To What Extent Do Latin Americans
Trust and Cooperate?

Field Experiments on Social Exclusion in Six Latin American Countries

The puzzle of cooperation among humans remains central and relevant. In fact, in 2005 Science magazine listed “How did cooperative behavior evolve?” as one of the twenty-five most relevant scientific puzzles to be solved in the next quarter century. However incompletely understood, cooperation and collective action have represented pervasive and ubiquitous features of human experience across time, place, and income level. Cooperation—and the failure to cooperate—have affected societies in ways that range from the organization of hunters and gatherers and “tragedies of the commons” in the use of shared resources to automobile traffic control and responses to global warming. Moreover, many social interactions involving cooperation risk losses of efficiency as a result of externalities or problems with the provision of public goods, and some individuals “free-ride” on the contributions of others. When there is a lack of trust such free-riding intensifies, thus reducing opportunities to produce socially efficient outcomes and accentuating inequalities in the distribution of outcomes.

Free-riding and lack of coordination represent daily challenges in the lives of communities. For instance, when the state fails in the performance of its expected functions, communities attempt to solve collective action problems themselves, and in multiple ways. Households contribute labor to starting or maintaining local projects that benefit their neighborhoods, and neighborhoods contribute local funds that pay for security or playground maintenance. Child care, recreation parks, water provision, and street cleaning are all examples of projects for which groups contribute privately to public goods. Groups additionally organize to deal with problems other than those that arise from collective action, such as risk management involving credit, natural disasters, political violence, and crime. In these cases the formation of groups to deal with risk involves in itself a collective action problem, and payments for cooperation can be distributed across the group. Vulnerable groups in societies are more likely to face the challenges of collective action, either because they are left out of the regular channels that socie­ties use to cooperate or because they face more vital risks in their daily lives and have a greater need to pool than members of other groups.

Free-riding and coordination problems have been studied by sociologists, psychologists, and more recently by economists. The last have engaged in highly controlled experiments involving relatively small groups of individuals whose members are typically given a particular sum of money and allowed to invest in a group exchange or keep the money. If participants invest in the group exchange, the returns on that investment will depend on what the group as a whole invests in. The experiment is designed so that the private return from keeping the money exceeds the private return from the group exchange; the social return of the group exchange, however, is higher than that of keeping the money. This experiment yields a dominant strategy in which an individual will contribute zero to the group exchange while hoping that the individuals will invest in the group exchange, even though that does not represent an optimal allocation (see Andreoni, 1988, and Marwell and Ames, 1979, among others). In other words, experiments try to replicate cooperation and related dilemmas as predicted by economic theory.

Economic experiments in this area have so far yielded two key results. First, economic theory overestimates the prevalence of free-riding. In fact, even though experiments find that outcomes are closer to the free-riding result (where nobody contributes) than to the socially optimal one (where everybody contributes), experiments show that individuals still contribute more than would be implied by pure self-interest (Rabin, 1993; Andreoni, 1995). This is particularly true in “one-shot,” nonrepeated games, which contrasts with predictions based on a strong version of the free-riding hypothesis (Dawes and Thaler, 1988). A second and similar result is that violations of dominant strategies diminish with repetition and with game experience (e.g., Andreoni, 1988; Isaac and Walker, 1988; Kim and Walker, 1984). These findings have led researchers to explore possible explanations including “kindness,” concern with reputation, and confusion on the part of individuals in regard to the rules of the game and the consequences of their decisions (Palfrey and Prisbrey, 1997).

This chapter attempts to build on existing research by studying the microfoundations and mechanisms that may affect the possibility of collective action and group formation by different social groups. Of particular note is the reporting of results from a field approach, using survey and experimental methods, that focuses on the behavioral aspects of the collective action problem while taking into account the social and economic contexts in which microinteractions happen. The research summarized here has involved the direct observation of individuals facing problems of trust, collective action, and uncertainty under different levels of social heterogeneity and exclusion. The experimental design of this project thus captures key dimensions of problems at the intersection of trust and exclusion and makes it possible to derive lessons on collective action and group-oriented behavior in societies.

