had been worthy in their own eyes, irrespective of what others might think? It was made clear that the indicators could be anything—material things, achievements, knowledge, judgments, relationships, experiences, states. “Worthy” was made synonymous with “good,” “successful,” and “meaningful” in the instructions to accommodate Wierzbicka’s (2009) observation that these English words do not translate easily into many other languages. By using multiple terms that converge on the common notion of living well, we hoped to dilute the inevitable nonequivalence of any single term. The words were also chosen to accommodate all possible criteria of worth, however prudential or selfless, subjectively or objectively defined, and conventionally moralistic or amoral in content. We left the field of responses open. Participants were asked to come up with sixFootnote 2 indicators that were most important to them and write them down in the spaces provided.

Second, participants were asked to report the six indicators they thought the average student of their age and gender at their university would come up with in completing the same task. This allowed us to compare response patterns for self and average peer so as to address the influence of self-presentational concerns on the former. Presumably, any tendency to provide socially desirable but insincere responses would not apply as much (if at all) to describing what the average peer thinks is worthwhile in life as to disclosing one’s own commitments. Responses in both conditions would reflect normative cultural commitments, but be subject to different performative pressures. Where a national sample differed from other samples in the same direction on both self- and peer-referenced responses, a strong case could be made that the difference was not merely the result of dissimilar social concerns in the testing situation.

Third, participants completed the Short Schwartz Value Survey (SSVS; Lindeman and Verkasalo 2005). The SSVS is an abbreviated version of Schwartz’s (1992) measure of ten universal values (power, achievement, hedonism, stimulation, self-direction, universalism, benevolence, tradition, conformity, and security) and has been shown to have good reliability and validity. The method is simple and direct: Each value and its definition are provided and respondents indicate its personal importance on a 9-point scale ranging from opposed to my principles (0) to of supreme importance (8). The SSVS was included to allow for examination of basic value differences across nationalities and their relation to any differences found in what constitutes a good or worthy life. The former, we expected, would not account for the latter.

3 Results

3.1 Coding and Preliminary Analyses

Chinese and Japanese responses were translated to English by bilingual students familiar with the idiomatic aspects of language in the localities sampled. Participants reported a wide range of indicators. To reduce this variety to a comprehensive set of categories that would allow for reliable comparative analysis, three readers independently produced lists of repeatedly reported indicators. The lists were then harmonized and refined to produce a final list of 30 categories that occurred with enough frequency (2%) in the combined set of responses to merit inclusion in a fixed coding scheme. Two coders then independently assigned each reported indicator to one of the 30 categories in the list. If the indicator was not a clear example of any of these categories, it was assigned to the miscellaneous category. If the indicator could not be understood from the information provided, it was assigned to the uninterpretable category. The coders agreed in 96% of cases, a remarkable level of consistency for a 32-category scheme. Disagreements were resolved through discussion.

Only 0.62% of indicators reported in the self-reference condition were classified as miscellaneous. A Mantel–Haenszel test of general association revealed no significant differences across nationality (controlling for gender) in the number of such indicators reported, χ2(6) = 4.78, p = .57. The same was true of the 0.85% of indicators classified as miscellaneous in the peer-reference condition, χ2(9) = 8.55, p = .48. An additional 0.46% of indicators reported in the self-reference condition were classified as uninterpretable. Mantel–Haenszel testing revealed a marginally significant difference across nationalities in the number of such events reported, χ2(6) = 12.91, p = .04. Specifically, 0.77% of indicators reported by Indians and 1.10% of indicators reported by Japanese, but none reported by Canadians and Chinese, were classified as uninterpretable. A parallel but marginally nonsignificant difference emerged for the 0.81% of indicators in the peer-reference condition that were classified as uninterpretable, χ2(12) = 19.63, p = .07. Again, a slightly greater number of such indicators were reported by Indians (2.31%) and Japanese (0.79%) than Canadians (0.15%) and Chinese (0%). The reasons why Indian and Japanese participants were somewhat more likely to provide inadequately described indicators are unknown. In any case, the marginality of the difference, the minuscule relative frequencies for this category, and the fact that the vast majority (84%) of those Indians and Japanese contributing to it provided only a single uninterpretable indicator together suggests that the finding has little bearing on the main results reported below.

