Rationality and observed behavior

In this Department of Economics seminar, Pedro Rey Biel from Universitat Autònoma de Barcelona discusses his paper on Rationality and observed behavior.

Abstract

Rationality and higher-order rationality are both central assumptions implicit in most of economic theory. Yet, growing evidence suggests that, while a large majority of experimental subjects play rational actions in simple laboratory games, only a considerably smaller fraction satisfies higher-order rationality. An important step towards identifying higher-order rationality from experimental data was made recently by Kneeland (2015).

We argue that her class of games, while substantially improving on previous methods used in the literature, suffers from problems of both framing individuals to choose actions that are consistent with higher-order rationality, as well as of allowing for an easy inductive step to identify such actions.

To overcome these two shortcomings, we propose a novel class of games with incomplete information, the e-ring games, that builds on the ring games used by Kneeland (2015), while using an information and communication structure related to the email game of Rubinstein (1989). We use our new class of games to identify higher-order rationality, and, for the first time, provide a comparison of different methods of measuring higher order rationality.

We find that, for the same subject, higher-order rationality is generally very variable across games and that the ring games used by Kneeland (2015) may indeed be subject to both framing and easy inductive reasoning.