Robust Dynamic Optimization and Natural Resource Management: Numerical Methods and Fisheries Management

Most resource management problems exist in a setting of not only risk, but of Knightian uncertainty. That is, not only is the problem stochastic, the underlying distribution for this randomness is not known with certainty and is difficult to recover. When faced with such problems of uncertainty or ambiguity, many decision makers do not follow a simple subjective expected utility type behavior (Ellsberg) and under some conditions (Gilboa and Schmeidler, 1989) the rational approach is to make choices as if nature is picking the worst possible probability distribution. This decision framework has been operationalized as robust control, and has received substantial attention by economists in the last decade. For the most part, however, the economists’ analysis has been abstract, yielding qualitative results, but failing to provide specific policy recommendations and bypassing an important modeling choice that must be addressed if robust control is to be used for applied policy analysis. In this paper we present a numerical approach to robust dynamic optimization using an approach developed by Nilim and El Ghaoui (2005) and apply it to a problem of fisheries management. The approach has the advantage of being directly tied to the data, using statistical characteristics of the data to define the robust bounds. Our results