Choosing a Good Toolkit:
An Essay in Behavioral Economics
Alejandro Francetich. University of Washington Bothell School of Business, and
David M. Kreps, Graduate School of Business, Stanford University
September 2015
Abstract: The problem of choosing an optimal toolkit day after day, when the distribution
of values of different tools is uncertain and can only be learned by carrying the tools, is
a multi-armed bandit problem with non-independent arms. Accordingly, except for very
simple specifications, this problem cannot (practically) be solved. Decision makers facing
this problem presumably resort to “sensible decision heuristics, employing past experience
and, perhaps, what they know about the problem. We examine and compare the performance
of several heuristics, from very simple and naïve to sophisticated. Asymptotic results
are obtained, concerning the long-run performance of the heuristics, that indicate how our
heuristics perform for discount factors close to one. But our focus is on the relative performance
of these heuristics for discount factors bounded away from one, which we study
through simulation of the heuristics on test problems. (139 words)
Keywords: Heuristics, multi-armed bandits, behavioral decision making
