Abstract

- preliminary and incomplete-We show how standard learning rules can be interpreted as small departures from rationality in the context of an asset pricing model. We propose a distinction between ‘internal rationality’, as agents that maximize discounted expected utility under uncertainty given consitent beliefs about the future, and ‘external rationality’as agents that know perfectly the true stochastic process for fundamentals (dividends) and market determined variables (asset prices). Naturally, this distinction is irrelevant with complete markets and homegeneous agents. However, we show that the required amount of information and computational ability for achieving external rationality is gigantic, once one allows for weak forms of heterogeneity and market incompletness. We show how simple models of learning that satisfy internal rationality can be interpreted as small deviations from rationality.

Citations