ریسک پذیری در تنظیمات اجتماعی:اثرات گروه و همسالان

We investigate experimentally the effect of consultation (unincentivized advice) on choices under risk in an incentivized investment task. We compare consultation to two benchmark treatments: one with isolated individual choices, and a second with group choice after communication. Our benchmark treatments replicate findings that groups take more risk than individuals in the investment task; content analysis of group discussions reveals that higher risk-taking in groups is positively correlated with mentions of expected value. In our consultation treatments, we find evidence of peer effects: decisions within the peer group are significantly correlated. However, average risk-taking after consultation is not significantly different from isolated individual choices. We also find that risk-taking after consultation is not affected by adding a feedback stage in which subjects see the choices of their consultation peers.

The standard economic approach to the analysis of choice under risk emphasizes the role of individual risk preferences. In deciding how much to invest in a risky asset, individuals weigh up the costs and benefits referring to these preferences. By contrast, in many important real-world settings individuals do not take choices in isolation, and the social settings within which choices are made may influence behavior. For example, individual choices may be swayed by the opinions and decisions of others. In this paper we investigate how a common social setting, consultation with a group of peers, affects choices under risk in the laboratory. There is abundant field evidence that people's choices are often influenced by their peers.1 While field studies can provide compelling evidence of correlated behavior within peer groups, identifying these as peer effects is complicated by confounding factors (Manski, 1993). Moreover, it is difficult to assess the influence of peer effects from field data, as naturally occurring control treatments where peer effects are absent but other variables are held constant are typically not available. For these reasons we use a controlled laboratory experiment, described in Section 3, to investigate the effect of social settings on investment decisions over multiple periods. Our experiment has two benchmark treatments that replicate Sutter's (2009) experiment on decision-making in groups. In one treatment, decisions are made by isolated individuals without any communication with peers. In the other treatment, decisions are made by groups whose members can communicate and have to agree on a single group decision via electronic chat. In our two consultation treatments, subjects are also allowed to freely communicate with their peer group, as in the benchmark treatment with groups, before making a decision. However, each subject's earnings depend only on his or her own choices and not on the choices of others. We use this framework because direct communication between peers is an important feature of many settings where peers may influence one another. Our focus on consultation contrasts with related laboratory studies of peer effects, discussed in Section 2, in which subjects are informed of each other's choices and may be influenced by these, but there is no direct communication between subjects (for example, Yechiam et al., 2008, Cooper and Rege, 2011 and Lahno and Serra-Garcia, 2012). We do, however, control for the additional influence of seeing others’ choices by varying the degree of feedback we offer to subjects across the two consultation treatments. In one treatment we ensure that subjects are fully informed of the choices of others in their group while in the other treatment they do not receive such feedback. Our experiment is also related to experiments where subjects give and take advice (also discussed in the next section). However, our framework departs from these studies in that we do not incentivize giving advice. Instead, the only motivations for our subjects to give or take advice are intrinsic motivations independent of financial consequences (as in many examples of peer advice in everyday life). Also, our subjects face the same task at the same time as their peers, whereas in other experiments on advice the experimental design induces differences between the experience and/or expertise of advice givers and takers. In Section 4 we report our results. The benchmark treatments replicate previous findings of higher risk-taking by groups relative to isolated individuals (Sutter, 2007 and Sutter, 2009). Content analysis of the messages sent by group members shows that higher levels of risk taking are associated with messages referring to expected value maximization. We also find some evidence that risk-taking is higher in consultation groups where expected values are mentioned, although this effect is only marginally significant. Furthermore, we find that consultation does not increase average risk-taking beyond that observed among isolated individuals. Thus, simply providing direct communication between peers does not result in the higher risk-taking observed when decisions are made by groups. We do, however, find evidence of peer effects in our consultation treatments. Within consultation groups, variability in choices is significantly lower than the variability in choices between individuals from different groups; this result holds whether we explicitly inform subjects of their peers’ previous round choices or not. More generally, we do not find any evidence that informing subjects of the previous round choices and earnings in their peer group influences risk-taking when subjects already have the ability to consult with these peers through electronic chat.

Using a simple investment task we compare choices under risk by three types of decision-maker: isolated individuals, groups, and individuals who can consult each other. In line with previous research using the same investment task (Sutter, 2007 and Sutter, 2009), we find that groups take more risk than individuals. When individuals can consult one another we find that communication among peers leads to significant correlation of decisions within the consultation group. However, consultation has a weak effect on the level of risk-taking: average risk-taking after consultation is not significantly different from the average risk-taking of isolated individuals. We also find that, if subjects can already consult with others in their peer group, explicitly showing them the choices and earnings of peers does not change their behavior. Although consulting individuals can discuss the task in the same way as group decision-makers, content analysis reveals some important differences between treatments. Perhaps most importantly, subjects in the consultation treatment exchange fewer messages than in the group treatment, including messages discussing expected values. This may explain why consultation fails to increase average investment, since mentions of expected value have a strong effect on average investment in the group treatment. These results suggest that having to make a group choice under risk is quite different from giving people the opportunity to communicate with peers. Our consultation treatments were designed to isolate the effect of unincentivized communication between peers. If subjects had been financially motivated to provide others with investment advice – for example, if they had been paid a percentage of others’ earnings – it is possible that consultation would have had a significant effect on the level of investment. Similarly, we chose not to direct subjects to use communication in any particular way. If we had made it mandatory for subjects to justify their choice to their peers, this might have induced them to think differently about the task (and perhaps about the expected value of their choices), and may have resulted in a higher level of investment. Thus, our finding that consultation does not translate into higher levels of investment than are made by isolated individuals may reflect particular features of our design. Nevertheless, it is notable that even in our relatively simple consultation setting subjects’ decisions are influenced by their peers, as evidenced by the similarity of investment decisions within consultation groups. Further investigation of how features of the social setting influence risk-taking among peers seems a promising direction for future research.