COOPERATION AND SOCIAL EXCLUSION

Over time human societies have attempted to minimize losses resulting from problems of collective action by harnessing the conflict between individual and social outcomes through incentives, generally in the form of norms and laws.[1] Although the possibility of cooperation within a group is determined by multiple factors, one of the most controversial is group heterogeneity. Some argue that heterogeneity offers the additional incentives necessary for a small subgroup to be interested in providing a public good (Olson, 1965; Bergstrom, Blume, and Varian, 1986), yet others claim that heterogeneity creates difficulties in agreement and problem solving (Alesina and La Ferrara, 2000).

Since the benefits of economic and social progress are often unequally distributed, social heterogeneity is intrinsically linked to the problem of exclusion. Nonetheless, winners and losers, haves and have-nots, and the included and the excluded can engage in mutually beneficial interactions if the collective action problem is solved. The importance of addressing this problem can hardly be overestimated, as few individuals have the option of living and working only with persons like themselves. Vulnerable individuals, for example, must interact with nonexcluded groups and individuals in environments including labor, housing, and informal and formal credit markets. Likewise, on other occasions heterogeneous groups share common spaces and must make decisions that affect their common interests, quite often with asymmetric stakes across subgroups. Riding public transportation, using public parks, participating in debate on a public issue, and voting all represent instances in which members of a society must make decisions that result in varying benefits and costs depending on the actions of other group members. Social scientists often refer to decisions of this type as a game, in order to facilitate the study of behavior in these situations.

Cooperating or forming groups to produce a group-beneficial outcome is usually costly in monetary or other terms. Sometimes a coordination game is involved in which each individual will benefit more if everyone else behaves in a socially optimal way, and the resulting payments thus drive individuals towards the best outcome without conflicts between individual and group interests. Other instances involve collective action games in which the optimal individual behavior would be not to cooperate, although everyone in the group would benefit if everyone cooperated. In either case, the group needs to find—and create—conditions under which individuals will choose to make decisions that benefit all members of a group even when those decisions are individually costly. These conditions are defined by several behavioral issues. Individuals may make decisions, for instance, according to a sense of group or subgroup affiliation, or social distance from or sympathy toward others in the group. Their personal evaluation of the benefits and costs of forming a group or cooperating in a collective action dilemma may be mediated by their expectations of the actions of others as well as their valuation of the distributional and efficiency consequences of their actions.

Formal and informal institutions play a major role in shaping individuals’ decisions, as they provide key information for a person who is bearing the cost of group-oriented action. Individuals use information from the context in which a game operates in order to inform their decisions and therefore provide the best possible benefits for every player. It should be noted that those benefits may include an increase in payments to others, an increase in social welfare, or a decrease in inequality, because his or her preferences involve prosociality as part of his or her interests. In any case, the individual will collect information from his or her personal, group, and social contexts and transform the relative payments of the game when making his or her decision. Figure 7.1 illustrates how such a cognitive process may operate (Cárdenas and Ostrom, 2006), involving layers of information obtained from various contexts. The bottom left layer of information in the figure (“static game layer”) involves the initial calculations about benefits and costs of a decision at a certain time t as a one-shot game, within the formal rules and action sets of the individual. Within our framework, individuals do not base their decisions to engage in collective action solely on this layer. They also consider the aspects of other elements, including dynamic aspects of the game in previous (t–1) and future (t+1) rounds (“dynamic game layer” in the figure) (e.g., whether the game will be repeated with the same players, and previous experience with the same players). Also, players will consider the composition of the group with whom they are playing (“group context layer”) (e.g., an individual might be more willing to cooperate with certain players in his or her group than others, based on group membership, identity, or social distance). Finally, at the top right of Figure 7.1 (“identity layer”), the framework suggests that there might also be individual normative aspects or values that constrain the initial action set (e.g., certain values may eliminate antisocial or cheating behavior in games irrespective of their counterparts or the context). As will be shown in the results of our experiments, and as has been shown in previous behavioral literature, individuals seem to use these layers of information when deciding to trust others or to cooperate with others. The key proposition of this behavioral and institutional approach to the problem of collective action and group-oriented behavior is that individuals use information from their personal, group, and social contexts to transform the game situation and make the best decision according to their own personal and other-regarding preferences.

