To what extent can long-differencing capture climate adaptation?

Understanding the degree to which we are able to adapt to climate change is central to economic assessments of future climate damages. Economists increasingly use comparisons between long differences and fixed effects estimators to measure climate adaptation. We show that such comparisons can be misleading. Neither estimator is consistent for its intended parameter, as both the long-difference (LD) and fixed effects (FE) estimands are weighted averages of the long- and short-run responses to climate and weather. As a result, the difference between the two understates the true extent of adaptation, and the standard test based on this difference --while controlling size -- tends to be substantially underpowered in the settings researchers typically encounter. An empirically-calibrated simulation shows this difference understates adaptation by about 30--80%, depending on the averaging window.