RELIANT BEHAVIOR IN THE UNITED STATES AND JAPAN.

BEARD, T. RANDOLPH ; BEIL Jr., RICHARD O. ; MATAGA, YOSHIHARU 等

YOSHIHARU MATAGA [*]

Japanese economic success is often attributed to culturally reinforced psychological conditioning that promotes interpersonal reliance, cooperation, and a group interest orientation. This article provides direct experimental evidence on differences in behavior among future businesspeople in the United States and Japan. Utilizing a simple, two-person extensive form game of perfect information introduced by Selten (1975), we provide evidence that, contrary to some views, the Japanese can be less reliant on the behavior of others and are more likely to take actions at variance with group welfare in some settings. Thus, popular explanations of Japanese economic achievements may require further exploration. (JEL A14, F00, M14, P52)

I. INTRODUCTION

The role of culture in economic behavior is an important--but controversial--topic. Although it is widely believed that cultural influences affect economically relevant behavior, such effects can present difficulties for "conventional" economic theory. Discussions of the differences between the United States and Japan in the business sphere provide a good example of both the high degree of academic and popular interest in this topic and the complexities involved in obtaining useful characterizations.

A common popular conclusion in the Japan versus United States "debate" is that Japanese business practices differ fundamentally from United States practices and that these differences stem from dissimilar sociological conditioning in the two countries. While United States (or "Western") culture stresses individualism, personal initiative, and risk-taking, the Japanese are said to place a higher value on consensus, cooperation, favoritism toward "within-group" peers, and stability. [1] Therefore, a citizen in Japan may be more interested in how he or she fits into the social context and his or her relationship with others. [2] These contrasting orientations are assumed to manifest themselves in different patterns of behavior having economic implications.

This article offers a direct, controlled experimental test of potentially significant differences in behavior between future Japanese and American businesspeople. Utilizing a simple, two-person game of perfect information, we examine the degree to which United States and Japanese business students make differing assessments of the reliability of the behavior of their compatriots, and we provide evidence on the implications of these assessments for observed behavior. Our simple game template, introduced by Selten (1975), was the focus of a study by Beard and Beil (1994) and allows comparison of a kind of behavior, reliance on maximizing play by others, that may have important economic implications. In our usage, "reliance" means merely that one player assigns a sufficiently high probability to self-interested behavior by other players to induce the first player to take actions that are desirable only when others maximize their own monetary payoffs. Thus, reliance differs from "trust" as usually defined in gam e theory, and the former appears considerably weaker.

Our experiments use as subjects business students in roughly comparable American and Japanese universities, and we offer nontrivial, immediate money payoffs to participants in order to solicit serious efforts on their parts. Several versions of the basic experiment are conducted to establish payoff dominance and evaluate hypotheses on subject behavior. Our findings are quite surprising. First, subjects in both the United States and Japan are loath to rely on others' maximizing behavior, even when that behavior is, as in the United States, highly reliable. Second, contrary to the assumptions underlying some analysis of Japanese business practices, subjects in Japan appear less reliant than their United States counterparts.

The article is divided into five sections. Section II provides a literature review on the role of culture in U.S.--Japan business behavior. Section III describes the game form used to design our experiments. Section IV describes the experiments and their results, and a conclusion completes the article.

II. BUSINESS PRACTICES: UNITED STATES VERSUS JAPAN

The literature on Japanese business practices and ethics is vast, and we treat it very briefly here. Nevertheless, several common themes can be identified. First, an emphasis on collective decision making pervades Japanese business practices: Important decisions are subject to lengthy internal debate with an eye toward obtaining strong consensus prior to taking action. This pattern of collective choice is sometimes attributed to Confucian ethical influences, which stress harmony (Wa). There are, however, conflicting theories about how this collective action affects behavior. Weisz et al. (1984), Kagitcibagi and Berry (1989), Triandis et al. (1988), and Yamaguchi (1994) all offer excellent surveys. Most authors believe, however, that individualism in its Western interpretation is frowned upon. Rather, a hierarchical structure subject to social reinforcement establishes strict roles for business associates to play, and corporate governance is often likened to running a well-ordered household.

