Estimates of marginal products and rates of return to cash grains, dairy, poultry, and other livestock research in the United States were made by Bredahl and Peterson using 1969 Census of Agriculture data. Their results showed national returns to crop and livestock research to be in the 36 to 46% range. These estimates of returns, several times higher than market rates, have proven useful to agricultural researchers and administrators in supporting budget requests. Bredahl and Peterson provided marginal products by commodity groups by states which have been used by economists in particular states to calculate rates of return to research on commodity groups in those states (Mitchell, Coffey, Babb, and Pratt). More recently, Davis has provided evidence that the production coefficient on the research variable in aggregate agricultural production functions has declined since the 1950s but remained stable for the past 10-15 years. Stability in the aggregate, however, does not necessarily imply stability over time across commodity groups or states. Stability of the research coefficient is an important issue since estimates from studies such as Bredahl and Peterson's are used in making projections of returns to future research spending. Instability over time would indicate that one should not make projections which make use of research coefficients from only one cross-section. The main focus of this paper, therefore, is to provide additional evidence on the efficiency of allocation of research resources among commodity groups and regions within the United States. Data from the 1969 and 1974 Censuses of Agriculture are employed in aggregate agricultural commodity group production functions to test if the research coefficient for any or all of these groups has remained stable from 1969 to 1974. A second purpose of the study is to examine the effects on the research coefficients of certain variables not tested in the Bredahl and Peterson study. Variables are included to account for research spillover, weather differences, and land quality differences across states. Alternative research lags are tested and the importance of the assumed research lag on the rates of return is also illustrated. The question of research spillover is an important one and has recently received increased attention in the literature (Evenson, White and Havlicek, Davis, Garren and White). It is really the lag in spillover or the incomplete spillover of research results from one state to another that allows one to pick up any variance with a state level research variable in a cross-sectional production function. The spillover that occurs, if unaccounted for, will likely bias the state marginal products derived from commodity group production functions.