Solving the prisoner’s dilemma, the tragedy of the commons, or any collective action dilemma requires individuals to trust their partners in the interaction. Trusting others under incomplete contracts, however, involves the possibility that the trusting action results in receipt of no benefits from the trustees and creates net losses for the trusting person. If the trustees reciprocate, though, the group increases its social net welfare. If the game is repeated, players can engage in a virtuous circle of trust and reciprocity, building a reputation for being trusting and trustworthy and collecting information about the trust and trustworthiness of others in the group (Ostrom, 1998). If the game is played only once, players may still be willing to cooperate if institutions and personal characteristics provide sufficient positive information for players to involve themselves in group-oriented behavior.

The uncertainty of the intentions and actions of the other actors is a crucial part of the problem of collective action. Individuals may have information about past actions of specific individuals or more general patterns of past behavior by groups, as well as information on the social norms that usually guide the behavior of those interacting with them. As some uncertainty remains nevertheless, understanding the willingness to trust, cooperate, or engage in costly group-oriented behavior involves understanding individuals’ risk preferences (Bohnet and Zeckhauser, 2004; Ashraf, Bohnet, and Piankov, 2006).

The behavioral literature on collective action suggests that a number of factors feed the virtuous circle of trust, reputation, and reciprocity (Figure 7.2). Several of these factors are associated with social exclusion and group inequality. The seminal work of Olson (1965) discusses how heterogeneity may affect cooperation and collective action, and social exclusion may affect the cycle of trust, reciprocity, and reputation in several ways. According to Ostrom’s approach, exclusion may create not only different interests, but also different endowments and resources for those engaged in a group activity. Under these circumstances face-to-face communication may be more difficult or impossible if some individuals in the group are excluded, and in heterogeneous groups whose members know little about one another, there is likely to be only limited availability of information on the past actions of others. In-group and out-group effects thus make the development of shared norms across excluded and included groups particularly costly.

Equally relevant is the question of whether certain homogeneous groups are more or less likely to engage in cooperative behavior than others because of their socioeconomic level, wealth, or human capital. Some may argue that the poor are less likely to solve a collective action dilemma because their opportunity cost of cooperating is much higher than that for those less constrained by income or wealth. On the other hand, some would contend that the lack of assets or smooth income creates conditions for the poor to rely on their social networks and on others like them to provide key goods and services offered by neither the state nor private providers based on their market possibilities.

EXPERIMENTAL SETUP

The research undertaken for this chapter attempts to study the interaction between social exclusion and collective action in Latin America using a field experimental approach. The project began with the identification of a representative sample of individuals from six cities in the region who were then asked whether they would be willing to participate in a set of experiments that involved economic incentives. The full sample consists of more than 3,000 observations, or roughly 500 individuals per city from different backgrounds, socioeconomic levels, and age cohorts and both sexes in Bogotá, Buenos Aires, Caracas, Lima, Montevideo, and San José.

A team of researchers with experience in survey and field methods was selected to undertake the experiments and surveys in each city, and to guarantee homogeneity in the application of experimental protocols, those researchers in charge of each city participated in a training workshop at the launching of the project.[2] This workshop provided participants with a uniform approach to implementation and related fieldwork details such as sampling procedures, timing of actions (i.e., invitations, presurvey, experiments, postsurveys), elements to be included in experimental sessions, and the construction of questionnaires. Each survey team agreed to conduct twenty-five experimental sessions with an average of twenty participants each. [3]

With the sampling quotas defined, the first step of the fieldwork consisted of inviting individuals to experimental sessions. The sessions were arranged so that at least three sessions per city included only individuals from high-income strata and at least three other sessions included only individuals from low-income strata; the rest combined individuals from all strata. Around thirty individuals were invited for each session, under the assumption that approximately one-third would not show up for the session, with each experimental session allowed to go forward with roughly twenty participants.

Potential participants were invited several days before the scheduled sessions, and at the time of invitation, individuals were asked a set of basic demographic questions in order to enable the researchers to fulfill the sampling quotas described previously.Participants were additionally promised remuneration and provided with information on the expected monetary gains from their participation in the experiments. The day before each experimental session, the invited participants were reminded of the invitation with a phone call or home visit, and research staff worked with potential participants to arrange transportation if necessary. On the day of the sessions, the participants were welcomed by experimental teams, and at the appointed time, the sessions started. Following the experiments, participants in each session completed surveys designed to collect additional sociodemographic information and determine their attitudes, beliefs, and preferences regarding social exclusion, discrimination, minorities, and prosocial norms. To reduce the possibilities of idiosyncratic measurement error due to individuals’ reading ability, the surveys were administered by the monitors of the experiments and supported by a group of pollsters specially trained for this purpose. After participants completed the surveys, they were paid, based on the results of one of the experiments (randomly selected by the monitors).

