Simulation Analysis of the Intra-Urban Migration within Kofu City

Abstract

This study focuses on the intra-urban migration. It can be assumed that it is an explanatory variable in terms of the urban spatial structure. Recent quantitative studies applying simulation method to the migration have employed the analytical procedure including three probabilistic models (Fig. 1). The comparison of the each output of simulation with the actual pattern of migration has appealed to the further investigation into the local component unexplained by the distance friction.
In this study, firstly, we tried to investigate the local component by extracting the behavioral pattern of migrants, and secondly, to simulate the actual pattern in consideration of the causes. Monte Carlo simulation models of the Hägerstrand type are formulated in Eqns. (5) to (7). And we applied them to the built-up area of Kofu city, Yamanashi prefecture (Fig. 2). The results are summarized as follows:
1. There are two dominant flows on the movement of migrants (Fig. 5). The first is in the northern part of this study area. The flows start and arrive within the relatively small area. The migrants are mainly composed of the personal and young from twenty to twentyfour years old, including forced migration by the movement of a dormitory. The second is characterrized by in-migrants moving into the southern urban fringe and they move from relatively long distance. There exists the housing development that is newly constructed and offers much opportunity of housing to the migrants. They are mainly composed of the single families with children and the parents are mainly twentyfive to thirtynine years old. Generally speaking, the number of migrants decreases with the increase of distance and the pattern is well fitted to the Pareto function (Fig. 8).
2. Comparing with the output from a complete random process (Fig. 7 a)), the model which includes the distance friction effect based on the above mentioned Pareto function gives us better output (Fig. 7 b)) and it simulates the actual pattern (Fig. 6 b)). However the correlation coefficients between the outputs and the actual pattern are not significant at 5% level (Table 1). Thus it is not true that the distance friction is enough as the determinant of the actual intra-urban migration.
3. Subsequently we employed four indices, i. e., population, number of households, number of households living in owned house and number of households living in rented house, representing the local component. And we applied an estimated value to the housing development including exceptionaly much opportunity of housing. Furthermore we eliminated the number of migrants concerned with the forced migration from Monte Carlo simulation, and then added the number to that of the unit area where they move in. As a result, when using the number of households living in rented house, the output is almost coincident with the actual pattern visually (Fig. 7 c)) and stochastically (Table 1). Thus it can be concluded that such index represents the local component best.
Summing up, this study reveals the importance of the opportunity of housing as the local compornent of the intra-urban migration. The existence of such opportunity pulls the migrants and controls the migration directly. It is the indirect relationship that the migration is closely correlated with the distance as shown as Fig. 8. For enhancing the approximation of the simulated output to the actual pattern, it is necessary to assess, more precisely, the local component of the unit area within the city. According to the outputs of this study, the behavior of each migrant is neither random nor determined by the distance friction. For the future, it is expected that we should formulate the attributes of migrants who assess the above mentioned opportunity individually. However, at that time, the availability of the simulation model as an explanation one must be considered carefully.