Political Learning from Rare Events: Poisson Inference, Fiscal Constraints, and the Lifetime of Bureaus

David E. Lewis Woodrow Wilson School and Department of Politics, Princeton University, Princeton, NJ 08544 e-mail: delewis{at}princeton.edu How do political actors learn about their environment when the"data" provided by political processes are characterized byrare events and highly discontinuous variation? In such learningenvironments, what can theory predict about how learning actorswill take costly actions that are difficult to reverse (e.g.,eliminating programs, approving a risky new product, revisinga security policy, firing or recalling an appointed or electedofficial)? We develop a formal model for this problem and applyit to the termination of bureaucratic agencies. The conventionalwisdom that "the older a bureau is, the less likely it is todie" (Downs 1967, Inside Bureaucracy) persists but has neverbeen properly tested. This paper offers a learning-based stochasticoptimization model of agency termination that offers two counterintuitivepredictions. First, politicians terminate agencies only afterlearning about them, so the hazard of agencies should be nonmonotonic,contradicting Downs's prediction. Second, if terminating agenciesis costly, agencies are least likely to be terminated when politiciansare fiscally constrained or when the deficit is high. We assessthe model by developing a battery of tests for the shape ofthe hazard function and estimate these and other duration modelsusing data on U.S. federal government agencies created between1946 and 1997. Results show that the hazard rate of agency terminationis strongly nonmonotonic and that agencies are less likely tobe terminated under high deficits and divided government. Forthe first 50 years of the agency duration distribution, themodal termination hazard occurs at five years after agenciesare enabled. Methodologically, our approach ties the functionalform of a hazard model tightly to theory and presents an applied"agenda" for testing the shape of an empirical hazard function.With extensions, our model and empirical framework are applicableto a range of political phenomena.