As one of the main goals of the study is to observe the effect of social heterogeneity on individuals’ decisions, information on the socioeconomic composition of the groups in each particular session was made as salient and clear as possible. The participants met throughout the session in one room where they were able to see each other, although they were not allowed to communicate during the session. As the sessions progressed, participants received information about their peers, depending on the particular activity.

More than 3,000 people participated in 148 sessions in six cities, providing a unique data set that combines detailed data from their socioeconomic and demographic backgrounds with behavioral data from their decisions during the experiments. Each of the city teams conducted sessions of various group sizes, from ten to thirty-nine people, and each session followed the same protocol, with the same sequence of activities. This is the most comprehensive experimental data set to date in Latin America, given the number of countries included and the replicability of the designs in each city.

Table 7.1 provides basic demographic statistics of the sample, by city and overall, comparing it with a representative sample of individuals extracted from the national household surveys of the countries in the project (restricting the computation to the capital cities). The comparison reveals remarkable similarities between the experimental samples and those of the national household surveys; the only potential difference of note is that individuals in the experimental samples were slightly younger and more educated than the average individual in their respective cities.

Interactions among participants took place in a controlled setting where it was possible to observe how incentives, institutions, and norms may affect behavior, and the experiments were conducted in a manner that allowed measurement of how the degree of group heterogeneity affects individual decisions and group outcomes. The design follows the approach shown in Figure 7.3, which proposes a series of relationships among the key explanatory factors of collective action.[4] In this framework the working hypothesis is that exclusion may affect trust and prosocial behaviors such as group formation and collective action in various ways, according to the relationships in Figures 7.1 and 7.2. Exclusion may negatively affect trust and voluntary contributions to a group project, as it reduces the availability of information about past actions and implies more asymmetric interests among players. Exclusion also reduces the sense of belonging to a group-oriented activity and therefore can affect voluntary contributions and group formation. In-group/out-group effects have additionally been reported (Marwell and Ames, 1979; Ostrom, 1998) as key factors in explaining collective action in groups.

The experiments provide key information about individual behavior and group outcomes regarding the possibilities for and limitations of collective action in groups, as well as clues regarding individuals’ motivations and limitations in making decisions that may solve collective action problems. Ultimately, the results help to provide an understanding of how social heterogeneity and social exclusion affect possibilities for actions that create greater benefits to groups.

The experimental design was based on four activities (A.1–A.4) in which participants made individual decisions that had economic outcomes for themselves and for the others in their group.[5] Participants were informed before the first activity began that they would receive a payment based on the outcome of one of the activities (randomly selected by the experimenter). The four activities were undertaken in a session of two to three hours. The sequence of activities was as follows:

• Experiment 1 (Trust Game): In this game all participants in each session were randomly assigned to pairs, half of the participants assuming the role of player 1 and the other half player 2. The two groups of players were located in different rooms. Identities were never revealed, but each player was given information about demographic characteristics of the other player in his or her pair: age, gender, education, and an indication of socioeconomic level of the neighborhood (high, medium, or low) where the player lived.[6] Both players received an endowment, and player 1 was then asked to decide how much of this endowment to send to player 2. The amount sent was then tripled on its way from player 1 to player 2. In the other room, player 2 was asked to decide the amount to be returned to player 1 for each possible offer from player 1. Each player knew in full the rules governing the operation of the game before being asked to make his or her allocation decisions. Immediately before making his or her decision, each of the players was asked to predict the decision that would be made by the other player. After both players had made their decision, the choice of player 1 was matched with the corresponding response for that amount by player 2, determining the outcome of the game.

