On the Timing of Relevant Weather Conditions in Agriculture
Abstract
A growing literature is analyzing the effects of weather fluctuations on agricultural outcomes.When estimating these weather-agriculture linkages, researchers often make ex ante arbitrary
choices about the “season”, the time period over which weather conditions are believed to be
relevant to the agricultural outcome. In this paper, we propose a tractable approach to identify the timing of relevant weather conditions in such econometric models. The approach consists in a grid search of models based on all possible calendar periods within the year. In simulations, we find our approach is effective at recovering the “true season”, sometimes even when the selected weather variables do not match the data generating process. We also find that imposing an incorrect season introduces non-classical measurement error which biases estimates of weather impacts in either direction. We apply our approach to a panel of state-level agricultural TFP growth rates. We find considerable heterogeneity in seasonality across regions. Importantly, we find that imposing arbitrary seasons can bias weather effects in either direction, in line with our simulations.