鶏卵需要の計量的分析

Two objectives underlie this study: The first is a quantitative analysis of the demand for eggs in Japan and the second is an examination of the merits of the simultaneous-equations method. The selection of eggs for investigation has been influenced by the facts that the poultry industry of Japan is a important contributor to gross farm income, and eggs are the most important component of gross poultry income. It was seemed advisable to choose an industry where current period economic influences on supply might be expected to be strong and where interrelations between demand and supply would consequently be seen. In general, it is in these situations that the simultaneous-equations method should be expected to assert its superiority and therefore provide a suitable empirical testing ground for the alternative procedures. The selection of egg industry also has been influenced by this aim. In order to work with the simultaneous-equations approach, three complete models of the egg industry were formulated. Model I and Model II were formulated to measure the demand relationships of eggs in the post war years and Model III to measure the ones in the pre war years. Model I and Model II involve equations relating to production, domestic use, margin, and an identity that states that the domestic use must equal the total supply. In addition to the above four equations, Model IIIinvolves more an equation relating to import. Model II differs from Model I to the extent that the variable contained in the supply relationship in Model II is substituted for a part of the variables contained in the same relationship in Model I. The demand relationship in Model III differs from the same relationships in Model I and Model II to the extent that the demand relationship in Model III contains variables representing quantities of supply of items thought to be substitute or complementary goods with eggs. Model I and Model II were estimated with annual datas from the period 1950-59 and Model III from the period 1924-37. These three egg models are nonlinear equation systems. Several modifications were made in fitting parameters to the models. Because the equations in the models must be stated in linear terms, therefore the endogeneous variables in the models used in nonlinear combinations were transformed into linear approximations by use of formula in Klein. The linear approximations were thensubstituted for the original variables. Combinations of variables that are assumed to be entirely predetermined were treated as a single composite without lineari-zation as they are assumed to be given. In fitting parameters to the models, both the least-squares method and the limited-information method were used. Coefficients obtained by fitting the demand relationships in all the models by the limited-information approach differ from the least-squares results for the comparable demand relationships. But coefficients obtained by fitting the demand relationships in Model I and Model II by the limited-information method differ more from the least-squares results for the comparable demand relationships than do coefficients obtained for the same relationships in Model III. This seems intuitively, not statistically, to be partly due to the structural changes in the egg supply. In the pre war years the general distribution of farms had poultry and where flocks of very small size were predominant and eggs were produced largely or wholly for the use of the farm family. This small scale egg production was a contributory factor to the nonse-nsitiveness to price. In the post war years technological improvements in egg production have substantially reduced costs. Many of the technological improvements have been most practical for the owners of large flocks. Accordingly, there is a trend away from numerous small flocks and toward fewer and larger flocks. It is considered that this tendency has moderated the degree of the nonsensitiveness to price. Coefficients obtained by fitting the demand relationships in Model II by the limited-information approach differ from coefficients obtained by fitting the same relationships in Model I by the same approach. These results indicate that the limited-information estimates of a given equation may be quite sensitive to addition or omission of predetermined variables in the remainder of the systems. The values of the Durbin-Watson d-statistics for some of the equations fitted by the least-squares method are small or large, indicating that the hypothesis of serial independence in the residuals is rejected, but the values of the d-statistics for all the equations fitted by the limited-information method show non-serial-correlation of residuals or fall in the inconclusive test range. It appears to be impossible at the present time to draw reliable over-all conclusions about the superiority of the limited-information method, but it seems fairy safe to state that the limited-information method is superior to the single-equat ...