Very little attention has been given to the modeling of yield distribution for crops and regions in which yields exhibit irregular behavior. We undertake a statistical case study of Texas upland cotton and propose an alternative mixture distribution based on regime-switching model in which the conditional distribution of yield depends upon an observable drought index. The results show that the mixture distribution model provides a better fit to the data than conventional parametric distributions and produces higher implied premium rates than the current published Group Risk Plan insurance rates in more than two-thirds of Texas counties examined.