A statistical learning approach to pasture, rangeland, forage (PRF) insurance coverage selection

Abstract

This study uses statistical learning methods to identify robust coverage alternatives for the Pasture, Rangeland, Forage (PRF) insurance program. Shrinkage and ensemble learning techniques are adapted to the context of the PRF coverage selection process. The out-of-sample performance of the proposed methods is evaluated on 116 representative grids throughout Texas during 2018–2022. Ensemble learning methods generated more stable coverage choices compared with the other selection strategies considered. Depending on the target return, a reduction in the prediction error between 5% and 14% was observed. Furthermore, the proposed coverages can provide a broader protection than current coverage choices made by farmers.