% Students' attitudes toward social dilemmas in Argentina, Mexico, and the United States

Students' attitudes toward social dilemmas in Argentina, Mexico, and the United States

Sebastian Chaitas, Arlen Solodkin, and Jonathan Baron
University of Pennsylvania


We gave questionnaires about five hypothetical social dilemmas and two personal dilemmas to university students in the U.S., Argentina, and Mexico. Latins (students in Argentina and Mexico) did not differ from Americans (U.S. students) in expressed willingness to cooperate in social dilemmas, but Latins thought that others were less likely to cooperate. Belief that others would cooperate was generally correlated with willingness to cooperate. Latins more often believed that cooperation had long-run advantages for the self, and this belief was correlated with willingness to cooperate among Latins but not among Americans. We found no national differences in personal dilemmas.

In a social dilemma, each of several people is faced with a choice between two options, one of which is better for the individual and the other of which is better for everyone all together (Baron, 1994). For example, it is easier for individuals to drop their litter on the ground rather than putting it in the trash can, but pitching in is better for everyone. The selfish option is called defection; and the beneficial option is cooperation. Voluntary compliance with the law is often a form of cooperation. Rampant corruption can be seen as a breakdown of this form of cooperation. National differences in attitudes toward cooperation can affect, or be affected by, the level of corruption. On the other hand, spontaneous cooperation may be lower in cultures that emphasize individualism, even when corruption is not widespread.

The present study compares university students' attitudes toward cooperation in hypothetical social dilemmas in three countries, which differ in culture and level of corruption, Argentina, Mexico, and the U.S. One of our dilemmas concerns littering. The subjects were mostly college students. We also examined personal dilemmas involving conflict between short-term and long-term individual goals, such as eating unhealthy desserts. Such dilemmas are analogous to social dilemmas in many ways (Elster, 1985), but we see no reason to expect national differences in them for the three countries we examine. These situations thus serve as a kind of control for national differences in answering hypothetical questions in general.

In both kinds of dilemmas, we asked what the subject would do and what percent of others would ``cooperate,'' and we asked for reasons for both cooperating and not cooperating. Expectations of the behavior of others are of particular interest. These may reflect true national differences more than what people say they would do themselves. They may also play a causal role in people's own decisions about whether to cooperate or not.

We were interested in the three countries largely because of national differences in perceived corruption. On a scale of corruption, where 0 is most corrupt and 10 is least corrupt in 1995, the U.S. scored 7.79, Argentina 5.24, and Mexico 3.18 (Transparency International and the University of Goettingen, 1995. The scale is based on a combination of several surveys. At the ends of the scale were New Zealand at 9.55 and Indonesia at 1.94.) Corruption, real or perceived, may hinder economic development (Duncan, 1993).

Although our study is not relevant to questions about the origin of corruption, it may bear on the way in which corrupt behavior maintains itself. Specifically, people may defect more when they believe that others are defecting. This would occur if people were motivated by ``fear,'' that is, by a disproportionate fear of cooperating when others are defecting, as opposed to ``greed,'' a willingness to defect just when others are cooperating. Greed induces a negative correlation between the behavior of self and the expected behavior of others.

Some studies have found a positive correlation between cooperation and beliefs that others are likely to cooperate (e.g., Dawes, McTavish, & Shaklee, 1977; Dawes et al., 1986). In the 1986 study, a public good would be provided if a minimal number cooperated. Subjects cooperated, even when they believed that their own cooperation was redundant, provided that they believed that others would cooperate as well. Correlations, however, could result from effects of one's own behavior on beliefs about what others would do, rather than effects of beliefs on behavior. It is important to manipulate expectations experimentally in order to find out whether they affect cooperation.

