Misclassification Error in Remote Sensing Matters: The Effect of Ethanol Plants on Local Cropland Transitions

The quality of land cover data substantially affects estimates of treatment effects in land use change studies. We estimate the local impact of US ethanol plant expansions on cropland retention and conversion. Due to misclassification error, we show that using raw data from the Cropland Data Layer gives mixed results or results counter to economic expectations, while a cleaned version gives consistent and intuitive results. Our findings also highlight the importance of other methodological choices. Our preferred specification shows that plant expansions increase the probability of cropland conversion by 1.5 percentage points and cropland retention by 0.1 percentage points in plant neighborhoods.