HETEROSKEDASTICITY IN CROP YIELD MODELS

This study examines three alternative models of correcting for heteroskedasticity in wheat yield: the time trend variance, the GARCH, and an econometric model that includes the potential sources of heteroskedasticity. Nonnested test results suggest that modeling the sources of heteroskedasticity is the preferred procedure. Including potential sources of heteroskedasticity as explanatory variables removed the heteroskedasticity in the sample wheat yields. The results also suggest that the GARCH specification is a promising model of correcting for heteroskedasticity when the sources cannot be identified. The time trend variance model alone may misspecify the true variance structure.