Bounded rationality and thick frontiers in stochastic frontier analysis

Abstract

Recent research has proposed a statistical test based on the notion that agents have bounded rationality, if and only if more attractive states are chosen with larger probability. We propose and implement a statistical test for bounded rationality in the context of stochastic cost frontiers. Bounded rationality is related to probabilistically cost-efficient distributions. The test is based on comparing a discrete set of probabilities with the theoretical distribution under bounded rationality. Implementation is shown to be quite easy in a Bayesian framework using the Bayes factor for model comparison between estimated and theoretical probabilities. The bounded-rationality model introduces only an extra parameter in frontier models and, therefore, it is quite practical to use in applications as a general semi-parametric model for inefficiency.