Psychological motivations for collectivist behavior: comparison between Japan and the U.S

People have several motives to conform to their groups. Following the previous literature on social psychology and economics, we consider three psychological factors motivating conformity: (a) the efficiency factor, (b) the comfort factor, and (c) the satisfaction factor.

The efficiency factor refers to the individuals’ belief that cooperation in a group promotes achievement. In human society, it is widely known that cooperative action within a group often yields better economic outcomes than when individuals in a group act independently. This is supported by evidence suggesting that groups that cooperated tended to survive and expanded more than other groups (Bowles and Gintis 2011). If people believe in the efficiency of in-group cooperation, they will cooperate by conforming to the group’s decisions even when their opinions are different.

The comfort factor indicates the comfort that individuals feel when behaving similarly to others. In uncertain events and situations, people are not confident in the accuracy of their own information. Accordingly, following others can lead to safer and better decisions (Griskevicius et al. 2006). Thus, people often feel more comfortable when following group judgments based on the information of various members in the group (Castelli et al. 2001; Quinn and Schlenker 2002).

The satisfaction factor corresponds to satisfaction individuals feel when cooperating with others. Some people conform to groups in order to gain approval and affiliation and increase the likelihood that they will be liked by others (Castelli et al. 2001; Cialdini et al. 1999; Renkema et al. 2008). In addition, some people obtain psychological satisfaction from going along with others because they may gain “identity utility” when their behavior is consistent with the group’s code of conduct (Akerlof and Kranton 2010). Others try to avoid conflict by emphasizing harmony within the group; that is, they tend to suppress their own opinions and conform to the opinions of the group (Yamaguchi 1994).

Previous studies in economics, social psychology, and political science have shown that institutions in each country significantly affect human behavior and outcomes (Greif 2006; North 1991; Williamson 2000). North (1991) states that “institutions are the humanly devised constraints that structure economic and political interaction. They consist of both informal constraints (sanctions, taboos, customs, traditions, and codes of conduct), and formal rules (constitutions, laws, property rights)” (p. 97). There are differences between Japan and the U.S. not only in formal rules such as laws but also in informal institutions such as customs, traditions, and norms (Hofstede 1980; La Porta et al. 2008). These institutional differences would give rise to differences in psychological motivations for conformity between the two countries. In particular, we conjecture that differences in informal institutions between Japan and the U.S. will lead to a difference in the impact of the efficiency factor on people’s conformity.

In Japan, given its historical practices, the belief in achieving better economic outcomes through intra-group cooperation is likely to motivate people’s conformity. First, Japan's high population density, coupled with its limited natural resources, may have forced people to cooperate as it seemed more efficient for survival (Matsumoto et al. 1996). Second, in the past, most Japanese individuals secured their livelihood through agricultural activities that require cooperation (Triandis et al. 1993). In particular, they were engaged in rice farming, which requires irrigation systems and an extraordinary amount of labor, making cooperation more valuable (Talheim et al. 2014).⁴ Reflecting this history, Japanese elementary education has encouraged pupils to study in groups and teaches the restraint of the ego (Vogel 1979). These historical and educational traditions can inculcate belief in the economic value of cooperation and conformity among the Japanese people.

Furthermore, Japan was a country with low mobility of people, which may also have contributed to the efficiency of in-group cooperation.⁵ As is well known from game theory, long-term interactions among the same members promote greater gains through cooperation (Fudenberg and Levine 2009). Individuals will not pursue self-interest for the benefit of the group as long as they expect rewards from the group in the long-run or punishment by other group members (Greif 1994; Yamaguchi 1994). This may also have fostered the belief that cooperation promotes achievement, implying that as belief in the economic efficiency of in-group cooperation becomes stronger, people will behave in a more collectivist manner. Therefore, we conjecture that the efficiency factor will predict Japanese conformity behavior.

In contrast, the U.S. is characterized by abundant resources and a vast land area. In such an environment, the need to cooperate for economic gain is relatively small. Therefore, Americans are taught to cultivate independence rather than cooperation (Matusmoto et al. 1996). Moreover, the U.S. has been well known as a country with many immigrants and high mobility of people (Campbell and Kean 2012). In highly mobile societies, relationships and interactions among people tend to be shorter-term than in low-mobility societies. Game theory has accurately predicted that individuals in short-term interactions with others may experience failures of coordination and socially inefficient outcomes (Bowles and Gintis 2011). Therefore, in the U.S., even if a person recognizes that in-group cooperation efficiently achieves better outcomes, he or she may behave individualistically due to a perceived risk of exploitation by others. This suggests that the efficiency factor is less likely to predict conformity behavior in the U.S. compared with Japan.

We will examine the three psychological factors that can lead to conformity in Japan and the U.S., and thereby attempt to provide an insight into whether and how institutional differences between the two countries influence the effects of these factors.

