The decision-making process by which academic departments within an experiment station allocate funds among commodities is examined. The decision to conduct research on some commodities and not on others introduces a problem of censored dependent variables. In order to overcome this problem, a simultaneous equations model with selectivity was used; it was applied to data from the Idaho Experiment Station. The results indicated a simultaneous relationship between research funding levels and expected benefits. Marginal products of one dollar in research investment were $53.80 for applied research, and $8.49 for maintenance research.