Author : Mendoza, Maria Nimfa F.

Summary : Risk is primarily defined as variability of outcome in conjuction with expected outcome. In a broader sense, risk estimation covers, in the case of rice production, the estimation of the probability distribution of rice yield. The major intent of this study was to investigate the sensitivity of risk measurement to functional form of an estimation technique for the production function. Specifically, the effect of nitrogen input on the probability distribution of rice yield outcome in nine agronomic zonal levels stratified according to season and water stress conditions was measured. Stochastic production functions incorporating manageable inputs and environmental factors were estimated for 1972-1977 rice data from farmers' fields in Central Luzon, Philippines. Five basic models with different sets of linear and interaction stress variables were used. The multiple regression model with normally distributed zero mean constant variance errors was assumed to approximate the production process. Initial coefficient estimates are obtained by ordinary least squares. For a sweep among empirical models indexed by transformation of the response variable, Box-Cox regression was performed. With the obtained response functions as input information, rice yield distributions at fixed levels of nitrogen and given solar radiation and water stress conditions were simulated. As descriptors of risk effects of nitrogen, the means, variance, and skewness of the simulated yields distributions were analyzed. It was observed that the risk effects of nitrogen is highly conditioned by solar radiation and water stress. Results indicated that the method of risk measurement used was not sensitive to the production function estimation technique. Therefore, the superiority of the simple and computationally economical least squares estimation for the standard multiple regression model was asserted. Differences in inferences on risk affects between models indicate that model formulation poses a greater problem.