This experimental game permits the measurement of the extent to which an individual trusts another person of similar or different socioeconomic characteristics, and whether the actions and characteristics of that individual affect the response of his or her partner in the game. In short, the game measures trust and reciprocity. Higher offers by the first individual are interpreted as signals of trust, and higher returns from the second individual are taken as signals of reciprocity. The game-theoretical prediction for this game is that player 1 will send an offer of zero, as there is no assurance that player 2 will return any amount. Replications of this game around the world have shown that people on average send half of the initial endowment to player 2, and that the returns from player 2 to player 1 generate a net positive return for player 1 of about 10–20 percent over what was originally sent.

• Experiment 2 (Voluntary Contributions Mechanism): In a second experiment, participants gathered in a single room and participated in a voluntary contributions mechanism (VCM) or public goods game. Each player was given a token that could be kept or invested in a group project. If a player kept the token, he or she earned an amount (e.g., $10). If the player invested the token in the group project (i.e., if he or she was a “cooperator”), his or her token and the rest of tokens in the group account each yielded a return of $1 to every participant in the group. A player who kept the token also received $1 times the number of tokens in the group account. Before they made their individual (private) decision whether to contribute to the group, the monitor announced orally and wrote on a board the gender, age, education, and socioeconomic composition (i.e., the number of individuals from high, medium, and low socioeconomic neighborhoods) of the group. Also, in order to capture expectations, the monitor requested that every participant write his or her prediction about the fraction of group members who would be cooperators.

The public goods or VCM experiment captures a similar dimension of trust, though in this case regarding a group instead of an individual, by measuring willingness to contribute a token to a public good and provide benefits to all group members. The decision to contribute to the group increases the benefits for all, but not contributing will always yield greater individual payments and thus provide an incentive to free-ride. Full cooperation yields greater payments to everyone than if full free-riding occurs, and the gains from cooperation increase with the number of players. A key element of the game is that no player knows in advance how many of the group members will contribute. The players know only general socioeconomic characteristics of the other players before making the decision.

• Experiment 3 (Three Risk Games): Each player individually made decisions in three games measuring individual attitudes in regard to risk, ambiguity, and losses.
o The first game, measuring risk aversion, offered the participants a choice among six 50/50 lotteries with known probabilities and known outcomes that ranged from a sure low payment to an all-or-nothing higher payment (with the lotteries in between increasing gradually in expected value and in the spread of the low and high payment).
o The second game, measuring ambiguity aversion, offered the same payments for the six lotteries mentioned. In contrast to the first game, however, individuals did not know the exact probabilities but were informed only that at least 30 percent of the chances were for the low payment and also at least 30 percent for the high payoff.
o The third game, measuring loss aversion, used six lotteries with 50 percent probabilities for each outcome but included the possibility of negative payments in some cases.[7]

The individual risk games are based on three components of risk behavior. These three games allow a distinguishing of risk attitudes in terms of risk aversion, ambiguity aversion, and loss aversion. The first game measures risk aversion, based on known probabilities and known outcomes for six 50/50 lotteries. Choosing lotteries with lower payments can be interpreted as greater risk aversion. The second game measures risk ambiguity, and the third, loss aversion. The purpose of this activity is to generate measures of risk behavior in order to link them with trust and cooperative behavior.

• Experiment 4 (Risk Pooling): Each player was given the opportunity to choose whether to form a group (i.e., a “subgroup” of the group participating in the session) to share equally the gains from another risk aversion game or to play the risk aversion game individually. Once players decided whether to form the group or not, the total number of people forming the group was announced, and the players then made their decisions in their individual risk game.

This game measures individuals’ willingness to join a group and to accept an even distribution of payments after choosing again a lottery like those available in the first of the individual risk games. As in the VCM game, the purpose of this game is to determine whether an individual will base his or her decision to join a group on the potential socioeconomic composition of the group (in this case, based on the socioeconomic composition of the group in the session). Again, players were not allowed to communicate with one another and were given only basic information about the composition of the session group (age, education, gender, and socioeconomic composition). It should be noted that in this game the most profitable group outcome would occur when all players joined the group and chose higher-risk lotteries (at 50 percent chance of the high payment, the expected value should yield greater payments to everyone in the group).

At the end of the last activity the monitor randomly selected one of the activities to be paid, and while one monitor calculated individual earnings and privately called upon each participant to distribute them, the remaining monitors interviewed each participant, filling out an individual survey with detailed information about socioeconomic characteristics and attitudes, beliefs, and preferences in regard to various dimensions of social exclusion.