The results of experiments in which subjects are given false feedback about others' behavior, in order to manipulate their expectations, conflict (Fleishman, 1988). It seems that subjects are influenced by both fear and greed. Sometimes one predominates, sometimes the other. The relative strength of the two motives may be affected by the framing of the dilemma. Fleishman (1988) used ``give some'' and ``take some'' games to study the effects of perceptions of others' behavior. In the give-some game, each subject in a five-person group received 100 points (each worth 1/3 cent) on each trial. The subject could give any number of those points to a central pool or stock. At the end of each trial, the number of points in the pool was doubled and divided among all the subjects. (Each subject would thus get 200 points per trial, if all contributed all their points.) In the take-some game, the pool received 500 points at the beginning of the trial, and each subject could take up to 100 points. The games are identical except for the description in terms of giving or taking. Not giving a point is the equivalent of taking it, and vice versa. To manipulate expectations about others, subjects were told after the first trial that the pool contained either 380 or 120 points. In the give-some game, subjects gave less on subsequent trials when they thought that the pool contained 380 after the first trial than when they thought it had 120. That is, behavior was opposite from what others were perceived as doing. In the take-some game, they took less when they thought that the pool had 380. Behavior was the same as what others were perceived as doing. Give-some subjects were apparently motivated by greed, take-some, by fear. Fear seems to prevent harmful actions but not harmful omissions. (These effects were small and should be replicated, but they help to make sense out of previous conflicting results.) We did not systematically manipulate the presentation of our situations, but most can be seen as concerning harmful acts rather than harmful omissions.

Personal dilemmas involve a conflict between our present and future self. It might be better for us in the present not to exercise, if we are busy and we have to go out of our way to do so. However, if we think in terms of the future, we will find that it is much better for us to do so. In general, people are biased in favor of their present self because it is less distant and claims more attention (Baron, 1994, ch. 24). By taking into account our present as more important than our future, we distance ourselves from our ultimate and most significant goals in life. Personal dilemmas involve decisions about health, education, and retirement.

Our questionnaire consisted of seven hypothetical situations involving either personal or social dilemmas. The questionnaires were given to university students in the U.S., Mexico, and Argentina. After each situation the subjects were asked to respond as truthfully as possible about their views towards the dilemma, their attitudes, their possible behavior, and their final decisions. In addition they were asked to give their views about the problem's magnitude, and to account for alternative behaviors. The situations were chosen to represent common choices faced in one or another of the three countries yet meaningful to students from all three countries. We expected that differences would arise more in social dilemmas than in personal ones, because social situations are those that differ the most between countries. Personal dilemmas may show a difference among individuals but not necessarily among countries.


The 76 U.S. subjects were students from the University of Pennsylvania and from the Philadelphia College of Pharmacy and Science, paid $6/hour for completing this questionnaire and others. Sixty-one Mexican university students completed the same questionnaire (in Spanish) without pay. Most were from the Universidad Anahuac and Universidad Autonoma de Mexico. Fifty-eight Argentinean students, mainly from Universidad de Buenos Aires and Universidad de Belgrano, also completed the questionnaire in Spanish without pay. The difference in payment was determined by the customary practices and expectations of the respective universities. We call the Mexicans and Argentinians ``Latins'' and the U.S. subjects ``Americans.''

The questionnaire consists of seven hypothetical situations, five social dilemmas and two personal dilemmas. (One additional situation, in retrospect, could not be classified, so we ignored it.) We classified situations as social or personal according to the extent to which each situation's outcome influenced society as a whole. In the social dilemmas, the decision clearly affected other people. In the personal dilemmas, the decision influenced mainly the decision maker. The questionnaire was written in English by the three authors, then translated separately for Argentina (by S.C.) and Mexico (A.S.) - the separate translations designed to accommodate country-specific linguistic features. The two Spanish-speaking authors (S.C. and A.S.) then checked each other's translations against the original English.

Each situation was accompanied by a set of seven or eight questions, depending on the case. The description of the hypothetical situations will follow. The questions asked after each of the cases were practically the same for the eight situations, and they are described after the situations. We describe the social dilemmas first, but the numbers indicate the order in which the subjects read them.

Social dilemmas:

2. Imagine yourself living in a very polluted city, a city with so much pollution that many new laws have been made to fight it. One of these laws requires everyone to have their car inspected every 6 months, and if the emission levels are high, the car has to be fixed. Suppose your car fails the inspection, and imagine you have two alternatives: One is to pay $100 to have your car fixed. This will not make it less likely that you will need it fixed after the next inspection, six months later. The other is to pay $45 to the mechanic so that he gives you a permit to drive for six months.

4. Imagine that you are walking in the middle of a park eating and drinking something. When you finish your food you do not see any trash-can nearby to throw your paper cup and bag. The only one that you see is about 50 yards away.

5. You are a customs officer working in an international airport. A tourist brings a new computer which costs around $3,000. He offers to give you $500 if you let him through, so that he doesn't have to pay the 50% tax on electronic imports.

