پایه های خرد سرمایه اجتماعی

Research on social capital routinely relies on survey measures of trust which can be collected in large and heterogeneous samples at low cost. We validate such survey measures in an incentivized public good experiment and show that they are importantly related to cooperation behavior in a large and heterogeneous sample. We provide evidence on the microfoundation of this relation by use of an experimental design that enables us to disentangle preferences for cooperation from beliefs about others' cooperation. Our analysis suggests that the standard trust question used in the World Values Survey is a proxy for cooperation preferences rather than beliefs about others' cooperation. In contrast, the “fairness question”, a recently proposed alternative to the standard trust question, seems to operate through beliefs rather than preferences.

Trust has been proposed as an important determinant of various economic phenomena, including growth (Knack and Keefer, 1997 and Zak and Knack, 2001), financial development (Guiso et al., 2004), civic participation (La Porta et al., 1997), investment decisions and patterns of international trade (Guiso et al., 2009). Such studies suggest that survey measures of trust like the standard trust question (“Generally speaking, would you say that most people can be trusted or that you can't be too careful in dealing with people?”) are a good proxy for “social capital” and that social capital promotes economic efficiency by facilitating cooperation and the enforcement of incomplete contracts. However, this literature has been challenged on the grounds that it is unclear what survey measures such as the trust question actually measure (see e.g. Sobel, 2002, Durlauf, 2002 and Beugelsdijk, 2006). This paper combines survey measures and experimental data to shed light on the behavioral validity of such measures. By showing that the survey measures importantly relate to cooperation when actual money is at stake we provide evidence for the behavioral validity of standard survey measures, and by uncovering that the standard trust question is related to preferences for cooperation rather than beliefs about others' cooperativeness we provide a more accurate interpretation of what is actually measured by “survey trust”. Thus, our paper adds to the literature on the microfoundations of social capital. We report results from a public good experiment with close to 1500 randomly selected participants from the Danish population. We find that both self-reported trust and observed cooperation levels are high, and regression analysis shows that trust attitudes have significant explanatory power for cooperation behavior. While these results are interesting per se, the main focus of this paper is to study the microfoundation of this relation. We argue that cooperation choices are driven by preferences and beliefs. Some people have no preference for cooperation, and choose to free ride regardless of the contribution level of others (15% in our sample are free riders), but most have a preference for cooperating given that others do (69% are conditional cooperators). Beliefs about other peoples' inclination to cooperate do not matter for free riders but are a main determinant of contribution choices of conditional cooperators. Our main finding is that trust attitudes appear to be a proxy for the strength of people's cooperation preferences but not for beliefs. In particular, we find that responses to the standard trust question (Trust for short) explain how much people contribute given their beliefs about others' contributions, but not how optimistic they are about other peoples' tendency to cooperate. We also find that the fairness question (“Do you think most people would try to take advantage of you if they got a chance, or would they try to be fair?”), an alternative to the trust question that has recently been added to the World Values Survey, is a proxy for beliefs but not for preferences. In particular, we show that responses to the fairness question (Fairness for short) explain how optimistic people are, but do not explain the strength of their preferences for cooperation. We speculate that the two survey measures capture different aspects of social capital because Trust evokes thoughts about what the respondent generally does (“you can't be too careful”) while Fairness evokes thoughts about how other people generally behave (“would they try to be fair?”). Our paper contributes in several ways to a recent stream of research combining survey and experimental measures of social capital. First, we relate Trust to cooperation behavior while the literature has to a large extent focused on behavior in experimental trust games.1 This focus on trust experiments in the literature is surprising given that “social capital” is a multifaceted concept (Dasgupta and Serageldin, 1999), and that most definitions of social capital involve notions of trust and cooperation. In fact, many contributors to the economics literature see trust and cooperation as intimately related concepts (e.g. Knack and Keefer, 1997 and La Porta et al., 1997). In social psychology, the notions of trust and cooperation have long been thought to be closely related. For example, Yamagishi (1986: 111) argues that “mutual trust is the key to actual cooperation”. The public good game used in this study is played in groups and may therefore better reflect important aspects of everyday cooperation problems which are often multilateral rather than bilateral as in the experimental trust game.2 Second, we provide strong evidence that survey measures of social capital are significant predictors of cooperation behavior in the Danish population. Evidence from other samples on the relation between survey trust and cooperation suggests that there is a positive relation between the two measures. For example, an early study by Yamagishi (1986) finds that “high-trustors” contribute more than “low-trustors” in a sample of Japanese subjects, while Ahn et al. (2003), using prisoner's dilemma games, find no relation between a survey measure of trust and cooperation behavior in a sample of US students. Anderson et al. (2004) find in a sample of US students