Experimentally Identifying Constraints to Risk Sharing: Separating Limited Commitment, Moral Hazard and Hidden Income

Project Outcomes Statement

We study how individuals, specifically entrepreneurs, in the developing world cope with risk. Facing poor institutions, individuals are at the mercy of both moral hazard considerations (there is nothing to force their workers into applying much effort) and adverse selection considerations (there is no reason why they should know their workers' abilities). Therefore, they rely on their social networks. By hiring their friends, individuals can both deal with moral hazard, since friends may be unlikely to behave poorly due to reputational or altruistic considerations and with adverse selection, since an individual may be better equipped to know their friend's ability. In our empirical work, we document across markets in Bangalore, Delhi, and Patna that individuals (a) pay all workers, friends or strangers, the same; (b) maintain small businesses, proportional to their network size; (c) tend to hire more out of their friend group when they run businesses where shirking is less observable; among other things. Empirical evidence suggests that individuals may choose to avoid hiring strangers from the general market due to the substantial risk of getting an undesirable type or a situation where effort cannot be controlled, and instead lean on hiring out of their sparse set of friends. We develop a new model to study this phenomenon theoretically. In the model, friends provide two benefits: (a) they are of known ability; (b) they are unable to act in a morally hazardous way. Despite these local benefits, we demonstrate that there can be substantial global costs. Namely, individuals will inefficiently stick to hiring their friends, rather than risk hiring out of a general labor pool. Not only will firms be too friend-biased in composition, but they may remain too small, and worse, there is no market level force that will correct this. That is, at the market level, "social learning" may lead to the entire market never correcting this inefficiency. Our findings suggest new directions to focus in understanding why small to medium enterprises in the developing world may be inefficiently small, too network reliant, and unable to endogenously correct these patterns in the market at large. Multiple research assistants who participated in the work here went on to Master's and PhD programs and a PhD student went on to a top tier faculty position.