6. You have to validate your passport at a local office. Because of inefficiency and insufficient staff, there is a 5-hour line. An officer offers you to validate it for you in twenty minutes if you give him a $20 tip.

8. You are a government official in charge of importing a certain product for your country. You get two offers from foreign companies. One offers to sell you the product you need for $10 million. The other, offers you the exact same product for $13 million but also gives you a $1 million gift if you decide to buy it from them.

Personal dilemmas:

1. Imagine that you like to eat lots of junk food. Your doctor finds that you have high levels of cholesterol and recommends that you go on a low fat diet, which means no more junk food. You are then presented with this kind of food in a social environment.

3. You wake up early to go to work, spend all day working and then go home very tired late in the afternoon. You know that exercising is good and that everybody recommends it. Your doctor recommends that you start getting some regular exercise, such as running a couple of miles every day. Imagine that it is the morning after your visit to the doctor.

Situation 7 was not analyzed (except for the justifications). It involved paying extra money to the ticket seller for good seats in a movie house. (We originally thought of it as a social dilemma, but we then realized that it was really just a marketing device.)

The questions that followed situation 2 (modified accordingly for other situations) were:
a) What thoughts would go through your head?
b) What do you think you would finally do?
c) Imagine someone like you who would make the opposite decision. What thoughts do you think would go through that person's head?
d) What percentage of people in this situation do you think would pay the $100?
e) Would you be more likely to pay the $100 if you thought that almost everyone else would do it in the same situation?
f) In the long run, is it better for you to pay the $45 or the $100?
g) Is it better for others for you to pay the $45 or the $100?

In situations 5 and 8, subjects were also asked, ``Would you take the money if you knew that you would not get caught?''


Latins (Mexican and Argentinians) differed from Americans (U.S.) in being less optimistic that others would cooperate in social dilemmas, but they were no less likely to say they would cooperate themselves. Latins were much more likely, however, to see cooperation as being in their long-term self-interest. Females were more likely than males to say that they would cooperate themselves. Details follow.

In general, we analyzed the results separately for social and personal dilemmas. Since our main interest was in individual differences, we formed summary indices across the relevant situations for each subject.

Table 1 shows the responses for questions b, e, f, and g. The percents are those of people giving the ``good'' answer (cooperate or exert self-control) or saying that this behavior is better for themselves in the long run or for others. (We counted only clear answers, treating``maybe'' and other such responses as missing data.) Subjects saw the personal behaviors as more beneficial to themselves in the long run than the social-dilemma behaviors, and as less beneficial to others, thus indicating that they distinguished the situations somewhat. Note that the low rate of cooperation in situation 6 is easily understood in terms of the large self-sacrifice required for cooperation, a five-hour wait instead of 20 minutes. Even here, however, most subjects understood that their behavior affected others.

- Insert Tables 1 and 2 about here. -

Basic results for social dilemmas

Table 2 presents individual differences as a correlation matrix for social-dilemmas. The numbers entered in each correlation are means of all relevant non-missing items on each question (b, e, f, g and d) for each subject. The variable USNAT is 1 for U.S.\ subjects and 0 for Latins. Our strategy was to use this variable as the main nationality index. We report differences between Argentina and Mexico separately. Sex was coded 1 for males, 0 for females. (Of course, correlations with one dichotomous variable are equivalent to t tests, but we report them as correlations to facilitate comparison with other correlations in Table 2.)

Latins were less inclined to believe that others would cooperate in social dilemmas, although the subjects themselves showed no national differences in expressed willingness to cooperate. The differences in beliefs were significant in all five situations (p<.02) These results may reflect reality. Since fewer Latin Americans go to college, the Latin sample was less representative of the population as a whole. Latins may be truly less cooperative, but Latin college students may be more cooperative than their fellow citizens. Alternatively, Latins may have more optimistic bias than Americans about whether they would do the right thing.

The correlation between cooperation and beliefs about others' cooperation is positive but not significant. However, Fleishman's conclusion that the correlation is positive only when defection is an action implies a positive correlation only for situation 6, involving the ``tip'' to avoid waiting. The correlation is indeed positive for this situation (r=.32, p<.0005). It is also positive and significant for situation 4 (r=.16, p=.025) and situation 5 (r=.28, p<.0005), and it is not significantly different from zero for situations 2 and 8. In sum, people tend to conform to what they think others are doing, in some situations, but we find no situations where they tend to do the opposite.

