THE NEED FOR SENSITIVITY ANALYSES OF DYNAMIC OPTIMIZATION MODELS

The properties of dynamic optimization models are not obvious, since they lack explicit functional form and must be solved by computer simulation. Models should be accompanied by sensitivity analyses demonstrating (1) that optimal behavior offers a significant advantage over alternatives, and (2) that variation in parameters does not significantly alter the outcome. Re-examination of Clark's (1987) model on lion foraging group size demonstrates that calculating numerical optima at a few points does not adequately describe the "landscape" of a model's predictions. A brief survey reveals limited or no sensitivity analyses in other published models.