This paper compares three models of input allocation in multicrop systems. In addition to the variable input and satisficing models analyzed in previous research, an allocatable fixed input model of short-run input use is derived. The empirical application studies irrigation water use in the Central Plains region of the United States. Based on results from model specification tests and prediction accuracy measures, the allocatable fixed input model dominates both other models in explaining multicrop water allocation. In addition, the paper presents an alternative approach to the study of deficient data on multicrop production. By transferring econometric results from analysis of 'non-deficient' crop-level data, input allocations in deficient data sets can be predicted.