This study used data collected in Japan and the U.S. during February 2006 in a panel survey conducted by the Center of Excellence (COE) project at Osaka University. The survey gathered data suitable for the analysis of human behavior and preferences in both countries, and particularly examined respondents’ preferences (e.g., time discounting, risk aversion, personal values). The questionnaire contained 87 questions, some of which included sub-questions, and the same questions were asked in both countries. Questions were initially composed in Japanese and then translated into English by a Japanese person who had stayed in the U.S. from ages 10 to 18 years. The translation was conducted with assistance from a specialist at a U.S. survey company. Finally, a prominent bilingual Japanese American economist assessed the semantic identities of the Japanese and English surveys. The survey was conducted annually from 2003 in Japan and from 2005 in the U.S., until 2013. However, five questions concerning IC were included in only the 2006 and 2012 waves of the survey.

This paper analyzed data collected in the 2006 survey from the five questions concerning IC.⁶ In Japan, 4879 people aged 20–75 years nationwide were selected using double stratified random sampling. Respondents were visited at their homes and handed the questionnaire. Completed questionnaires were collected several days later; 3763 questionnaires were returned (response rate: 77.1%). In the U.S., 4868 people aged 15–99 years were randomly selected from the registered membership of a large survey company, which covered all U.S. states, except Alaska and Hawaii. Questionnaires were distributed by mail; 3120 were returned (response rate: 64.1%).

We assume that individuals’ group conformity predicts their tendency to follow group decisions in their workplace and at home. Thus, we assume the following factors of conformity: (i) workplace conformity (W-CONF; i.e., the individual’s tendency to follow group decisions in the workplace), and (ii) home conformity (F-CONF; i.e., the individual’s tendency to follow family decisions at home). In our survey, respondents rated their conformity on these factors by responding to the following questions: “At work, I should follow opinion as a group” and “At home, I should follow my family’s opinion.” Responses to all questionnaire items used a 5-point Likert scale (1 = it doesn't hold true at all for me; 5 = it is particularly true for me). Therefore, higher scores indicate greater conformity on each factor.

As discussed above, we assume that individual conformity behavior stems from three psychological factors, namely, the efficiency factor, the comfort factor, and the satisfaction factor. First, we measure respondents’ prioritization of the efficient factor (EFFICIENCY) using the item “Working as a group results in greater achievement than working individually.” Higher scores on EFFICIENCY indicate a stronger belief that cooperation more efficiently promotes outcomes. Second, we measure respondents’ prioritization of the comfort factor (COMFORT) using the item “Behaving similarly to people around me makes me feel comfortable.” Higher scores on COMFORT indicate more comfort from behaving similarly to other group members. Third, we measure respondents’ prioritization of the satisfaction factor (SATISFACTION) using the item “I am more satisfied when I achieve a goal by cooperating with others than by myself.” Thus, higher scores on SATISFACTION indicate greater satisfaction from cooperation itself.

We analyze the following ordered probit models to estimate respondents’ psychological motivation to conform at work and at home:

In these models, \(i\) represents a respondent from either Japan or the U.S. and \(j\) represents the W-CONF or F-CONF score of 1–5. \({\text{CONTROL}}_{i}\) represents a set of individual attribute variables such as sex, age, family structure, education, occupation, and religion.⁷ Table 1 presents definitions and summary statistics of the variables. In models (1) and (2), underlying scores are estimated as the probability that the linear function of the three psychological factors and individual attributes, plus random error, is within the cutoffs. \(\kappa_{j}\) represents a set of cut-points corresponding to an ordinal value \(j\). \(\varepsilon_{i}\) represents normally distributed random error. We estimate the ordered probit models (1) and (2) separately for the Japanese and U.S. samples, thereby examining the ability of EFFICIENCY, COMFORT, and SATISFACTION to predict W-CONF and F-CONF among Japanese and U.S. respondents, respectively.

Respondents did not necessarily answer all questions related to the variables we analyzed. We excluded responses without data for these variables. Thus, the final number of responses included in the data analyses was reduced to 2797 in Japan and 2177 in the U.S. Table 1 summarizes statistics for all variables included in our analyses. Table 2 compares the mean values of conformity variables and psychological factors in the Japanese and U.S. samples. Mean W-CONF and F-CONF scores in the Japanese sample were significantly larger than in the U.S. sample, suggesting that Japanese people were generally more motivated to conform to their group than American people (Yamagishi et al. 2008).⁸ Additionally, this inference remained supported after we controlled for differences in response style between the two countries, using a within-culture standardization procedure for each variable (Table 3).⁹ Figure 1 presents response distributions for each conformity variable in the Japanese and U.S. samples, respectively. The distributions of W-CONF and F-CONF were both skewed to the right for Japan, and to the left for the U.S. These results support the view that Japanese people are more collectivist than American people (Hofstede 1980; Triandis 1995).¹⁰

Mean differences in scores for psychological factors were also found between the two countries (Table 2). EFFICIENCY and SATISFACTION were significantly higher in Japan than in the U.S. (p < 0.01); however, COMFORT was significantly higher in the U.S. than in Japan (p < 0.01). A comparison of means after within-culture standardization was consistent with these results (Table 3). Additionally, a considerable proportion of U.S. respondents assigned high scores to COMFORT (Fig. 2), although individualism and independence are commonly regarded as representative American values (Matsumoto et al. 1996).

