Endogenous social interactions
This thesis comprises three chapters. All chapters have a common theme which at it's base deals with how interactions in social non-market contexts may shape individual and aggregate economic outcomes. Chapter 1 entitled "Whom Should I Observe", studies a model of observational learning in the context of a simple learning problem. Players have heterogeneous prefer ences over outcomes. A player can learn about the underlying outcome distribution by observing one other player. We characterise optimal link formation. The main result is that some players prefer to form a link with a player who experiments with actions that they axe not willing to experiment with themselves. This is interpreted as an informa tional micro-foundation for a preference for diversity. Applications are discussed. Chapter 2 is entitled "Are Gifts-in-Kind Inefficient". It is often argued that gifts- in-kind are inefficient transfers (Waldfogel 1993). We study a model of partnership formation, where players have incomplete information about the desirability of the part nership. Prior to the players simultaneously deciding whether to form the partnership, one player gets a signal of the partner's type and can send gifts, which may be either a gift-in-kind or cash. The model has multiple equilibria. Under certain conditions the efficient equilibrium payoffs involves the transfer of gifts-in-kind. The reason is that gifts-in-kind reveals more to the receiver about the giver's beliefs about the receivers type than do other transfers. An evolutionary argument, in the spirit of Kim and So- bel (1995), is given. In the long run the efficient equilibrium is played with positive probability. Chapter 3 entitled "Revisiting Schelling's Spatial Proximity Model" formalises the model of Schelling (1969, 1971) of interaction in one-dimensional neighbourhoods. We show, via numerical simulations, that the rest points of the adaptive process tends to select neighbourhood configurations which are relatively segregated in the aggregate. We test the robustness of rest points to the introduction of noise in the adaptive process. The long run prediction is that complete segregation occurs. The model is simulated and results show that the wait until the stochastic process reaches the set of segregated states increases rapidly in the size of the population. Variations, with better long run properties, are suggested and analysed. We also analyse a model where residents have a strict preference for integration. Nevertheless the only stochastically stable states are segregated. We test the robustness of this prediction by allowing for heterogeneity in preferences. Interestingly this turns the prediction on it's head: only integrated states are stable. Schelling's original model is robust to this pertubation.