This paper addresses the key challenge of how to promote intergroup cooperation. Using a lab-in-the-field experiment in Bougainville, Papua New Guinea, we study individuals’ willingness to cooperate with another person in an incentivized social dilemma game. In a between-subjects design, we vary (i) whether individuals interact with a member of their own or another community, and (ii) whether individuals are observed by a key in-group member or not. We find that when individuals are not observed, they are less likely to cooperate with members of other communities. Yet, observation by an in-group member increases the share of people who cooperate with members of other communities from 17 percent to 70 percent, such that there is no in-group bias in cooperation when participants are observed. We relate the results to a shared understanding among the participants that intergroup cooperation is socially desirable. Our findings suggest that communities, organizations, and policymakers may leverage in-group observation to improve intergroup cooperation.

People often care about how they are perceived by others, and this motivates many individuals to act in a way that sends a favourable signal about their character. In this paper, we introduce a decomposition of signalling into its direct and indirect components. An observed behaviour can influence a person’s image directly when the behaviour itself is image-relevant, and it can influence a person’s image indirectly by changing people’s beliefs about an unobserved, image-relevant behaviour. In an experiment on charitable giving, donors decide on (i) what charities to donate to, and (ii) how much to donate. We vary which of the donors’ decisions are observed by a third-party. We find that individuals engage in wasteful, indirect signalling. They engage in a behaviour that in itself is not image-relevant (donating to many charities), but only if doing so can change observers’ beliefs about an unobserved, image-relevant behaviour (donating large amounts). Avoiding wasteful signalling is key to designing effective institutions, and decision-makers therefore need to consider the two components of signalling.

Income and wealth grow over time, and this leads to changes in the level of inequality in a society. Yet, a large literature in cognitive psychology suggests that individuals often struggle to understand the effect of exponential growth. Failing to grasp how inequality develops may lead to biased preferences for policies with long-term effects, from taxation to investments in education. In an incentivised experiment, I examine (i) whether individuals are able to predict how exponential economic growth influences inequality, and (ii) whether informing individuals about the actual development of inequality influences their preferences for redistribution. I find that most people underestimate how much inequality increases in the presence of growth. However, informing individuals about the actual level of inequality does not affect their preferences for redistribution. Rather, what matters is whether people know if redistribution will come at a personal cost to themselves.