BASIC SOCIOECONOMIC CHARACTERISTICS

Before turning to the experimental results, it is worthwhile to consider participants’ answers to the questions on attitudes, beliefs, and preferences in regard to trust, collective action, and exclusion that they were asked after the completion of the experimental activities. The participants in the six cities of the project reported low levels of participation in organized groups (Figure 7.4). Cultural and athletic organizations accounted for the group participation of the largest number of participants, followed by religious groups; only about one out of seven individuals, however, reported participating in one of these types of organizations. Interestingly, state-sponsored and ethnic organizations had the lowest participant-reported participation rates among the choices.

After the question regarding group participation, participants were asked to agree or disagree with a series of statements regarding the scope and scale of the welfare state. Figures 7.5a and 7.5b show the results, separating positive from negative statements about the welfare state. In general, the positive statements that met with the most agreement involved equality of opportunities, lack of discrimination, and collective welfare. Statements proposing higher tax collection for redistributive purposes met with the least agreement among the positive statements.

Participants were additionally surveyed in regard to unmet financial or career desires (Figure 7.6). The most common unmet desire was buying a house, cited by one in three participants. Other important unmet desires were obtaining a bank loan, studying, and working, cited by one in every four participants.

Participants’ perceptions of rights reveal areas of both frustration and satisfaction (Figures 7.7a and 7.7b). When participants were asked to identify areas in which they believed their rights had been violated, the top three choices from a list of twenty items were the opportunity to have a decent job, freedom of opinion, and justice and equal treatment under the law. The least cited areas were voting rights, freedom from torture, freedom of association, and the right to run for public office. As a matter of fact, almost 80 percent of participants reported having voted in their country’s most recent presidential election. When participants were asked why they believed their rights had not been respected on at least one occasion in the preceding five years (Figure 7.8), the most frequently cited reasons were lack of connections, lack of money, and age. These results are consistent with those found in other opinion surveys of the region (e.g., Latinobarometer).

Regardless of their own situations, participants expressed a significant level of agreement on what groups in society are most vulnerable (Figure 7.9). Two-thirds of those surveyed cited the elderly as the most vulnerable social group. Approximately one-third of respondents viewed children as the most vulnerable group.

Finally, the participant survey explored the notion of social distance and perceived causes of social conflict (Figure 7.10). Political differences, cited by almost half of respondents, represented the leading perceived cause of conflict. Approximately one-third of respondents cited differences in income and level of education as a cause of conflicts.

LESSONS LEARNED FROM THE EXPERIMENTS

The following sections describe the most relevant and robust results that emerged from the group-level and individual data generated by the experiments described in this chapter.[8]

Finding 1: Latin Americans Are Willing to Trust and Cooperate

Consistent with previous observed experimental behavior, the game-theoretical prediction that people in the trust game would not send any amount as either player 1 or player 2 is rejected. Only 12 percent of the observed decisions made by the individuals who participated as player 1 involved sending player 2 nothing. The average offer was 44 percent of the initial endowment, and the median offer (made in 32 percent of the decisions) was 50 percent of the initial endowment. Although social efficiency is maximized when player 1 sends player 2 the entire endowment, letting player 2 decide the allocation of the tripled amount, this occurred in only 9 percent of the cases. An additional 15 percent of participants sent 75 percent of the initial endowment.

With respect to players 2, the experimental results also reject the prediction of selfish behavior; the results confirm that reciprocity is the major driver of the behavior of players 2. Only 14 percent of players 2 decided to keep the entire amount in their hands after player 1’s decision; half of these were those who offered a zero return to player 1 when player 1 had also offered them a zero amount. About 11 percent of players 2 did not return any amount to those players 1 who had sent them their entire endowment.

Expectations of the other player’s behavior largely explain the amounts sent by player 1 and the reciprocal responses of player 2. Additionally, individuals’ behavior in other games works as a good predictor of behavior (i.e., more risk-loving individuals both made higher offers as players 1 and returned higher amounts as players 2). When demographic characteristics are analyzed, it is found that females sent slightly smaller amounts and also returned less than males, a result consistent with other findings in the economics literature. Education and socioeconomic status do not seem to explain variations in behavior of players 1, but players 2 of low socioeconomic status tended to return a smaller percentage of the amount received than did other players.