Among the most noted consequences of the above is the apparently very high level of "trust" inherent in some Japanese business practices. These practices, which often seem astounding to Westerners, include the formation of apparently binding but noncontractural "conglomerates" (Keiretsu). Sours (1982, 28) notes that "these relationships ... rarely rely on written contracts. ... (this) demonstrates, by contrast [to the United States], the high degree of unspecified trust and obligation assumed by Japanese managers in their mutual dealings, both intra-firm and inter-firm." Although Sours writes of "unspecified trust," there is an important distinction between "trust," as that term is used in game theory, and "reliance." Consider, for example, the canonical experiment to measure "trust" proposed by Berg et al. (1995). Subject A makes a voluntary transfer of money to B. This money is trebled, then B may return any desired share of it to A. Although this template undoubtedly measures trust as usually conceived, t he economic relevance of this circumstance is not universal, and few market transactions resemble this experiment.

In contrast, "reliance" signifies only that a player expects others to pursue their own interests in a reliable fashion. Unlike the "trust" scenario described above, reliance gives party B, say, a positive (if small) incentive to take an action that benefits A. Of course, one expects players to act in their own self-interest in most cases. However, this claim is quite different from the notion that players "rely" on the maximizing play of others. This latter idea we term "reliance," and we would argue that reliance is at least as important in economic relationships as trust. Trust is needed to promote efficient outcomes in the absence of contracts (if only implicit ones), and may allow players to forgo costly legal arrangements, whereas reliance is precisely that which makes contracts work.

Several recent papers have examined differences in trust among Japanese and Americans. Yamagishi (1998a, 1998b), Saijo and Nakamura (1995), Cason et al. (1997), and Brandts et al. (1998) have conducted voluntary contribution mechanism (VCM) experiments in both countries, and such investigations are related to measurements of trust. These experiments have revealed mixed results. Yamagishi (1988a), Yamagishi (1988b), and Saijo and Nakamura (1995) all find that free riding in VCM experiments is higher in Japanese populations than it is in identical experiments in United States populations. However, Brandts et al. (1998) found no difference in behavior between Japanese and U.S. participants in their VCM environment, whereas Cason et al. (1997) find higher rates of contribution in the VCM environment in the Japanese treatments than those exhibited in their U.S. experiments.

Related to the issue of trust is the phenomenon of "spite," which has also received some experimental examination. Spite refers to a player forgoing benefits to reduce the benefit to another. Saijo and Nakamura (1995) find more spite in Japan than in the United States, and speculate that Japanese aversion to payoff inequality may cause this result.

Observed variations in trust between U.S. and Japanese subjects have been explained by two opposing theories. The first, described by Fukuyama (1995), conventionally holds that cohesive social institutions engender greater trust, and that Japan is a very "high-trust" society. Alternately, Yamagishi and Yamagishi (1994) argue that greater interdependence in decision making reflects both a lack of trust and a habitual reliance on monitoring and sanctions.

Although the literature on trust in Japan and the United States is relatively extensive, much of the discussion of differences between U.S. and Japanese business behavior has tended to blur the important distinction between "trust" and "reliance." This is unfortunate because, at the sociological level, these two behaviors may be substitutes or complements. The purpose of this article is to examine reliance. We turn now to an examination of how reliance might be studied.

III. AN EXPERIMENTAL TEMPLATE

Reliance refers to the willingness of a player to eschew a secure action and place his or her fate in the hands of another. Unlike analyses of trust, however, this second player has a positive (though perhaps small) incentive to "reward" the first player's action. To see how this can be evaluated, consider the following simple game first discussed by Selten (1975) (Figure 1).

Figure 1 is a two player, one-shot extensive form game of perfect information. Player A moves first, selecting action L or R. if A selects L, the game ends with payoffs (X, Y) to A and B, respectively. However, if A selects R, then B observes this and makes a choice between actions l and r. If B selects l, both players receive zero payoffs. If B chooses r, however, A receives some large payoff (say, $1,000,000) while B receives some small payoff (say, $1). All of this is known to the players.