Moreover, cooperation is also significantly correlated with questions f (long-run) and g (effect on others): cooperators tended to think that cooperation helped others and themselves in the long run. Latins are also much more likely to see cooperation as being in their long-run self-interest. When these two variables, f and g, were included with beliefs (question d) in a multiple regression with cooperation (question b) as the dependent variable, both beliefs and long-run benefits were significant (p<.028) and effect on others was almost significant (p=.056 two tailed).

When USNAT is added to the regression, it is also significant, with Americans more cooperative when other predictors are controlled. Recall that, without the other predictors, USNAT is not a significant predictor of cooperation. It appears the overall equality of Latins and Americans was the result of two factors pulling in opposite directions. Latins believe that their compatriots are less cooperative, so they are less cooperative, but they also believe that cooperation is in their long-term self-interest, so, as a result, they are more cooperative than they would be otherwise. When both of these factors are controlled statistically, Americans are significantly more cooperative, perhaps because people are more influenced by the behavior of their compatriots than their belief ratings would indicate.

In an additional analysis, we regressed cooperation (b) on long-run (f), effect on others (g), beliefs (d), USNAT, and the interactions of the last three with USNAT. Only the interaction between USNAT and long-run was significant (p=.001). The correlation between long-run and cooperation was higher for Latins than for Americans. In fact, while the correlation was .56 for Latins (p<.0005), it was nonsignificant for Americans (r=.06). Not only do Latins think that cooperation is better for themselves in the long run but their opinion about this effect also has a stronger relation to their own willingness to cooperate.

The correlation between long-run and cooperation was .68 for Argentina and .45 for Mexico. Both were significant, but the former was significantly higher at p=.02. But Mexico and Argentina did not differ in overall endorsement of long-run, and they both different equally from the U.S.

Although beliefs about others differed between the Latins and the Americans, they also differed significantly between Mexico and Argentina (r=.22, p=.007). The mean expected percent of cooperation by others for the three countries, respectively, was 49% for the U.S., 38% for Mexico, and 32% for Argentina. This difference between Mexico and Argentina is opposite from what we would expect from the corruption ratings. However, all the differences we found between Mexico and Argentina are consistent with the Mexican sample being more Americanized.

Basic results for personal dilemmas

Behavior in personal dilemmas (question b) was not correlated with sex, nationality, or beliefs about the behavior of others. It was correlated only with beliefs about the effect on others (question g; r=.15, p=.030), but a multiple regression predicting behavior from beliefs, long-run benefit, and effect on others - analogous to the same regression done for cooperation - was not significant overall. Although these results are based on two situations rather than five, they make it more likely that the results for social dilemmas were specific to these kinds of situations.


We coded justifications for defection according to the following categories. These justifications were responses to either question a or c, depending on the subject's willingness to cooperate in question a. Although inter-rater reliability was tested and was reasonably high, different countries were scored by different raters, so we must take national differences in justifications as suggestive at best. The justifications are ordered in terms of their frequency of use, shown in parentheses (out of a maximum possible of 8, since all 8 situations were included).

LITTLE. Doing the wrong thing just a little will not hurt. (0.47)

RIPOFF. The situation described is a rip-off: the rule being broken is unfair or bad. (0.37)

DROP. Drop in the bucket: if too many people are doing the wrong thing, the action will not make a significant difference. (0.26)

FEAR. It is OK to do the bad thing because everybody does it (or most people do it). (0.20)

NOBODY. Nobody is hurt by breaking the rule. (0.20)

COOL. It is cool to beat the system. (0.15)

GREED. If too many people do the good or the right thing I can do the wrong thing and it will not significantly affect others. (0.03)

FEAR justifications were significantly more frequent than GREED justifications (t=4.33, p<.0005). This is consistent with our finding that people are more willing to cooperate when they think that others would cooperated.

There were no significant sex differences in frequency of the justifications listed above (in a logistic regression predicting sex from the frequency of each measures for each subject). USNAT, however, was associated positively with COOL (to beat the system; r=.26, p<.0005) and negatively with RIPOFF (OK to break the rule because the rule is wrong; r=-.18, r=.006) and with NOBODY (is hurt; r=-.26, p<.0005; the overall predictability of USNAT was confirmed by logistic regression before testing individual predictors). These correlations are suspect, however, because different raters scored the Americans and Latins.

