Non-rationalizable Individuals, Stochastic Rationalizability, and Sampling

TL;DR

This paper explores the relationship between deterministic and stochastic rationalizability in individual choices, showing that populations can be stochastically rationalized even with many non-rationalizable individuals, and that detection of such individuals becomes harder with more data.

Contribution

It establishes that stochastic rationalizability can hold even when many individuals are not deterministically rationalizable, and analyzes the limitations of identifying non-rationalizable individuals from population data.

Findings

Populations can be stochastically rationalized despite many individuals lacking deterministic rationalizability.

Detection of non-rationalizable individuals from data decreases as the number of observations increases.

Stochastic rationalizability does not imply individual deterministic rationalizability.

Abstract

Experimental work regularly finds that individual choices are not deterministically rationalized by well-defined preferences. Nonetheless, recent work shows that data collected from many individuals can be stochastically rationalized by a distribution of well-defined preferences. We study the relationship between deterministic and stochastic rationalizability. We show that a population can be stochastically rationalized even when half of the individuals in the population cannot be deterministically rationalized. We also find the ability to detect individuals who are not deterministically rationalized from population level data can decrease as the number of observations increases.