A reconsideration of the Rothschild-Stiglitz insurance market model by information theory

This paper reconsiders the Rothschild-Stiglitz insurance market model by information theory. The seminal work by Rothschild and Stiglitz (1978) has become one of the standard models of insurance markets. However, there are a couple of issues that economists find it difficult to understand with this model. First, this model shows that all the equilibria are separating equilibria, and such separating equilibria may not exist under certain conditions. Second, this model is based on the indifference curves of the insurance consumers' utility functions which could be difficult to obtain empirically. This paper attempts to overcome these difficulties by using what Spence called "indices" of observable and unalterable attributes of insurance consumers. This paper approaches the problem as a sequential Bayesian game (a dynamic game with incomplete information), and uses information theory to estimate the costs and benefits of the information contained in indices. Consequently, this paper shows that if the costs and benefits of information fall within certain ranges, using indices is sufficient to avoid the non-existence of equilibria. Furthermore, obtaining indices is easier than specifying the utility function for each insurance consumer.