Individual differences in willingness to cooperate in social dilemmas (question b) were predicted by the justifications in a multiple regression (p=.05, removing the effect of USNAT so as to remove the effect of different coders). Cooperation was negatively correlated with LITTLE (just a little won't hurt; standardized coefficient -.14, p=.045) and NOBODY (is hurt; coefficient -.15, p=.033). Behavior in personal dilemmas was not related to justifications.

We analyzed one justification for cooperation because of its interest: fear of punishment, embarrassment or being noticed (frequency 0.43). Latins used this justification more often than Americans (r=-.17 with USNAT, p=.012), but it was not correlated with any other variables of interest.


The most important results concern the correlates of the tendency to cooperate and national differences in these correlates. Cooperation increased with the belief that others will cooperate (in some situations) and with the belief that cooperation is in one's long-run self-interest. The latter belief may be an error, especially in large-scale social dilemmas involving strangers. If so, it may be a beneficial error, except that it may also prevent people from seeing the need for coercive regulation by making them think that education will suffice to insure cooperation (Baron, 1995). Belief that others will cooperate may also be irrelevant to the benefits (to others) of cooperating oneself (Baron, 1994).

Belief in long-run benefit is correlated with cooperation in Latins only, and this belief is more prevalent among Latins. Cultures may differ in the kinds of reasons they take to be good ones for cooperating. Americans may be more prone to adopt a Kantian (Protestant) view that we are obliged to do the right thing irrespective of our self-interest. Latins may be more prone to generalize from a Catholic moral tradition that views good works on Earth as rewarded in the afterlife. In this perspective, even long-term benefits on Earth become more acceptable as a justification for moral behavior, even if the benefits are imagined rather than real.


Baron, J. (1994). Thinking and deciding (2nd Ed.). New York: Cambridge University Press.

Baron, J. (1995). The long-run fallacy as a reason to cooperate in social dilemmas. Manuscript.

Dawes, R. M., McTavish, J., & Shaklee, H. (1977). Behavior, communication, and assumptions about other people's behavior in a commons dilemma situation. Journal of Personality and Social Psychology, 35, 1-11.

Dawes, R. M., Orbell, J. M., Simmons, R. T., & van de Kragt, A.\ J. C. (1986). Organizing groups for collective action. American Political Science Review, 80, 1171-1185.

Duncan J. (1993). Corruption seen still plaguing Latin America. Business Week, June 16.

Elster, J. (1985). Weakness of will and the free-rider problem. Economics and Philosophy, 1, 231-265.

Fleishman, J. A. (1988). The effects of decision framing and others' behavior on cooperation in a social dilemma. Journal of Conflict Resolution, 32, 162-180.

Knez, M., & Camerer, C. (1994). Creating expectational assets in the laboratory: Coordination in `weakest-link' games. Strategic Management Journal, 15, 101-119.

Parks, C. D., Vu, A. D. (1994). Social dilemma behavior of individuals from highly individualist and collectivist cultures. Journal of Conflict Resolution, 38, 708-718.

Transparency International and the University of Goettingen.\ (1995). International corruption rankings. http://gwdg.de/ uwvw/icr.htm

Table 1
Percent ``good'' answers to questions for each dilemma.

Would do Conform Long run Affect others Belief about others
Question b e f g d
Social dilemmas
2 Pollution 80.0 41.7 86.9 76.4 47.7
4 Litter 91.5 48.6 73.4 96.2 42.0
5 Customs 65.0 39.7 67.9 84.1 36.6
6 Passport 22.4 26.9 38.0 64.5 28.0
8 Bribe 69.7 30.5 66.7 77.1 Personal dilemmas
1 Junk food 73.0 51.9 99.1 57.3 39.7
3 Exercise 68.2 42.1 96.8 42.2 45.0

Table 2.
Correlations relevant to social dilemmas (* p<.025, ** p<.01).

Sex USNAT B (would do) E (conform F (long run) G (affect others)
B (would) -189** -041
E (conform) 008 081 -018
F (long run) -018 -754** 319** 017
G (affect others) 060 -103 220** 166* 257**
D (belief) 053 434** 115 158* -281** 023

File converted from TEX by TTH, version 0.7.