So far some economists have studied the pattern of traffic assignment theoretically and empirically. But it seems that there are few proper stochastic or statistical models for analyzing the traffic assignment problem on the base of an economic theory. The purpose of this paper is to formulate the economic behavior of traffic assignment and to derive some statistical models which seem useful for empirical study or the prediction of the future course.
Three diffirent models for traffic assignment were developed. The first is similar to Quandt-Baumol model (1966) by which they analyzed the social traffic demand in California. In that model it is suggested that individual will choose the transportation modes so as to minimize his loss prescribed by traffic time and cost of each mode. However in deriving the social demand function from individual's one, we have had no consistent procedure concerning aggregation.
The second model was developed by a stochastic approach to overcome the difficulty in the aspect of aggregation of the first model. The traffic time and cost are main factors to assign the modes in that model too.
The thrid model was proposed for prediction in the future course so that model always satisfys the condition that the sum of predicted values of each assignment ratio be strictly unity.
Finally some empirical applications on the basis of the second and thrid model were attempted. It seems that such applcations were rather succesful.