When $0 [less than] X [less than] $1, 000, 000, the game has two sorts of Nash equilibria. In the first, A selects R and B selects r, leading to the $1,000,000 payoff for A. This equilibrium is subgame perfect. Alternately, however, if A fears that B would (with some suitably large probability) respond to R with l, then A will select L. This leads to another Nash equilibrium, which, however, is imperfect: For A to select L, A must believe B could play l. Yet such a belief is inconsistent with maximizing play by B, and thus the equilibrium this belief supports is imperfect.

The game depicted in Figure 1 is an important one for game theory and has been the subject of considerable debate. Rosenthal (1981) suggested that, when X was suitably large but less than $1,000,000, reasonable people in the place of player A would select L. This conjecture implies that, when the stakes are sufficiently large, people will not act as if the hypothesis that others are maximizers is totally reliable. Notice that this claim does not require that others deviate [Graph Omitted] from maximizing play, only that, in suitable circumstances, players will act as if deviation is conceivable (has nonzero probability). Further, this "nonreliance" is not equivalent to not "trusting" the other player: It is quite conceivable a player A might "rely" on B's choice in Figure 1, but give B $0 in the trust game of Berg et al. (1995).

Beard and Beil (1994) subjected the "Rosenthal conjecture" to extensive testing using experimental designs derived from the game in Figure 1, though of course using lower payoffs. Utilizing numerous anonymous treatments based on this template, they found that: (1) subjects acting as A players frequently selected secure strategies (equivalent to A selecting L); (2) the proportion of A players selecting secure strategies varied across treatments (as the payoff X varied) in a manner consistent with Rosenthal's discussion; (3) A player subjects were very likely to select the secure action L when the game payoffs satisfied the condition Y [greater than] $1, a phenomenon suggesting a fear of "revenge." Beard and Beil (1994, 261) provide extensive discussion of possible explanations for their results, including incomplete information arguments, and conclude that "although subjects were observed to make own-payoff maximizing choices about 98% of the time, a majority of experimental participants evidenced an unwillin gness to rely on the maximizing behavior of others."

Although the Beard and Beil experiments provided strong evidence for the existence of an unwillingness by subjects to "rely" on the maximizing behavior of others (at least when the stakes were sufficiently high), subjects in the experiments were American college students enrolled in business classes at a large public university. Yet if the analyses of Japanese business ethics presented earlier is correct, the findings of Beard and Beil (1994) may merely reflect cultural attributes of Americans, and results from experiments with otherwise similar Japanese subjects might be quite different.

The game of Figure 1 provides an attractive framework for evaluating claims about Japanese business behavior. In particular, the "nonreliant" behavior conjectured by Rosenthal corresponds fairly directly with a test of the degree of unspecific "reliance" subjects have in the behavior of others. If some popular commentators on Japan are correct, we might observe greater willingness by Japanese subjects to rely on maximizing play by other Japanese participants than is seen in the United States, at least when an appropriate group identification exists between them. On the other hand, if Yamagishi and Yamagishi are correct, then we might instead see a lower level of "reliance" in the experiments conducted in Japan. For example, the intensive social structure of Japan may make the Japanese more accustomed to punishing individuals who do not behave "cooperatively." Of course, given the anonymous pairing of subjects in a one-shot experiment, the opportunities for punishment would appear to be limited.

Additionally, the game in Figure 1 can provide insights into a second common claim made about the Japanese. In particular, the hypothesis that Japanese decision makers place great emphasis on their group (as opposed to their individual) welfare can be evaluated using the template of Figure 1. When, for example, Y [greater than] $1, selection of the perfect equilibrium strategy R by player A indicates that A both expects B to engage in maximizing play and that A is willing to enrich himself at B's expense. This latter effect may be inconsistent with the "cooperative" orientation supposed to characterize Japanese business behavior and with cultural traditions that place "group" welfare over that of the individual.

IV. EXPERIMENTAL DESIGN AND RESULTS

We designed and executed three experimental treatments based on the game form in Figure 1 at the University of Marketing and Distribution Sciences (UMDS) in Kobe, Japan. The purpose of this exercise is to compare results obtained from the Japanese experiments with those results for the equivalent experiments analyzed by Beard and Beil (1994). The payoffs for the U.S. and Japanese experiments appear in Table 1.