3.2 Comparing Indicators Across Nationalities

3.2.1 Method of Analysis

Due to the large number of indicator categories, participants rarely reported more than one instance of any particular category among the set of six. This justified analyzing responses in relation to the likelihood of reporting at least one indicator of a given category as a function of the explanatory variables of interest. Table 1 lists the 30 categories in order of sample prevalence in the self-reference condition. The ten most prevalent categories for each national sample are listed in Table 2. Prevalence was compared across samples for each of the 30 categories using repeated measures logistic regression. Specifically, the log odds of reporting at least one instance of a given category rather than none was modelled as a function of nationality (Canadian, Chinese, Indian, Japanese), the repeated factor of reference (self vs. average peer), the Nationality × Reference interaction, and the covariates of gender, age, and SES (father’s education).Footnote 3 The method of generalized estimating equations (GEE; Diggle et al. 2002) was used to produce efficient parameter estimates and standard errors in a context of correlated within-subject responses. To control familywise Type I error across the 30 models, a Bonferroni-corrected alpha level of .002 was used in testing the significance of the six predictors. One or two of the four national samples had no reported instances in either the self- or peer-reference condition for 7 of the 30 categories. These cell counts of zero led to quasi-complete separation and convergence failure in the estimation of the corresponding logit models (see Allison 2008). Following a solution recommended by Allison (1999, 2008), we combined the zero-cell samples in these instances with the sample that had the next lowest relative frequency after first confirming through Mantel–Haenszel testing that the combined frequencies did not differ significantly between themselves. Combining samples in this manner and re-estimating the problematic models led to successful convergence with appropriate parameter estimates and standard errors.

Table 1 Indicator categories in order of overall prevalence
Full size table
Table 2 Ten most prevalent indicator categories by nationality
Full size table

3.2.2 Results for Gender and Reference

The results revealed that men and women did not differ significantly on any of the indicator categories, controlling for age and SES. Reference emerged as significant for 13 of the 30 categories. Specifically, participants were more likely to report at least one instance of the categories knowing that I’ll be remembered after I’m gone for who I was or what I did, having had a positive impact on others or having made the world a better place, having lived a moral life according to my personal principles, having gained wisdom, having taken full advantage of opportunities and lived up to my personal potential, having raised my children well, and good relationships with family members, when reporting what was most important for assessing the worth of their own lives than when reporting what they thought was most important to the average peer. In contrast, participants were less likely to report at least one instance of the categories high social status or celebrity, power or influence over others, having had a successful career, being highly educated or possessing great professional skills/knowledge, a lot of wealth or assets, and having a happy and healthy family for themselves than for the average peer. With the exception of the last category, the pattern suggests greater willingness to claim morally laudable indicators than attribute them to others, and less willingness to claim indicators that might be seen as egoistic or self-seeking than attribute them to others. This self-enhancing tendency is consistent with the social desirability of the categories. The Nationality × Reference interaction, however, was not significant for any of the categories, indicating that self-enhancement was evident to the same extent across national samples. Reference, in other words, did not qualify any of the group differences described below.

3.2.3 Differences in Category Prevalence Across Nationalities

The GEE score statistic for nationality, the primary predictor in the logit models, was significant for 13 of the 30 categories: having had close and enduring friendships, χ2(3) = 26.99, p < .0001; having a happy and healthy family, χ2(3) = 51.83, p < .0001; having had a positive impact on others or having made the world a better place, χ2(3) = 24.17, p < .0001; a lot of wealth or assets, χ2(3) = 17.40, p = .0006; having had a successful career, χ2(3) = 42.10, p < .0001; having had lots of fun and other pleasurable experiences, χ2 (3) = 30.92, p < .0001; good relationships with family members, χ2(3) = 20.58, p = .0001; having travelled the world, χ2(3) = 27.21, p < .0001; financial security and comfort, χ2(3) = 19.20, p = .0002; high social status or celebrity, χ2 (3) = 20.20, p = .0002; having had a lot of involvement in my community, χ2 (2) = 28.55, p < .0001; having had children who are successful, χ2(2) = 15.34, p = .0005; and having pursued hobbies or leisure activities that were personally fulfilling, χ2(1) = 20.07, p < .0001. In each case, the effect was decomposed by conducting all possible pairwise comparisons. GEE parameter estimates for these comparisons were tested using empirical standard errors (White 1980). The results of these comparisons for each the 13 categories yielding differences across nationality are described in turn below.

