Dynamic Decision-making in Operations Management

Many Operations Management (OM) models assume that people act as forward-looking optimizers in dynamic environments. We experimentally examine this assumption. To cover a wide range of settings we look at several common classes of dynamic decision problems, and characterize behavior as either optimal, or as consistent with one of several non-optimal policies. At a high level, our results suggest that behaviors are not uniform and depend on the features of the problem. Specifically we find that: (1) Decisions are generally forward-looking (though not always optimal) in Technology Adoption and Capacity Allocation problems, but not in Search/Stopping problems. (2) The optimal policy is a good, but not the best, representation of behavior in both Technology Adoption and in Capacity Allocation; in both tasks simpler forward-looking heuristics achieve a better fit. (3) Performance (payoff) is correlated within-subject for different dynamic tasks but the specific policy usage may vary even within-subject. Together, these results provide new micro-foundations for researchers interested in building more descriptive models of dynamic behavior