Estimating the Efficacy of Fighting Fire: Propensity Score and Instrumental Variable Methods

In this paper we estimate the treatment effects wildfire suppression (fire crew response time) and fuel management (prescribed fire) have on wildfire size and intensity. Since wildfire management (suppression and fuel management) is likely endogenous to wildfire behavior, we develop instrumental variables (IV) and propensity score matching (PSM) methods to provide consistent estimations of the returns to management. The suitability of IV versus PSM methods depends largely on the assumptions made and the data available, and in the case of wildfire modeling, it seems PSM may be more appropriate. While the vast majority of PSM literature focuses on binary treatments, we explore the ability of recently developed continuous treatment propensity score models. We find that in general wildfire management does limit wildfire size and intensity. However, our results differ depending on the estimation strategy (OLS, IV, PSM) employed. † Economist, USDA Forest Service, RTP, NC and Ph.D. Candidate (Economics) at North Carolina State University, Raleigh, NC.