  1. 1.

    Having had close and enduring friendships. Canadians differed significantly from Chinese, χ2(1) = 9.12, p = .003, and Japanese, χ2(1) = 11.70, p = .0006. Indians also differed significantly from Chinese, χ2(1) = 4.67, p = .03, and Japanese, χ2(1) = 4.62, p = .03. There were no other significant differences among nationalities. The predicted odds (adjusting for gender, age, and SES) of reporting (at least one instance of) this category (rather than none) were 2.50 times greater for Chinese and 2.68 times greater for Japanese than for Canadians. Similarly, the predicted odds of reporting this category were 1.87 times greater for Chinese and 2.01 times greater for Japanese than for Indians. See Table 3 for the relative frequencies, expressed as percentages, for each nationality in both reference conditions.

    Table 3 Relative frequencies for indicator categories by nationality
    Full size table
  2. 2.

    Having a happy and healthy family. Chinese differed from Canadians, χ2(1) = 9.06, p = .003, Indians, χ2(1) = 22.66, p < .0001, and Japanese, χ2(1) = 11.90, p = .0006. The latter three nationalities did not differ. The predicted odds of reporting this category for Chinese were 2.59 times greater than for Canadians, 4.31 times greater than for Indians, and 3.03 times greater than for Japanese.

  3. 3.

    Having had a positive impact on others or having made the world a better place. Chinese differed from Canadians, χ2(1) = 4.33, p = .04, and Indians, χ2(1) = 7.62, p = .006, whereas the latter two groups did not differ from each other. The predicted odds of reporting this category were 2.68 times greater for Canadians and 3.71 times greater for Indians than for Chinese. Finally, Japanese did not differ significantly from any of the other nationalities.

  4. 4.

    A lot of wealth or assets. Chinese differed from Canadians, χ2(1) = 8.29, p = .004, Indians, χ2(1) = 6.05, p = .01, and Japanese, χ2(1) = 11.70, p = .0006, whereas the latter three groups did not differ among themselves. The predicted odds of reporting this category were 2.60 times greater for Canadians, 2.12 times greater for Indians, and 3.13 times greater for Japanese than for Chinese.

  5. 5.

    Having had a successful career. Japanese differed from Canadians, χ2(1) = 19.89, p < .0001, Chinese, χ2(1) = 22.18, p < .0001, and Indians, χ2(1) = 15.76, p < .0001, whereas the latter three groups did not differ among themselves. The predicted odds of reporting this category were 4.29 times greater for Canadians, 5.34 times greater for Chinese, and 4.38 times greater for Indians than for Japanese.

  6. 6.

    Having had lots of fun and other pleasurable experiences. Chinese differed from Canadians, χ2(1) = 17.39, p < .0001, Indians, χ2(1) = 5.95, p = .01, and Japanese, χ2(1) = 21.34, p < .0001, and Indians differed from Japanese, χ2(1) = 4.84, p = .03. No other comparisons were significant. The predicted odds of reporting this category were 4.77 times greater for Canadians, 2.50 times greater for Indians, and 5.73 times greater for Japanese than for Chinese, and 2.26 times greater for Japanese than for Indians.

  7. 7.

    Good relationships with family members. Indians differed from Chinese, χ2(1) = 4.28, p = .04, and Japanese, χ2(1) = 5.90, p = .02. No other comparisons were significant. The predicted odds of reporting this category for Indians were 2.32 times greater than for Chinese and 3.13 times greater than for Japanese.

  8. 8.

    Having travelled the world. Canadians differed from Japanese, χ2(1) = 11.70, p = .0006, with no other differences among nationalities. The predicted odds of reporting this category were 8.93 times greater for Canadians than for Japanese.

  9. 9.

    Financial security and comfort. Japanese differed from Canadians, χ2(1) = 7.24, p = .007, Chinese (marginally), χ2(1) = 3.76, p = .05, and Indians, χ2(1) = 8.35, p = .004, whereas the latter three groups did not differ among themselves. The predicted odds of reporting this category were 5.77 times greater for Canadians, 3.72 times greater for Chinese, and 7.04 times greater for Indians than for Japanese.

