This paper aims at describing and explaining the changing distribution of university graduates in Japan in order to clarify the effects of interregional migration on changes in the social class composition of the Tokyo metropolitan area (TMA) and the rest of Japan (Non-TMA) . In Japan, it is still under discussion, whether the university graduates (UGs) have been intensified their concentration into TMA for the last a few decades or they have dispersed to Non-TMA. The analysis thereby begins with the distribution of the UGs by 5-year cohort in 1970, 1980 and 1990 measured by location quotients (LQs) of UGs in the TMA. It yields the following results; First, if compared the LQs within a cohort, for example the LQ of the population aged between 30 and 34 in 1970, the LQ of the population aged between 40 and 44 in 1980 and the LQ of the population aged between 50 and 54 in 1990, the LQs show almost the same values, that is, after their thirties each cohort group does not change their spatial distribution of the UGs, and the distribution pattern at the age around 30 determines the later distribution of the cohort. Second, if compared the LQs between cohorts, for example the LQ of the population aged between 30 and 34 in 1970 (1936-1940 cohort) , the LQ of the population aged between 30 and 34 in 1980 (1946-1950 cohort) and the LQ of the population aged between 30 and 34 in 1990 (1956-1960 cohort), we can find a tendency that the newer cohort has a disperse distribution pattern of the UGs.
The paper then proceeds to an analysis of the determining process of the distribution pattern of the UGs by 5-year cohort. The results are as follows; First, the determining process can be split into two stages, the ratios of the teenagers who go on to universities and net migration of UGs and Non-UGs after their graduation from high school. Second, concerning the first stage, the disparity of the ratios between TMA and Non-TMA has been reduced obviously for the last a few decades. The ratios of the cohort born between 1941-1945 are 32.6% (male) and 7.8% (female) in TMA, while 16.1% (male) and 3.2% (female) in Non-TMA. The disparity of the ratios of the 1961-1965 cohort is far small, 41.5% (male) and 17.1% (female) in TMA, while 33.2% (male) and 11.5% (female) in Non-TMA. Third, concerning the second stage, the direction of the net migration is always from Non-TMA to TMA. If we observe the social composition in net migrants by cohort, we can find changing effects of net migration on the distribution of UGs. In the cohorts of 1941-1945, 1946-1950 and 1951-1956, most of the net migrants from Non-TMA to TMA are Non-UGs, thus the ratio of UGs in TMA lowered a lot by the net migration and the disparity between TMA and Non-TMA was reduced. On the other hand, in the cohorts of 1956-1960 and 1961-1965, the net migrants from Non-TMA to TMA decreased in number and in particular Non-UGs decreased apparently because of the transformation of industrial structure in TMA in the 1980s. As a result, the ratio of UGs in TMA rose and the net migration expanded the disparity between TMA and Non-TMA.
The tendency that the newer cohort has a disperse distribution pattern of the UGs has caused by reducing disparity of the ratio entering universities between TMA and Non-TMA (from the 1941-1945 cohort to the 1961-1965 cohort) and by the effect of the net migration (from the 1941-1945 cohort to the 1951-1955 cohort).