To what extent can long-differencing capture adaptation?

Understanding the degree to which we are able to adapt to climate change is central to economic assessments of future climate damages. Social scientists increasingly use comparisons between long-difference and panel fixed effects estimators to measure climate adaptation. Despite their empirical relevance, there is no formal framework to assess the extent to which these comparisons are valid. We demonstrate two limitations of this empirical strategy. First, standard implementations of these estimators yield biased estimates of both long-run and short-run population parameters. Second, the direction of this bias may substantially underestimate adaptation. We illustrate the empirical relevance of our theoretical results in an empirically-calibrated simulation design and an empirical application.