The experimental comparisons outlined here are facilitated by several factors. First, the U.S. experiments were conducted at Auburn University, a large public institution occupying a comparable national academic rank as the UMDS. Second, both experiments utilized only subjects enrolled in business classes at these institutions, and all the experiments were conducted in an identical manner. Additionally, a survey was conducted to determine students' perceived opportunity costs at both universities. The students in Japan revealed opportunity costs which were slightly higher than those revealed in the United States. The conversion from dollars to yen took this difference into account. Furthermore, all translations of instructions were evaluated by a bilingual psychologist. Thus, every effort was made to offer subjects equivalent experiences. [3]

Three experimental treatments were conducted in both countries. The first, treatment 1, constitutes a baseline experiment. The secure action L by player A pays almost as much as the best payoff A obtains from choosing R. B has a positive, though small incentive to reward A's choice of R with the perfect equilibrium payoff of $10 (1,500 yen). A payoff-maximizing player B would prefer that A select R in this treatment, as that choice allows B to obtain a higher payoff than that offered by A's secure strategy. The secure choice of L indicates that a payoff-maximixing player A entertains the possibility that B may not engage in maximizing play. Thus, the proportions of secure choices by A participants provide a direct measure of their reliance on maximizing play by their experimental counterparts.

Treatment 2 duplicates treatment 1 except for a reduction in the payoff to A of the secure strategy L. This variation is introduced to gauge the degree of "nonreliance" on maximizing play by B participants. With the secure payoff reduced from 97.5% to 70% of the A's perfect equilibrium payoff, only those maximizers who are either extremely risk averse or very uncertain of B's behavior should select the secure action L. Thus, one expects that treatment 2 will exhibit less (or, at least, not more) secure play than treatment 1.

Treatment 3 is a variation on treatment 1 in which B's payoff from A's secure choice is increased to a level above any payoff B could get if A picked R. Here, unlike the other treatments, a payoff-maximizing player B would want A to select the secure choice L. Thus, a choice of R by A indicates a great deal of reliance on B's response, since the choice of R is, in fact, damaging to B. Selecting R then indicates both a disregard for B's welfare, and a high degree of confidence in B's maximizing play. Complicating this interpretation somewhat is the fact that the selection of L by A also minimizes the differences between payoffs. Aversion to payoff inequality may animate some subjects.

All experiments were conducted in the same manner. Business students were recruited from classes and given a factual description of the range of money payoffs offered. After reading the instructions in common and answering any questions, each subject randomly selected a participation number. Participants were then paired together by an unrevealed rule, with one player acting as the A subject, the other as the B. The participants were then physically separated for the duration of the experiment. All pairings were anonymous, and no participant could ever learn with whom they were paired. After completing exit questionnaires, all subjects were privately paid their earnings, plus a participation fee. Typical sessions involved about 20 subjects (10 pairs) and took 30 minutes to complete. No subjects participated more than once, nor were the results or purpose of the experiments revealed at any time.

Table 2 summarizes the results of the experiments for both the Japanese and U.S. trials. Several points are of note. First, secure play is very common: In treatment 1, more than two-thirds of total participants selected secure strategies. Second, in the U.S. trials, nonmaximizing play is rare, with only 2 instances by B players in 35 opportunities. [4] In Japan, however, nonmaximizing actions were more commonly observed (6 of 21 opportunities), most such deviations arising in treatment 2. Finally, the levels of secure play vary among trials in a consistent manner for both the United States and Japan, with treatment 3 inducing the most secure play and treatment 2 the least. Thus, at a superficial level, the treatments appear to affect behavior in the expected manner.

We are now able to begin our evaluation of the differences, if any, between actions selected by Japanese and U.S. subjects. Table 3 presents pairwise equality tests for the proportions of secure play among treatments. To calculate these statistics, let [P.sub.ij] be the proportion of secure play exhibited in treatment (i) in country (j). Let [P.sub.ijrk] be the pooled frequency for treatments (i, j) and (r, k). We define the test statistic z as

(1) z = ([p.sub.ij] - [p.sub.rk])

x [square root][p.sub.ijrk](1-[p.sub.ijrk])([[n.sup.-1].sub.ij]+[[n.sup.-1].sub .rk])

where [n.sub.ij] is the sample size (number of pairs) for treatment (i, j). We take z to be approximately normally distributed with mean zero if the null hypothesis of equal probabilities of secure play is correct.

