Beyond marginal: Estimating the demand for water quality

Using micro-level data across Wisconsin covering over 100 inland lakes, we recover first- and second-stage hedonic welfare estimates for non-marginal changes in water quality. We overcome longstanding endogeneity concerns with Rosen (1974)’s second stage hedonic framework and recover slope estimates for water quality demand using instruments based on sorting behavior. For near lake Wisconsin households, we find the slope of their water quality demand function is bounded by − 2087 when imperfect instruments are employed, which is significantly more price inelastic than the naïve OLS estimate of − 895. Applying these estimates to a hypothetical policy scenario where water quality reduces by 24.2% due to a 30-year continuation of current trends, we find welfare losses of at least $7554 per household. These losses are 22% ($1658) more than what is predicted from marginal willingness to pay estimates recovered from the first-stage hedonic. For policymakers, our results highlight the importance of recovering underlying demand functions when evaluating non-marginal water quality improvements.