Most studies find that the presence of grandmothers is associated with higher child survival rates while the presence of a grandfather has no effect. In this paper we argue that the use of incomplete retrospective information on household composition may lead to substantial downward biases in the estimates of the grandfather’s effect. To illustrate our argument, we study the effects of the introduction of a non-contributory universal pension scheme on infant mortality in Nepal. We use cross-sectional data from the 1996 and 2001 Nepal Demographic and Health Surveys and implement a flexible difference-in-differences approach using as treatment indicator the presence in the household of an elderly person eligible for the pension benefits. Our results confirm the importance of the income of the elderly for infant welfare. In households that include an eligible elderly person, the pension scheme resulted in survival benefits of approximately 7–8 percentage points 12 months after birth from a baseline probability of approximately 89%. Importantly, these results do not depend on the gender of the elderly person. Moreover, we show that ignoring the lack of retrospective information on household composition leads to underestimates of the true effects when the elderly person is a male. These results are qualitatively similar across alternative definitions of the control group and do not depend on the gender of the infant. In addition, we explore potential channels through which these survival benefits are produced, and our results suggest that the pension scheme benefited infants only after their birth and mostly during the first month of life.