Denoting treatments 1, 2, and 3 conducted in Japan as J1, J2, and J3, and similarly for the U.S. (US1, US2, US3), Table 3 provides some direct evidence on differences in behavior between U.S. and Japanese businesspeople that arise in identical strategic circumstances. We note first that statistically significant differences in the rates of secure play (nonreliant behavior) exist in pairwise comparisons of trials J2 and US2, and J3 and U53. In each case, the Japanese subjects are less reliant on the behavior of their anonymous "partners" than the Americans are. This conclusion is at considerable variance with the views expressed by many commentators, is inconsistent with the popular media view of Japanese business behavior, and appears inconsistent with some of the psychology and sociology literature on Japanese versus American culture as exemplified by the analyses of Bachnik (1986), Cousins (1989), De Vos (1985), Shikanai (1978, 1983, 1984), and Weisz et al. (1984). The extent of nonreliance by Japanese sub jects appears quite sensible when it is noted that violations of maximizing play are, in fact, much more common among participants in Japan. This violation of payoff maximization may be similar to the spiteful behavior observed by Saijo and Nakamura (1995).

A second popular hypothesis about Japanese decision making holds that group interests are more important in Japan and that Japanese businesspeople will frequently put group interests above individual interests. Our experimental design offers two kinds of evidence on this point. Treatment 3 puts A player subjects in the position of actually reducing the earnings of B participants by the selection of the nonsecure action R. Player B subjects' responses to A players' selections of the "reliant" action R provides further evidence on the importance of the welfare of other players in subjects' choices.

We note that every Japanese subject who acted as a player A participant in treatment 3 selected the secure action L. While 86% of Americans did likewise, the Japanese rate is still significantly higher. Combine this result with the observation that Japanese subjects acting as B players were significantly more likely to adopt nonmaximizing play (z = 2.37), and one concludes that Japanese subjects acting as A players selected the secure strategies in treatment 3 more often because they (correctly) feared a nonmaximizing response by B players. The claim that they did so out of concern for the earnings of their B partners is inconsistent with the high level of nonmaximizing play observed in these trials. However, this finding is consistent with Yamagishi and Yamagishi's (1994) theory that cooperation in Japanese Keiretsu is not a product of culture but depends on the ability of participants to monitor and sanction behavior. In this case, B is able to "sanction" A for the choice R by selecting l.

All participants were asked to complete an exit survey. As a part of this exercise, each was asked what choice they would make as an A player, and their reasons for this choice. Though such choices are hypothetical, one may look at the results if great caution is applied. We can categorize the responses as follows. For those indicating they would select L, the natural categories of explanation are: secure ("I wanted to ensure my payoff"), fear ("I am scared that the person I am paired with will shortchange me"), rivalry ("I wanted to make the difference between my payoff and my opponent's the greatest"), and group maximizing ("I wanted to make sure we had the highest payoff possible for both of us"). [5] There may be no real difference between the "secure" and "fear" explanations. For those hypothetically selecting R, the categories of explanation are: reliant ("clearly the B player will want to maximize his or her earnings, so I will earn the most if I choose R"), group maximization ("if I choose R and my B player chooses r, then we will make the most available for both of us"), and gamble ("I'm taking a risk that the B participant will choose r"). With the exception of one or two responses, all of the participants' explanations easily fit into one of the above categorizations.

Table 4 summarizes the results when players are cast as A participants. Although some statistically insignificant differences are illustrated, the most striking result is the high proportion of "fear" as an explanation of Japanese subjects' choices of L (significant in every treatment), and the strange differences with respect to the "group payoff maximization" explanation in treatments 2 and 3. When one consolidates the similar categories "secure" and "fear," however, no significant differences in the rates of this explanation emerge between Japanese and American subjects who claim they would pick L as an A player.

The group payoff maximization results are confusing and may be plagued by the small sample of Japanese subjects who claim they would select R. In treatment 1, the Americans appear more "group conscious," while the reverse occurs in treatment 2. Pooling treatments 1 and 2 produces insignificant results.

