Qualitative models can be used for decision making under uncertainty. This provides a useful framework for evaluating the models. If the costs for every action/state of nature combination are known, decisions made using a well-calibrated model would result in actual costs being close to expected costs. In addition, the actual cost can be compared to the cost of perfect foresight actions, giving a bound on the value of a better model. Application of these procedures is made using a logit model developed to predict Missouri country grain elevator bankruptcy.