This paper uses prospective data from the Matlab surveillance system in rur
al Bangladesh to demonstrate that initially co-resident spouses and sons ha
ve a major impact on the subsequent mortality of old people, with significa
nt differences by the sex of the elderly person, and the age of the son. Sp
ouses significantly reduce mortality by similar magnitudes for both elderly
men and women. On the other hand, co-resident adult sons reduce mortality
for elderly women much more than for elderly men, with younger sons being m
ore beneficial than older sons. Furthermore, both married and unmarried fem
ales appear to benefit equally from co-resident adult sons. Finally, this a
nalysis suggests that the impact of spouses and sons on mortality in old ag
e is not substantially mediated through changes in elderly economic status.