Previous tests of the marginal productivity theory have been criticized on several grounds reviewed by the authors. One important deficiency has been the small number of factor inputs entered in the production functions. In 1978 Gottschalk suggested a method to estimate production functions with many inputs by assuming that the production process can be split into subprocesses. This reduces the probability of multicollinearity. The authors show that the method depends on an additional assumption. Tinbergen has developed a method for avoiding this assumption. Its application to American cross-section (state) data did not alter the estimated coefficients greatly.