A large amount of heterogeneity in conformity and psychological factors scores was observed within countries, in addition to differences between the two countries (Figs. 1 and 2); this result allowed us to analyze the motivations of conformist behavior in each country.

Tables 4 and 5 present the estimation results concerning workplace conformity [model (1)] and home conformity [model (2)], respectively.¹¹

Table 4 shows the estimation results for the ordered probit regression of W-CONF in the Japanese and U.S. samples. EFFICIENCY had significantly positive estimates in Japan (p < 0.05), but non-significant estimates in the U.S. The magnitude of the coefficient was significantly larger in Japan than in the U.S. (p < 0.05). This result suggests that Japanese people tend to conform in the workplace because they believe that cooperation more effectively promotes productivity, and that economic efficiency is less likely to motivate workplace conformity among Americans.

In contrast, non-significant estimates were obtained for SATISFACTION in both Japan and the U.S., suggesting that Japanese and American people do not tend to conform in the workplace because they find it satisfying.

Regarding individual attributes, the estimated coefficient of respondents’ sex was significantly negative in the W-CONF regression in both Japan and the U.S. (p < 0.05). This result suggests that men in both countries are less likely to conform at work than women, supporting earlier research (e.g., Cross et al. 2017; Triandis 1995). Except for management status, the other attributes were not significant for the Japanese sample, while some variables including occupation were significant in the U.S. sample. This last result suggests that degree of conformity in the workplace in the U.S. depends on occupation.

Table 5 presents the estimation results for the ordered probit regression of F-CONF. The estimated coefficients of EFFICIENCY were significantly positive in Japan (p < 0.05) and non-significant in the U.S. This result suggests that Japanese, but not U.S. individuals tend to conform at home because they believe it more efficiently promotes better outcomes.

Our empirical results are summarized as follows:

The above estimated coefficients for EFFICIENCY, COMFORT, and SATISFACTION in models (1) and (2) may partly involve an endogeneity problem: conformist behaviors could have caused individuals to hold pro-conformist beliefs, such as “I don’t feel satisfied or comfortable working individually because I have cooperated with others for so long.” Therefore, we employed the control-function instrumental variable estimation (CF) to manage potential reverse causality among the conformity variables and psychological factors.¹² CF provides consistent estimates for coefficients on endogenous regressors in parametric nonlinear models, including the ordered probit model, while the consistency is not guaranteed in two-stage predictor substitution, the usual IV method, for nonlinear models (Terza et al. 2008; Vella 1993; Wooldridge 2014, 2015). Thus, we used CF to estimate the two ordered probit models with the instrumental variables. In the first stage of estimation of CF, auxiliary regressions for endogenous regressors were conducted using instrumental variables the same as in the usual IV method. The second-stage regressions were subsequently performed by including the first-stage generalized residuals in the outcome equation of interest, that is, Eqs. (1) and (2).¹³ In CF, the significance of the first-stage residuals in the second-stage regression indicates the endogeneity of the regressors.

Tables 6 and 7 present the CF estimated coefficients for models (1) and (2). We adopted the individual attributes that were not significant in Tables 4 and 5 as the instrumental variables in the first-stage regressions and those that were significant as the control variables in the second-stage regressions.¹⁴ In these tables, \(e^{{{\text{EFFICIENCY}}}}\), \(e^{{{\text{COMFORT}}}}\), and \(e^{{{\text{SATISFACTION}}}}\) indicate residuals for EFFICIENCY, COMFORT, and SATISFACTION in the first-stage regression.

In Tables 6 and 7, no residuals of the first-stage regression are significant, suggesting that the three psychological variables were exogenous in all cases. As per the orthogonality conditions, Sargan’s J-test statistics reported at the bottom of Tables 6 and 7 could not reject the validity of our instruments. As for the weak-instruments problem, conventional tests in linear instrumental variable regression, such as the test of Stock and Yogo (2005) were not applicable to our non-linear regressions. Still, we confirmed that one or more instruments were significant at the 1% level in the first-stage regression for EFFICIENCY, COMFORT, and SATISFACTION, suggesting that the weak instruments problem was not serious in our case if it was present.

Accordingly, the estimates of psychological variables were qualitatively the same as those in Tables 4 and 5, except that COMFORT became non-significant in the F-CONF regression for the Japanese sample. Thus, the endogeneity problem was not serious in the estimation of Eqs. (1) and (2).

Persons without a regular occupation (resp., without a family) may not provide meaningful responses to the workplace (resp., home) conformity questions. Therefore, as an additional robustness check, we conducted a subsample regression of W-CONF by excluding respondents with no regular occupation, and a subsample regression of F-CONF by excluding single respondents.¹⁵ However, we found that the estimation results did not change substantially in the sub-sample regressions.¹⁶

We also used the ordered logit model instead of the ordered probit models; the estimation results on the three psychological factors remained qualitatively unchanged. Additionally, the estimation results remained qualitatively unchanged in the least-squared estimation. Furthermore, we repeated all estimations following application of a within-culture standardization procedure for each variable; the estimation results again remained qualitatively unchanged.