TIME‐SERIES MODELING OF GROSS MIGRATION AND DYNAMIC EQUILIBRIUM*
研究了迁入与迁出之间的高关联性，构建了双向反馈的双变量时间序列模型，并利用日本32个区域的数据验证了模型，发现人口从核心区向边缘区的扩散主要由回迁移民的反馈驱动。
Firstly, the high association between in- and out-migration is investigated in a time-series context and modeled according to three categories: 1) job transfer, 2) job search and marriage, and 3) return migration. Under certain coditions it is shown that aggregation of these migrations yields a bivariate time-series model having feedbacks in both directions. Secondly, the recent phenomenon of sharp changes in net migration seems to be discontinuous and, hence, catastrophic modeling [Casetti (1981) may be appropriate. However, this paper considers gross migration between cores (metropolitan areas) and peripheries (rest of the nation) for which a continuous function seems adequate. This is done by introducing a multivariate time-series model. This model is empirically supported, especially in Japan, divided into 32 regions, by t-tests and Durbin-Watson ratios, although it excludes economic variables such as employment growth and wage differentials. This may imply that the recent dispersal from core to peripheral regions could be explained primarily by feedback from return migrants. Finallym, provided future streams of gross migration follow the past trends given by simultaneous equation estimates, in-migration and out-migration would approach a stable state in most regions. Irrespective of random shocks in the future, in- and out-migration would tend to approach a stable equilibrium. According to the estimation of the stable states, the 45 core regions in the US would continue to lose population through net outflows while those in Japan would continue to gain. The present model may thus be valid only for short-term forecasts. By introducing feedback and lag structures, however, it does offer one explanation for the recent population turnaround.