Essays on Rationality: Modeling and Measurement

Abstract

The assumption of rationality is one of the cornerstones of economic analysis. Yet, substantial evidence has documented violations of rationality. This dissertation studies modeling and measurement of rationality.

Chapter 1 studies extending rationality to state dependent utilities in choice under uncertainty where the outcomes depend on the realized state of the world. The typical assumption is that utilities of outcomes do not depend on the realized state and are state independent. Without this simplifying assumption, it is difficult to separately identify utilities and beliefs. We provide novel general foundations for models with state dependent utilities: once we depart from expected utility, it is often possible to uniquely identify utilities and beliefs. Specifically, we show that with general models of non-expected utility under ambiguity we have complete identification of utilities and probabilities under full-dimensional uncertainty.

In Chapter 2, we study how to model cost of rationality or complexity in choice under risk. We consider a decision maker who cancels out common consequences with common probabilities when comparing lotteries. Under continuity assumptions, we show that there are two possibilities. 1) The decision maker is an expected utility maximizer with Shannon's entropy as an additive cost of complexity. 2) The decision maker is a power weighted expected utility maximizer where the probabilities are weighted by a power function. Additionally, we provide novel characterizations for expected utility. We consider a decision maker who evaluates each possible prize independently of the other possible prizes. If the preferences satisfy first order stochastic dominance or joint continuity in prizes and probabilities, then the decision maker is an expected utility maximizer.

Chapter 3 studies measurement of rationality violations. Over the past decades, several measures have been proposed to quantify rationality violations. However, little is known about their foundations and how they compare with each other. We provide novel axiomatizations and simulations for the most used measures of rationality in the standard budget-choice setting. This allows both theoretical and practical comparisons of the rationality measures. Additionally, the simulations suggest improving the current measures by overweighting small violations of rationality.