This article presents the first empirical test of Wildavsky's model of self-evaluation by public organizations. We elucidate Wildavsky's arguments and identify six variables that have theoretical effects on self-evaluation. A statistical model that incorporates these variables explains 46 percent of the variation in self-evaluation. The evidence suggests that self-evaluation is positively related to leadership support and employee involvement, and negatively related to the number of organizational sub-units undertaking evaluation at the same time. Refinements to the Wildavsky model are proposed, and conclusions are drawn on the theory and practice of self-evaluation by public organizations.