ANZWEE SEMINARS 2020

We conduct a series of laboratory experiments to study how groups aggregate information in settings in which group members repeatedly observe both private information about the state and either others members' actions and/or their actions and signals. Group members share common interests in guessing the state. The treatments vary the information structure group members have access to. We find that observing guesses of others improves information aggregation rates both when guesses of others contain valuable information and when they do not. Our results highlight the crucial difference between the analysis of imitation with short-lived versus long-lived agents. While herding leads to loss of valuable private information in the former case, it improves the group's ability to aggregate information in the latter case, and, thus, leads to higher welfare

Field studies have shown that women speak up less than men in professional environments where women are underrepresented. To shed light on the factors that may contribute to this gap, this paper studies willingness to speak up after someone has raised an opinion. We conduct a laboratory experiment to study why individuals override ideas in teams. Our design identifies the role of beliefs (i.e. disagreement), social image, and gender through two treatments that vary whether participants interact anonymously. Results show that social image concerns matter and cause participants override less when their interactions are not anonymous. Beliefs also play a role. Analysis by gender reveals that the gender of the decision maker does not affect overriding rates. However, the gender of the partner does; when participants disagree with their partner, they are more likely to override a woman than a man. We provide suggestive evidence that preferences in part drive the differential treatment of women. Studying group performance, we find that overriding helps groups on average, while the gender composition of groups does not affect team performance.

Economists model self-control problems through time-inconsistent preferences. Empirical tests of these preferences largely rely on experimental elicitation methods using monetary rewards, with several recent studies failing to find present bias for money. In this paper, we compare estimates of present bias for money with estimates for healthy and unhealthy foods. In a within-subjects longitudinal experiment with 697 low-income Chinese high school students we find strong present bias for both money and food, and that individual measures of present bias are moderately correlated across reward types. Our experimental measures of time preferences over money predict field behaviours better than preferences elicited over foods.

Policy-makers regularly overestimate the impact of development interventions. We explore how exposure to and interpretation of research may contribute to this phenomenon. We present results from three experiments run at or in collaboration with the World Bank and Inter-American Development Bank on how policy-makers, researchers, and development practitioners weigh results, seek information, and update in response to results from academic studies. We find that policy-makers care more about attributes of studies associated with external validity than internal validity, while for researchers the reverse is true. We also find evidence of asymmetric updating on good news relative to one’s prior beliefs and “variance neglect” - an insensitivity to confidence intervals. Finally, we show that information affects a real-life allocation decision.

We study how people perform risky experimentation to acquire information when they can also learn from each other. We develop and experimentally test a two-armed bandit model adapted from that of Keller et al. (2005). Our model predicts that the information generated by a group of players is no more than that generated by a single player. To implement this model in the lab, we design a novel dynamic information structure that can trivialize the posterior calculation for any sequence of signal realizations. We find that 1) when experimenting alone, the median subject generates almost exactly the same amount as the theoretical prediction, that 2) when experimenting with others, the subjects tend to generate more information than they would do alone, which is against the theoretical prediction, and that 3) the subjects only react to the posterior belief and do not condition their actions on what other players' past actions, thus excluding the folk theorem type argument as the explanation.

We report data on a social dilemma laboratory experiment to provide evidence of in-group bias in a leader-follower framework. We ?find that the presence of an in-group leader who advocates co-operation encourages the followers to co-operate in a social dilemma setting. However, the presence of an out-group leader has no such effects and therefore makes in-group leaders more effective than out-group leaders. Additionally, leaders also exhibit in-group bias towards followers. We conclude that group identity is an effective tool in promoting co-operation in groups only if the leader belongs to the same group as the followers.