One must evaluate hypothetical actions with great caution. However, taken with the economic results, the exit questionnaire suggests that, while the Japanese are more likely to act in a nonreliant fashion, their motives for doing so may be quite similar to those of American subjects. The desire for security, or the fear of others' actions, is simply more prevalent among Japanese decision makers.

One explanation for our findings consistent with the discussion above hinges on the relevance of "group identification" among U.S. and Japanese subjects. As noted by Clark (1983, 15), the establishment of group identification between individuals in Japan typically requires face-to-face meetings. The anonymous one-shot pairing used in our experiments may interfere with this necessary connection, although it would be surprising if college students in the same classes at the same university did not feel some group identification. In any event, group versus nongroup identification is common in all cultures, and the operative significance of this idea flows primarily from the strictness of the criteria required for group membership. The combination of the potential lack of group identification, and the inability of participants to sanction their partners' actions, leads to a reluctance by A participants to be "reliant." However, since our experiment is a single-period game, it is unlikely that the significant non maximizing play by the B players can be classified as anything other than spite.

V. CONCLUSION

Trust and reliance are different characteristics of behavior and, although trust can facilitate many economic interactions, reliance may be an important aspect of relations between "strangers." Our experimental results suggest that the Japanese, who are (perhaps incorrectly) popularly portrayed as more trusting, are not as reliant as American subjects. This attitude on the part of the Japanese is rational given the relative frequencies of nonpayoff-maximizing play but appears to arise from motives indistinguishable from those animating American subjects. We note further that our results appear to support the Yamagishi and Yamagishi (1994) interpretations, particularly with respect to spite.

If one accepts the popular characterization of Japan as stable and "homogenous," then the emergence of nonlegal institutions is possibly the result of the prevalence of "repeated play games" (with unknown endpoints) and the opportunity for sanctions, inherent in Japanese society. Such a finding would explain the emergence of significant, nonlegally binding social institutions without the assumption that the Japanese are "different" from the Americans (or anybody else). Yet our results suggest that there are differences in economic behavior that should be investigated.

In many psychological, sociological, and anthropological studies of Japanese culture, it has been shown that cooperation is the norm when individuals view themselves as belonging to a common group. In fact, it is not unusual for a businessman in Japan to travel several hundred miles to meet his partner face to face before making any transaction. On the other hand, these same studies suggest that cooperative behavior is unlikely if group identification is not present. In our experiment, we assumed that the group affiliation role might be fulfilled by the students' university and class affiliations. One explanation of our results therefore could be that the Japanese participants viewed themselves as being paired with people who belong to different groups. It would be interesting to conduct further experiments that clearly confine participants to specific groups and evaluate the resulting behavior within and across groups. It also seems likely that group identification criteria may differ between "trust" and "r eliance" phenomena. One would ordinarily expect that trust requires stronger group identification than does reliance. Further experiments that directly test issues of equality of payments and the ability to sanction behavior are also appropriate.

The results presented here are primarily suggestive, rather than conclusive. All cross-cultural studies suffer from various difficulties, and ours is no exception. One might plausibly argue that there are no real categories corresponding to "Japanese" or "American/Western" culture. It remains an important task, however, to identify differences in the economically relevant behavior of different peoples, as we urge further work on this topic.

(*.) We appreciate comments from Reinhard Selten, Ernst Fehr, Catherine Echols, Gary Becker, Jordi Brandts, Timothy Cason, Nick Feltovich, Simon Gacther, Celine Jullien, Rachel Croson, Arthur Shram, Andrew Schotter, James Walker, participants in the 1998 Economic Science Association meetings in Mannheim, Germany, William Neilson, and two anonymous referees. All remaining errors are our own.