  10. 10.

    High social status or celebrity. Canadians differed from both Indians, χ2(1) = 7.40, p = .007, and Japanese, χ2(1) = 7.84, p = .005, whereas the latter two nationalities did not differ from each other. The predicted odds of reporting this category was 3.10 times greater for Indians and 2.76 times greater for Japanese than for Canadians. Chinese did not differ significantly from any of the other nationalities.

  11. 11.

    Having had a lot of involvement in my community. Chinese differed from Japanese, χ2(1) = 9.49, p = .002, and from Canadians and Indians, χ2(1) = 22.75, p < .0001, who had been combined in this model to resolve the problem of quasi-complete separation. Japanese and Canadians/Indians did not differ. The predicted odds of reporting this category for Chinese were 6.30 times greater than for Japanese and 10.67 times greater than for Canadians/Indians.

  12. 12.

    Having had children who are successful. Chinese differed from Indians, χ2(1) = 4.71, p = .03, and from Canadians and Japanese, χ2(1) = 17.47, p < .0001, who had been combined in this model to resolve the problem of quasi-complete separation. Indians did not differ from Canadians/Japanese. The predicted odds of reporting this category for Chinese were 2.98 times greater than for Indians and 11.57 times greater than for Canadians/Japanese.

  13. 13.

    Having pursued hobbies or leisure activities that were personally fulfilling. Japanese differed from Canadian, Chinese, and Indians, χ2(1) = 21.07, p < .0001, who had been combined in this model to resolve the problem of quasi-complete separation. The predicted odds of reporting this category was 29.16 times greater for Japanese than for Canadians/Chinese/Indians.

The analysis of differences in category prevalence across nationalities yielded a fair number of group differences. Next, we examined the possibility of parallel differences in SSVS values.

3.3 Comparing Values Across Nationalities

3.3.1 Method of Analysis

Differences were evident in how the four national groups used the 9-point response scale of the SSVS. For example, Indians provided higher average importance ratings than all other groups for 8 of 10 values. The circumplex pattern of oppositions and compatibilities among the values (Lindeman and Verkasalo 2005; Schwartz and Boehnke 2004) implies that such a uniform difference between groups is due more to response bias than valid comparative measurement. To adjust for this bias, individual value ratings were standardized within-subject using the mean and standard deviation of the participant’s ten ratings. Overall and group means for the resulting relativized value scores are displayed in Table 4. To examine group differences, the scores were modelled as a function of nationality and the covariatesFootnote 4 of gender, age, and SES using ANCOVA. A Bonferroni-corrected significance level of .005 was used in effect testing to control for alpha inflation across the ten parallel models.

Table 4 SSVS values in order of overall relative importance and by nationality
Full size table

3.3.2 Value Differences Across Nationalities

The results revealed that men and women did not differ significantly on any of the values, controlling for age and SES. A significant effect for nationality emerged for 4 of 10 values: security, F(3, 424) = 13.28, p < .0001, conformity, F(3, 424) = 12.28, p < .0001, hedonism, F(3, 424) = 27.40, p < .0001, and stimulation, F(3, 424) = 22.95, p < .0001. In each case, the effect was decomposed by conducting all possible pairwise comparisons on the covariate-adjusted means. The results of these comparisons for each of the four values that yielded differences across nationalities are discussed in turn. These differences are summarized in Table 5.

Table 5 SSVS value differences across nationalities
Full size table
  1. 1.

    Security. Canadians assigned less relative importance to this value (mean within-subject standard score adjusted for gender, age, and SES = −.09) than did Chinese (adj. mean = .55), F(1, 424) = 24.96, p < .0001, Indians (adj. mean = .24), F(1, 424) = 6.68, p = .01, and Japanese (adj. mean = .52), F(1, 424) = 29.84, p < .0001. Also, Indians assigned less relative importance than did Chinese, F(1, 424) = 6.24, p = .01, and Japanese (marginally), F(1, 424) = 3.98, p = .05. Chinese and Japanese did not differ.

  2. 2.