that both Fairness and Trust are positively related to contributions in a public good experiment. In a comparable sample, Capra et al. (2008) find that the two survey measures are jointly (but not separately) related to contributions. Similarly, Gächter et al. (2004) find that Fairness is related to cooperation behavior in a sample from Russia and Belarus. Herrmann et al. (2008) report results from cooperation games with punishment and find that survey measured civic norms strongly relate to cooperation across a sample of 16 different participant pools. Using a sample of Peruvian microcredit borrowers, Karlan (2005) finds that a summary variable of several survey measures does not relate to contributions in a public good experiment when social capital is measured at the individual level. However, when aggregating the answers of the survey questions for each group of the public good game, the resulting social capital measure is positively related to the share of subjects in the group that contributes.3 Third, our findings suggest that different survey measures of social capital capture different determinants of cooperation and, thus, of social capital. More specifically, we find that Trust is related to cooperation preferences but not to beliefs about cooperation, while it is the other way around for Fairness. We are able to disentangle these two channels because we measure individual choices, beliefs and preferences using two versions of the public good game.4 The first is a standard one-shot cooperation game which we refer to as the Standard game. In this game, participants are endowed with money, approximately $10 each. Participants are anonymously matched into groups of four and simultaneously decide how much to contribute to a common project. All contributions are doubled and equally shared among the four participants. Not to contribute is therefore the individually money-maximizing choice, while contributing the total amount is the efficient choice since only money contributed will be doubled by the experimenter. Participants also indicate their expectation about the average contribution of others. In the second game, referred to as the Strategy game below (developed by Fischbacher et al., 2001), participants provide a complete contribution schedule conditional on the contribution choices of others. That is, they decide to contribute a, b, c given that others on average contribute x, y, z. Thus, beliefs about the average contributions of others do not matter for contributions in the Strategy game by design. Other large-scale studies have not been able to distinguish between the preference and belief channels of cooperation. The closest match to our study in this respect is Sapienza et al. (2007). In contrast to our results, these authors find that Trust captures the belief-based component but not the preference-based component of behavior in trust games.5 However, our finding is broadly in line with Gächter et al. (2004) for cooperation behavior, and with Fehr et al. (2002) for choices in a trust game. These authors find that the trust question remains significant for explaining trust behavior even when controlling for beliefs about the money sent back by second movers (i.e. beliefs about others' trustworthiness). On a methodological level, our findings validate widely used survey measures of social capital with measures of cooperative behavior in a controlled environment with monetary incentives.6 In a broader perspective, our paper therefore contributes to an ongoing debate on the relative merits of hypothetical vs. experimental (incentivized) measures (e.g. Camerer and Hogarth, 1999). Survey measures are inexpensive to collect in large and heterogeneous samples, but have been criticized by economists as being unreliable because answers are not incentive-compatible and respondents may therefore be inattentive or insincere. Controlled experiments come at a considerable administrative and monetary cost and may not be able to capture the heterogeneity present in the general population when implemented with relatively homogeneous convenience samples of students. We are able to provide a valuable validation because we implement a controlled and incentivized experiment with a sample that closely parallels the samples used in survey studies. We are able to do so because we use a “virtual lab” approach to recruit a large heterogeneous sample and to implement an experiment with a level of control comparable to the conventional laboratory. On a practical level, our finding that alternative survey measures capture different aspects of social capital has potentially important implications for the choice of policies aimed at clamping down on uncooperative behavior like tax evasion, petty crime or traffic rule violations. Recent evidence suggests that cooperation preferences are likely to be more stable and more difficult to influence than beliefs about cooperativeness in society.7 If so, policies targeted at correcting pessimistic beliefs about other peoples' compliance with cooperation norms may be a cost-effective alternative in improving efficiency to more conventional measures based on sanctions to deter non-compliance. For example, the OECD guidelines on improving tax compliance (OECD, 2004: 38) explain that “If a person believes that non-compliance is widespread they are much more likely not to comply themselves. Studies indicate that it is effective in reducing non-compliant behavior to ensure that taxpayers have an accurate understanding of the compliance behavior of others.” In line with this view, a field experiment by Fellner et al. (forthcoming) demonstrates that providing social information on compliance rates (paying fees for public TV) increased compliance in areas with low compliance rates. Asking the appropriate survey question can thus serve as an inexpensive diagnostic tool in evaluating policy choices. Our results suggest that policies aimed at correcting overly pessimistic beliefs might be effective if a (sub-)population scores low on Fairness. When a (sub-)population is found to score low on Trust, other policy measures – perhaps using the threat of punishment to deter uncooperative behavior – may be more effective.

