Dynamic Selection and the Puzzle of Slow Climate Change Adaptation in U.S. Agriculture

Long-run studies of agricultural productivity use balanced county-level panel data to document slow adaptation to climate change in U.S. agriculture. We revisit this treatment-effect puzzle using a selection model that incorporates the optimizing behavior of heterogeneous farmers. Using data from 1980 to 2020, we document three dynamic selection effects. First, counties that completely exit farming alter the composition of the balanced panel sample. Second, higher-skilled workers are more likely to leave rural areas as weather conditions worsen. Third, smaller farms consolidate into larger farms in counties affected by climate change. We document that the yield damage function’s slope depends on the county’s educational level and farm size. Causal estimates of weather impacts must account for selection effects along several adaptation margins of adjustment.