    Conformity. Again, Canadians assigned less relative importance to this value (adj. mean = −.40) than did Chinese (adj. mean = .34), F(1, 424) = 34.94, p < .0001, Indians (adj. mean = −.02), F(1, 424) = 8.87, p = .003, and Japanese (adj. mean = −.01), F(1, 424) = 12.51, p = .0005. Also, Chinese assigned more relative importance than did Indians, F(1, 424) = 9.26, p = .003, and Japanese, F(1, 424) = 7.42, p = .007. Indians and Japanese did not differ.

  3. 3.

    Hedonism. Chinese assigned less relative importance to this value (adj. mean = −.95) than did Canadians (adj. mean = .19), F(1, 424) = 64.96, p < .0001, Indians (adj. mean = −.03), F(1, 424) = 46.34, p < .0001, and Japanese (adj. mean = .10), F(1, 424) = 51.96, p < .0001. There were no other group differences.

  4. 4.

    Stimulation. Japanese assigned less relative importance to this value (adj. mean = −.82) than did Canadians (adj. mean = −.09), F(1, 424) = 36.97, p < .0001, Chinese (adj. mean = .15), F(1, 424) = 47.60, p < .0001, and Indians (adj. mean = −.46), F(1, 424) = 5.88, p = .02. Also, Indians assigned less importance than did Canadians, F(1, 424) = 7.06, p = .008, and Chinese, F(1, 424) = 22.12, p < .0001. Canadians and Chinese did not differ.

3.4 Comparing Indicators across Nationalities While Controlling For Values

Can the above differences in basic values account for those found in the reporting of indicator categories? To examine the potential redundancy of the two sets of differences, the 13 repeated-measures logit models that had yielded significant group differences were re-estimated with the relative importance scores for security, conformity, hedonism, and stimulation included as covariates alongside gender, age, and SES. The added variables effectively controlled for the four values on which the groups had been found to differ and tested for group differences in the reporting of indicator categories as before.

The results were clear-cut. The GEE score statistic for nationality remained significant at the Bonferroni-corrected level of p < .002 for all 13 categories. Furthermore, the pattern of odds ratios representing pairwise comparisons among nationalities remained much the same in every case. The high consistency supports our argument that the specific indicators by which people measure the worth of their lives cannot be inferred in any obvious way from their endorsement of abstract, generalized values.

4 Discussion

The purpose of this exploratory study was to examine the content and variability of university students’ conception of a good life across four nations. The nations were chosen for their cultural separation, as evidenced in the frequency with which they have been the source of cross-cultural contrasts over the years. We remained agnostic about how much variation we would find, owing to the theoretical distance of this research from the well-established tradition of mapping nations in regard to basic or universal values. This distance was confirmed in our finding that values, as measured by the SSVS, did not account for differences across nationalities in the reporting of what makes for a worthy life. The lack of predictive redundancy does not necessarily mean that basic values are not implicated in the sort of life an individual is aiming at. It only indicates that any generalized value (and what else could a universal value be?) can find many alternative expressions in the substantive commitments by which people define themselves, their purposes, and their actions. In fact, such local modulation provides reason to question just how “generalized” these values really are. In any case, the bridge from values to agency is surely composed of many other aspects of culture—customary beliefs, institutional structures, real and imagined opportunities and deterrents, the distribution of power, social identities, and so on. Considered as an axis of our moral bearings, the vision we have of a good life provides the interpretive ground on which our actions are defined and evaluated—and our abstract values inferred.

How alike, then, are students in Canada, China, India, and Japan in what they think their lives are, or should be, about? To begin with, it is worth noting that the same pattern of similarities and differences was found for participants’ reports of their own commitments as for those they attributed to their typical peer. The response frequencies in the two reference conditions did differ in a self-enhancing fashion, but these within-subject differences had no bearing on those found across groups. The absence of interaction suggests uniformity of self-presentational concerns and, more importantly, discounts the possibility that the reports were merely performative and not genuinely reflective of students’ life commitments. Next, a set of only 30 categories was sufficient for capturing the variety of open-ended descriptions that students provided. This sufficiency, as reflected in the small number of indicators relegated to the miscellaneous category, held equally across nationalities. By itself, this finding suggests a fair amount of common ground. Of course, it could have been that each nationality tended to refer to categories that the others did not. This, however, was not the case. Comparison of the ten most prevalent categories by nationality revealed that four were shared by all groups and three more by three of the four groups. This amounts to substantial overlap in the most commonly reported categories, enough to justify summarizing what students on the whole deemed important in defining a good life.

