Abstract
The theoretical framework of our Global METS Model was presented in the previous volume of Papers of this Association (Reference [6]). As a continuing project, we have estimated the empirical structure of the model and also have made some simulation analyses of the model for the future in order to trace the economic effect of the nation-wide transprot investment project in an interregional setting.
The statistical estimation of the model has been carried out by the direct least squares method after reformulating the model as a recursive system mainly using 1953-1967 regional data of Japan. This simplified method of estimation is inevitable because of a tremendous number of individual equations in our model. (We have 3100 equations approximately.)
We have been forced to change the functional form of a large number of the estimated equations to improve the performance of the structure as a whole when it has been simulated for the past by the so-called final test technique. Some of such changes have regrettably involved theoretical setback from the original formulation of our Global METS Model. The main points of the changes are:
(1) Abolishment of the gravity-type model concerning interregional commodity flows. As a substitute, a simplified version of the Frator method using the marginal distribution of the origin-and-destination table has been employed.
(2) Deletion of the congestion variables in the interregional and modal distance functions.
(3) Alteration in the formula to derive realized regional income produced. Instead of Min (TVP, TVE), we employed a weighted average approach: (αTVP+βTVE).
Three kinds of conditional forecasting based on three different hypothetical past and future time paths of the exogenous variables (1967-1985) have been performed. The three time paths can be described as trend type, priority-in-road construction type I, and the same type II (emphasis on the decentralization policy). Contrary to our expectation, the three methods have not produced any big difference between their forecasts. For the first case, we have had more than 10% average annual growth rate of the produced income for the whole nation until 1985 (Hokkaido_??_8.6%, Tohoku_??_8.3%, Kanto_??_12%, Tokai_??_10%, Hokuriku_??_10%, Kinki_??_9.7%, Chugoku_??_11%, Shikoku_??_12%, and Kyushu_??_7.4%).
Our forecasting model must be revised in many respeets including the establishment of a unified method of project evaluation.
