A COMPARISON OF CROP PRODUCTION-FUNCTIONS USING SIMULATED DATA FOR IRRIGATED CORN IN WESTERN KANSAS

A synthetic approach is used to deal with lack of available yield data for production function estimation. Corn yield data from the CERES-Ma ize simulator for western Kansas are used to estimate yield response t o nitrogen and irrigation water using five functional forms: quadratic , square root, Mitscherlich-Baule; linear von Liebig; and nonlinear vo n Liebig. Each is compared with the others using non-nested hypothesis tests. The Mitscherlich-Baule function is found to be favored over ea ch of the other specifications using pair-wise comparisons as well as a joint test. However, this model also was rejected in favor of the no nlinear von Liebig function in a pair-wise comparison. Optimal input l evels are comparable with research based on actual yield data. Optimal nitrogen levels are found to be somewhat higher using simulated data than using actual data, with optimal water levels being greater and le ss than the average amount used in the area. Costs of incorrectly usin g the Mitscherlich-Baule are relatively low. Substantial costs exist w hen the linear von Liebig is incorrectly used. The costs of using the nonlinear von Liebig when the Mitscherlich-Baule function is correct a re higher than incorrectly using the Mitscherlich-Baule. (C) 1997 Else vier Science Ltd.