It should further be noted that approximately one out of every four participants in the VCM game opted to contribute to the public good, which rejects the zero-contribution hypothesis in this game as well. Individuals did, in fact, cooperate, a finding consistent with other research. More interesting results emerge when the decision to contribute the token to the group account is explained as a function of game conditions, as well as of individual and group characteristics. First, behavior in this game was found to be consistent with behavior in other games. Decisions to cooperate in the trust game and in the risk-pooling game were significant predictors of the decision to cooperate in the VCM game. Similarly, expectations regarding trust and formation of groups largely explained cooperation in the public goods game. Socioeconomic characteristics do not seem to play important roles in the prediction of cooperation.

Finding 2: Even Though Latin Americans Trust and Cooperate, Social Distances Limit the Extent of Trust and Cooperation

A further important finding on the factors that drive trusting behavior on the part of trusting individuals in the first activity is that wider gaps in education between players 1 and players 2 were linked to reductions in the amounts sent from the former to the latter and in the percentages reciprocally returned. Figure 7.11 clearly illustrates the point, as the bars on both extremes (corresponding to large education gaps between players, either positive or negative) are shorter than those in the middle (corresponding to smaller education gaps). The result persists after individuals’ characteristics (economic and demographic) are controlled for. Figure 7.11 also presents the average foregone social welfare associated with each group of education gaps, measured as the percentage of the total endowments that a pair of players failed to earn as a result of less than complete trust. As the figure suggests, such foregone welfare is greater among pairs of players with larger schooling gaps.

Similar results are obtained in the VCM and risk-pooling games. At the session level, another proxy variable for social heterogeneity, namely, the standard deviation of the years of education, shows negative correlations with the fraction of cooperators in a session, as illustrated in Figures 7.12 and 7.13.

Finding 3: Lack of Trust and Cooperation Has Direct Consequences for the Collective Welfare Generated

Although participants clearly displayed trust and cooperation in the experimental games, they did not do so to the maximum extent possible. Had participants performed at socially optimal levels, they would have increased their gains over actual results by 40 percent in the trust game, 71 percent in the VCM game, and 22 percent (in expected value) in the risk-pooling game. These percentages provide an idea of the magnitude of the social welfare that societies in the region fail to generate as a result of limitations on trust and willingness to cooperate.

These limitations on socially optimal behavior could arise from many sources. Social distance (or specifically, differences in education), as noted above, represents one of these sources. In fact, that variable alone reduces the size of the resulting social welfare “pie” by approximately 9 percent of total wealth in the trust game, 15 percent in the VCM game, and 3 percent (in expected value) in the risk-pooling game.

Finding 4: “Tit-for-Tat” Motivates Latin Americans to Trust and Cooperate

In the trust game, players 1 expecting reciprocation made greater offers to players 2, and players 2 who expected greater offers were also willing to return greater amounts to players 1. In fact, players 2 were willing to allocate a return 2.5 times the size of 100 percent offers of initial endowments on the part of players 1. That return rate decreased with the amount sent by players 1.

Finding 5: Expectations Are Largely Met

Expectations about prosocial or group-oriented behavior have predictive power in explaining results. In both the trust and the VCM games, the participants were asked to predict the behavior of their peers in their sessions, based on demographic information about the groups’ composition. Only 10 percent of participants in the role of player 2 predicted that players 1 were going to send 0 percent of the initial endowment. Slightly more than 38 percent predicted they would be sent 50 percent, and 13 percent of players 2 predicted that player 1 would send them 100 percent of the initial endowment. The forecasts by players 2 proved remarkably accurate. On the other hand, participants in the VCM game predicted, on average, that 44 percent of players would contribute to the group account, and only 7 percent of participants predicted that no one would cooperate in the game. Participants were able to provide a rather accurate prediction of the actual rate of cooperation in the VCM game in their sessions and acted based on a reciprocal strategy. When players expected more people to cooperate in the game, they were more likely themselves to cooperate. Overall, the predicted fraction of cooperators could map the actual fraction of people contributing to the public good, as shown in Figure 7.14.