Beard: Associate Professor of Economics. Auburn University, 415 W. Magnolia Rm 203, Department of Economics, Auburn, AL 36849-5242. Phone 1-334-844-2918, Fax 1-334-844-4615, E-mail RBEARD@business.auburn.edu

Beil: Associate Professor of Economics, Auburn University, 415 W. Magnolia Rm 203, Department of Economics, Auburn, AL 36849-5242. Phone 1-334-844-2921, Fax 1-334-844-4615, E-mail RBEIL@business.auburn.edu

Mataga: Professor, University of Marketing and Distribution Sciences, 3-chome Gakuen-Nishi-machi Nishiku Kobe 651-21 Japan. Phone 81 78-796-3543, Fax 81 78-794-6149, E-mail mataga@umds.ac.jp

(1.) Shikanai (1984), and Weisz et al. (1984) are two good surveys of the psychological comparisons of Americans and Japanese.

(2.) We refer the reader to Sampson (1988) and Lebra (1976) for further reading.

(3.) The translation of the original instructions was done by Professor Mataga with the help of his wife, Etsuko. The translation was cross-checked by Professor Shin Hachiya, a psychology professor at UMDS.

(4.) In the original experiments of Beard and Beil (1994) the B's had 108 opportunities to select l when R was chosen. This only happened twice--once when a particular later told the experimenter that he had made a mistake, and once when the participant was clearly inebriated.

(5.) The group max explanation of the L choice only occurred in treatment 3.

REFERENCES

Bachnik, J. M. "Time, Space and Person in Japanese Relationships," in Interpreting Japanese Society: Anthropological Approaches edited by J. Hendry and J. Webber. New York: Oxford University Press, 1986, 49-75.

Beard, T. R., and R. O. Beil Jr. "Do People Rely on the Self-Interested Maximization of Others? An Experimental Test." Management Science, 40, February 1994, 252-62.

Berg, J., J. Dickhaut, and K. McCabe. "Trust, Reciprocity, and Social History." Games and Economic Behavior, 10, July 1995, 122-42.

Brandts, J., T. Saijo, and A. Sehram. "A Four Country Comparison of Spite and Cooperation in Voluntary Contribution Mechanisms." Instituto de Analisis Economico mimeo, 1998.

Cason, T. N., T Saijo, and T Yamato, "Voluntary Participation and Spite in Public Good Provision Experiments: An International Comparison." Purdue, mimeo, 1997.

Clark, G. Understanding the Japanese. Tokyo: Kinseido, 1983.

Cousins, S. "Culture and Selfhood in Japan and the U.S." Journal of Personality and Social Psychology, 68, January 1989, 163-71.

De Vos, G. "Dimensions of the Self in Japanese Culture," in Culture and Self, edited by A. Marsella, G. De Vos, and F. L. K. Hsu. London: Tavistock, 1985. 149-84.

Fukuyama, F Trust: The Social Virtues and the Creation of Prosperity. New York: Free Press, 1995.

Kagitcibasi, C., and J. W Berry. "Cross-Cultural Psychology: Current Research and Trends." Annual Reviews in Psychology, 40, 1989, 493-531.

Lebra, T. S. Japanese Patterns of Behavior. Honolulu: University of Hawaii Press, 1976.

Rosenthal, R. "Games of Perfect Information, Predatory Pricing, and the Chain Store Paradox." Journal of Economic Theory, 24, August 1981, 92-100.

Saijo, T., and H. Naka mura. "The 'Spite' Dilemma in Voluntary Contribution Mechanism Experiments." Journal of Conflict Resolution, 39, September 1995, 535-60.

Sampson, E. E. "The Debate on Individualism: Indigenous Psychologies of the Individual and Their Role in Personal and Societal Functioning." American Psychologist, 43, January 1988, 1203-11.

Selten, R. "Reexamination of the Perfectness Concept for Equilibrium Points in Extensive Games." International Journal of Game Theory, 4, 1975, 25-55.

Shikanai, K. "Effects of Self-Esteem on Attribution of Success-Failure." Japanese Journal of Experimental Social Psychology, 18, 1978, 47-55.

_____. "Effects of Self-Esteem on Attributions of Others' Success or Failure." Japanese Journal of Experimental Social Psychology, 23, 1983, 27-37.

_____. "Effects of Self-Esteem and One's Own Performance on Attribution of Others' Success and Failure." Japanese Journal of Experimental Social Psychology, 24, 1984, 37-46.

