The pattern of Japanese family structure has been marked by a high number of extended families living together. This characteristic indicates that the role of the family in securing a comfortable retired life for the elderly is quite im- portant in Japan and is a major source of “age selectivity bias” in household- based statistics. The economic position of the elderly and their savings rate as conventionally measured are largely overstated because these data tend to ex- clude the poor elderly who are economically dependent members of their chil- dren’s families. This not only explains why the savings rate of the Japanese elderly is so much higher than that of the U.S. elderly but also predicts a de- cline in savings rate in the near future given the persistent increase in elderly persons living alone. The economic position of the Japanese elderly on average is almost equiva- lent to that of the nonelderly. In addition, accounting for imputed income fur- ther narrows the gap between the elderly and the nonelderly. While the income of the majority of households headed by the elderly may be low, their asset level is significant, thanks to past asset inflation in Japan, so that the elderly may persuade their children to live with them by offering rich housing services. Empirical estimates suggest that this strategic bequest motive may well be im- portant in determining the living arrangements of the elderly, despite their chil- dren’s feelings about shared living. Finally, the income and wealth distribution of the elderly is much wider than that of the nonelderly and is more concentrated in the lower income classes. This implies that welfare policy targeting the elderly per se would be ineffi- cient, particularly given increasing fiscal constraints. The policy of concentrat- ing welfare resources on the poor elderly, rather than on the elderly population as a whole, would be effective.