Examination of the twelve categories reported by at least 25% of the combined sample reveals five that refer to good social relations (close and enduring friendships, happy and healthy family, positive impact on others, loving marriage or romantic partnership, good relationships with family members), three to practical or materialistic ambition (a lot of wealth or assets, successful career, achievement of great things), two to emotional life (well-being and contentment, fun and other pleasurable experiences), and two to character (moral life, wisdom). The prominence of social concerns in accounting for a good life is hardly surprising and consistent with past research (e.g., Twenge and King 2005). The remaining emphases suggest an equal spread of investment in a satisfying emotional life, conventional or socially recognized success, and what can be considered classical virtues. The hybridity of this profile hints at the cross-cutting vectors of aspiration and desire that characterize the “millennial generation” of university students around the world. Steeped in a global popular culture that promotes self-expression and consumerist individualism, tethered to incessantly streaming information and entertainment technologies, desperate for lasting approval and acceptance in an age of ever-increasing social fragmentation and dislocation, deeply cynical of institutional authority and tradition, and competitively anxious about their place in a doubtful economic future, today’s university students seem to be pulled in more directions than perhaps any generation before them. Accordingly, their “centrifugal” conception of a good life is a far from simple one.

Of the eighteen less commonly represented categories, six refer to social standing, recognition, or involvement (respect and admiration, being remembered after death, social status or celebrity, involvement in community, many friends, power or influence over others), five to personal achievement or fulfilment (realizing personal potential, fulfilling career, education and skills, overcoming challenges, fulfilling hobbies or leisure activities), two to children (raising children well, successful children), two to spirituality (harmony with nature or God, religious faith), two to quality of life (security and comfort, freedom), and one to travel (world travel). The above description provides an idealized common portrait. The findings also point to significant cultural variation in this pattern, to which we now turn.

Of the four groups, the Chinese stood out the most in regard to relative differences. Compared to the other groups, their conception of a good life was defined more by having a happy and healthy family, community involvement, successful children, close and lasting friendships, and less by the acquisition of wealth, impact on society or the world, and the pursuit of pleasure. Taken together, this suggests a socially circumscribed or localized, familistic orientation, consistent with the “morality of personal connections” (Fei 1992) that continues to define Chinese society despite 30 years of economic liberalization and relaxation of the hukou (residency permit) system.

The Japanese were similar to the Chinese in their greater emphasis (relative to Indians and Canadians) on close and lasting friendships, but less concerned than all other groups with career and money. Hobbies and leisure activities were more important for them than for the others. This pattern can be taken to reflect the ethos of a recessionary generation that grew up in the “lost decade” of the 1990s, the gradual recovery that followed, and the subsequent global financial crisis. To them, the ideology of lifetime employment, individual sacrifice, and national resolve that fueled Japan’s post-war “economic miracle” holds little credibility, relevance, or attraction (Matsumoto 2002). Depoliticized and more escapist than ambitious in orientation, this generation of Japanese places less emphasis than their parents on employment and salary in the quest for a good life.

Perhaps reflecting the internationalist leanings of their multicultural society, Canadians were more inclined than Japanese to include world travel in their description of a good life. Their lesser concern with social status than Japanese and Indians suggests more commitment to egalitarianism than exists in the latter, more hierarchical societies.

Finally, Indians were more concerned with good family relations than were Japanese and Chinese, owing perhaps to the practical and emotional demands of “joint family” life in India, in both its traditional and modified functional forms (Mullatti 1995).

These differences may seem modest in light of the considerable commonality described earlier. It should be recalled, however, that participants in this study were all students at nationally competitive universities. High-achieving students are more similar around the world than almost any other segment of the population. Socioeconomic differences across the four samples were further controlled for statistically. That group differences emerged for over 40% of the indicator categories despite this levelling and control suggests a fair degree of variation in moral orientation. Even so, it should be borne in mind that rural-urban differences within countries such as China and India far outweigh the sort of rarefied differences found here. Accordingly, it might be more accurate to view the differences described here as regional variations on globalization’s “new individualism” (Elliott and Lemert 2009) than as starkly different cultural visions of a good life.