Sours M. H. "The Influence of Japanese Culture on the Japanese Management System," in Japanese Management Cultural and Environmental Considerations, edited by S. M. Lee and G. Schwendiman. New York: Praeger, 1982.

Triandis, H. C., R. Bontempo, M. J. Villareal, M. Asai, and N. Lucca. "Individualism and Collectivism: Cross Cultural Perspectives on Self-Ingroup Relationships." Journal of Personality and Social Psychology, 38, February 1988, 323-38.

Weisz, J. R., F. M. Rothbaum, and T. C. Blackburn. "Standing Out and Standing In: The Psychology of Control in America and Japan." American Psychologist, 39, 1984, 955-69.

Yamagishi, T. "The Provision of a Sanctioning System in the United States and Japan." Social Psychology Quarterly, 51, September 1988a, 265-71.

_____. "Exit from the Group as an Individualistic Solution to the Public Good Problem in the United States and Japan." Journal of Experimental Social Psychology, 24, November 1988b, 530-42.

Yamagishi, T., and M. Yamagishi. "Trust and Commitment in the United States and Japan." Motivation and Emotion, 18, June 1994, 129-66.

Yamaguchi, S. "Collectivism among the Japanese: A Perspective from the Self," in Individualism and Collectivism, edited by U. Kim, H. C. Trandis, C. Kagitcibasi, S. Choi, and G. Yoon. California: Sage, 1994.

 Payoffs to A, B
 United States Japan
 A Chooses R
Treatment A Chooses L B chooses l, B chooses r A Chooses L
1 (9.75, 3.00) (3.00, 4.75) (10.00, 5.00) (1450, 450)
2 (7.00, 3.00) (3.00, 4.75) (10.00, 5.00) (1050, 450)
3 (9.75, 6.00) (3.00, 4.75) (10.00, 5.00) (1450, 900)
 A Chooses R
Treatment B chooses l, B chooses r
1 (450, 700) (1500, 750)
2 (450, 700) (1500, 750)
3 (450, 700) (1500, 750)
Notes: Payoffs in U.S. in dollars, payoffs in
Japan in yen. Japanese payoffs are higher than
the simple exchange rate equivalence would imply,
a result of opportunity cost student surveys.
 Experimental Results
 United States Japan
 A Picks R A Picks R
 A B B %
 Picks Picks Picks Secure
Treatment # Pairs L l r Play # Pairs
1 35 23 2 10 34 34
2 25 5 0 20 28 28
3 21 18 0 3 85.7 28
 A B B %
 Picks Picks Picks Secure
Treatment L l r Play
1 27 1 6 79.4
2 14 5 9 50
3 28 0 0 100
 Pairwise z-Statistics for Secure Play
Treatments US1 J1 US2 J2 US3
J1 1.27 -- -- -- --
 (0.102)
US2 3.49 4.53 -- -- --
 (0.0003) (0.00003)
J2 1.31 2.44 2.27 -- --
 (0.095) (0.095) (0.012)
US3 1.63 0.6 4.47 2.60 --
 (0.052) (0.2743) (0.00003) (0.0047)
J3 3.44 2.55 6.00 4.32 2.06
 (0.0003) (0.0054) (0.00001) (0.00003) (0.019)
Notes: Numbers in parentheses denote
the level of p-values. All z-scores are
given as positive numbers.
 Participants' Explanations for Their A Choices
 United States Japan
Treatment L R L
1 secure 14 reliant 15 secure 23(.098)
 fear 0 group max 5 fear 8(.003)
 rivalry 1 gamble 1 rivalry 1(.43)
2 secure 12 reliant 23 secure 18(.196)
 fear 0 group max 0 fear 6(.009)
 rivalry 2 gamble 10 rivalry 0(.063)
3 secure 26 reliant 10 secure 30(.363)
 fear 2 gamble 2 fear 15(.001)
 group max 4 group max 4(.375)
Treatment R
1 reliant 9(.023)
 group max 1(.026)
 gamble 1(.43)
2 reliant 17(.037)
 group max 9(.002)
 gamble 4(.022)
3 reliant 5(.033)
 gamble O(.057)
Note: The parenthetical numbers are the
p-values resulting from the equality of
proportions test for U.S. and Japanese
responses, conditional on A's choice.