Finding 6: Socialization, Trust, and Cooperation Are Remarkably Linked for Latin Americans

During the last activity, the risk-pooling game, an average of 48 percent of players decided to join the income-pooling group, with participation rates in individual sessions ranging from 11 percent to 100 percent. More interestingly, the fraction of those willing to join the income-pooling group was highly correlated with the fraction of contributors to the group account in the VCM game, as shown in Figure 7.15. Although these games are designed to measure different dimensions of group-oriented behavior, both might be driven by similar motivations such as in-group or sense of belonging effects. On average, groups who showed greater levels of contribution also showed greater levels of group formation.

An examination of trusting behavior by players 1 in the trust game shows that those who contributed to the group account in the VCM game sent on average 50 percent of their endowment in the trust game, while those who did not contribute sent only 48 percent. Likewise, those who joined the income-pooling group in the risk-pooling game sent offers in the trust game that were about 8 percent higher than the offers by those not joining the group. A similar pattern is confirmed for players 2 in the trust game. Those contributing in the VCM game returned about 8 percent more to their player 1 partners in the trust game, and those who joined the income-pooling group in the risk-pooling game returned almost 6 percent more to their player 1 partners.

The results also suggest interesting differences in regard to the average behavior per city. Participants in Bogotá and Lima showed lower cooperative behavior compared to their counterparts in the other cities of the project. On the other hand, players in Caracas seemed to be, on average, more cooperative than players in the other cities, as shown in Figures 7.12 to 7.15.

CONCLUSIONS

The results from the series of experiments reported here provide new evidence on how group-oriented behavior can emerge among members of groups and what factors may help or constrain choices that benefit individuals and groups. Like those of previous research, the present results show that trust, cooperation, and group formation are highly correlated. Experimental groups with clearly favorable conditions for trusting others were also those whose members were willing to contribute to a public good or to form a group to share income from an uncertain lottery. Finally, expectations regarding the behavior of others were found to be powerful predictors of actual behavior.

The findings on expectations are of crucial importance, for various reasons. First, if people can predict with some accuracy the behavior of those in the same room, based only on a short observation of those in the room and listening to very basic data about the demographics of the group, this means that individuals do pay attention and condition their group-oriented behavior to the immediate context and not only to individual traits of the group members. Secondly, expectations are key informants of economic decisions, and as such they can also misguide people towards behavior that is not group-beneficial, thus leading groups into traps or undesired equilibria.

Education, used here as a proxy measure of socioeconomic status, helps to explain the trusting and cooperative behavior of Latin Americans for several reasons. First, education in itself was found to be associated with higher levels of risk aversion among participants, as well as higher levels of trust and trustworthiness. (It did not, however, increase participants’ likelihood of contributing in the VCM game.) On the other hand, groups with higher degrees of heterogeneity in education showed reduced cooperation and group formation. In the trust game, pairs in which players displayed greater differences in education (education gaps) were characterized by lower amounts sent from player 1 to player 2. Thus, educational attainment on the part of some individuals is not an unmitigated benefit for groups. On the one hand, education can help to develop the cognitive skills required to overcome the limitations of risk aversion and thus can in turn enhance trust. On the other hand, education confers status, and differences in educational level can increase group heterogeneity, social distance, and out-group effects, while diminishing a sense of belonging; these factors can in turn hinder collective action. Consequently, if gaps in education within a group become a source of social distance, they can create barriers to trust and cooperation.

These differences in education, which were found to be linked to the cooperative and trusting behavior of Latin Americans, were in turn linked to possibilities for social welfare generation. Pairs of players with wider education gaps (in the trust game) and groups of players with higher heterogeneity in education (in the VCM and risk-pooling games) generated social outcomes that were smaller than those of their peers with smaller education discrepancies.

From a policy perspective, the findings in this chapter are quite compelling. As extensive research in economics and related disciplines shows (e.g., Putnam, 1994; Fukuyama, 1995), cooperative behavior and group formation leading to social capital buildup and trust are beneficial to societies’ economic growth (Knack and Keefer, 1997). In this context, policymakers should not lose sight of the fact that inclusive policies will yield not only short-run benefits, but also long-term, more durable results. If anything, the key lesson of this chapter for policymakers is that inclusion policies are investment policies.