Ranking Crop Yield Models Using Out‐Of‐Sample Likelihood Functions

Abstract

There has been considerable debate regarding which probability distribution best represents crop yields. This study ranks six yield densities based on their out-of-sample forecasting performance. The forecasting ability for each density was ranked according to its likelihood function value when observed at out-of-sample observations. Results show that a semiparametric model offered by Goodwin and Ker best forecasts county average yields.