We use data from an experiment with close to 1500 participants from all walks of life in Denmark to show that the most prominent survey measure of trust (Trust) is an important predictor of social capital in the guise of voluntary contributions to public goods. Our analysis behaviorally validates survey measures and contributes to the microfoundations of social capital by suggesting that Trust proxies the preference-driven component of cooperation. Trust is a stronger predictor of cooperation preferences than gender, age, education or salary. In contrast, our analysis indicates that an alternative survey measure of social capital that has recently been introduced to the World Values Survey, the Fairness question, primarily explains optimistic beliefs about cooperation in others. These optimistic beliefs map into increased cooperation because most participants have preferences to cooperate given that others do, i.e. because they are conditional cooperators. Apart from this belief-mediated effect, we find no direct influence of Fairness on cooperation. The effects of Trust and Fairness on cooperation are statistically significant in regressions which control for socio-economic variables like age, gender, salary and education. The effects are not only statistically significant, they are also sizeable. For example, those who trust contribute about 10% more than those who do not, and those who indicate full confidence in other people's fairness contribute about 20% more than those who express minimal confidence in other's fairness. These effects are remarkably strong, both compared to findings in the literature and compared to alternative explanations. First, research in social psychology suggests that the relation between attitudes and behavior is often rather weak (e.g. Eagly and Chaiken, 1993), and previous studies on the relation between survey and experimental measures of social capital tend to find mixed effects (see introduction for references). Second, in line with much of the literature, we find that cooperation behavior is mainly driven by beliefs. However, if beliefs are not accounted for in regressions, Trust and Fairness variables account for more variation than our socio-economic variables. We show that Trust and Fairness are systematically related to cooperation in a one-shot interaction with fully anonymous counterparts, i.e. in a situation without prior information about or experience with their counterparts. Thus, Trust and Fairness capture important aspects of “thin” trust towards a “generalized other” which has been found to be a relevant determinant of economic prosperity. Beliefs are likely to be particularly relevant in such anonymous one-shot interactions. After all, optimism and pessimism about others' inclination to cooperate matter most when little is known about actual cooperation. However, in everyday life, cooperation problems often loom large in groups whose members repeatedly interact, as in the workplace or repeat customer relations. Based on our results, we speculate that “thick” trust which is required in this type of repeated interaction is better predicted by Trust than by Fairness. The reason is that beliefs about cooperation are adjusted to observed contributions over time and, therefore, are likely to eventually become largely irrelevant as an independent determinant of behavior. A long-standing debate in the social sciences concerns the appropriate choice of policies to improve cooperation. A recent focus of research in (experimental) economics has been on the relative merits of formal vs. informal sanctions to deter uncooperative behavior (see e.g. Markussen et al., 2011 for a discussion). While often effective, both types of policies can be rather costly because meting out sanctions is wasteful. In contrast, policies that rely on “belief-management” as e.g. proposed by the “broken windows” theory, do not rely on wasteful sanctions and may be effective in some circumstances (see Keizer et al., 2008 or Fellner et al., forthcoming for field experiments). Such policies aim at improving cooperation by correcting exceedingly pessimistic beliefs about other peoples' inclination to cooperate. According to our results, the survey measure of Fairness provides an inexpensive diagnostic tool indicating when resorting to policies involving belief-management may be worthwhile.