A Heuristic Approach to Explore: The Value of Perfect Information

Abstract

How do people choose in a dynamic stochastic environment when they face uncertainty about the return of their choices? There is a growing literature that investigates the validity of boundedly rational models in this type of environment. In this research, we contribute to this literature by proposing a new heuristic decision process called Myopic-VPI, which extends the Value of Perfect Information (VPI) idea first proposed by Howard (1966) and Dearden et al. (1998, 1999) in engineering and computer science. This approach provides an intuitive and computationally tractable way to capture the value of exploring uncertain alternatives. In our approach, a decision-maker investigates the benefits of a subset of information, which can improve her myopic decision outcome. More specifically, the Myopic-VPI approach only involves ranking the alternatives and computing a one-dimensional integration to obtain the expected future value of exploration. In terms of computational costs, we show that Myopic-VPI is very attractive compared with dynamic programming, Index Strategy, and other heuristics, although its performance in accumulated rewards is not the strongest. Using individual-level scanner data, we find evidence that Myopic-VPI is able to capture actual consumers’ choices very well compared with other models under consideration (it provides the best in-sample fit, and very competitive out-of-sample fit). Our simulation and estimation results suggest that although consumers sacrifice some accumulated rewards by adopting Myopic-VPI, it allows them to save in cognitive costs. We argue that practitioners should consider Myopic-VPI as a serious alternative “as if” consumer model because of its relatively superior empirical performance in capturing actual consumer choice, and low implementation cost.