In this paper, we analyze changes in determinants of internal long-distance elderly migration using 1960, 1970, 1980, 1990 and 2000 National Censuses. First, we show the characteristics of elderly migration. Percent of elderly migrants has risen gradually from 3.0% (1960) to 10.3% (2000). Upheaval in migration rate at elderly ages with more aged people having higher migration rates, confirmed for 1970 and clearly seen in 1990 and 2000. Migration of young old has changed from concentration toward major metropolitan areas to that toward to rural areas. On the other hand, migration of old old continues to concentrate to major metropolitan areas. The adjusted gravity model and net migration model were used in this paper. Regression analysis were used for all cases. Dependent variables were numbers of migrants (AG model) and net migration rates (NM model). Explanatory variables were (1) index of potential migration living together with/near children's household (CHI), (2) index of potential return migration to 1940 living place (RMI), (3) index of amenity (average temperature: AT), (4) number of doctors per area (DA), (5) consumer price index (PI), (6) index of potential return migration to birth places (BPI), (7) number of inmates of nursing home per population (NH). Area divisions were 10 areas and 46 prefectures. The main results of our paper are summarized as follows. (1) Only CHI had strong explanatory power from 1960 to 2000 for both young old and old old. (2) DA had strong explanatory power in 1960 for young old and from 1980 to 2000 for old old. (3) For young old RMI, DA, AT and CHI had strong explanatory power in 1960, and the determinants of migration for young old changed to AT, PI, BPI and CHI in 1990 and to PI, BPI, NH and CHI in 2000. (4) For old old the same determinant of migration for forty years was only CHI, but AT and DA had explanatory power from 1980 to 2000. JEL Classification: R23