Random Matching and Aggregate Uncertainty

Random matching is often used in economic models as a means of introducing uncertainty in sequential decision problems. We show that random matching schemes that satisfy standard conditions on proportionality are not unique. Two examples show that in a simple growth model, radically di¤erent optimal behavior can result from distinct matching schemes satisfying identical proportionality conditions. That is, non-uniqueness has interesting economic implications since it a¤ects the reward and the transi- tion structures. We propose information entropy as a natural method for selecting unique matching structures for these models. Next, we give conditions on the reward and transition structures of sequential decision models under which the models are not a¤ected by non-uniqueness of the matching scheme.