High rates of pregnancy and childbirth-related mortality in developing countries have been linked to high rates of use of informal sources of childbirth care. Many of these deaths are believed to be preventable, leading to large investments in programs designed to promote use of formal care at the time of childbirth. These policies range from the benign, such as the distribution of cash transfers, to the punitive, such as government bans on the use of informal birth attendants. There is a contentious debate about the extent to which these policies will affect mortality. This paper sheds light on this question by examining the marginal returns to an institutional birth. To identify causal effects we exploit random variation in weather around the time of birth. The hypothesized channel is that inclement weather at the time of birth will deter some women from traveling to a health institution to give birth. We leverage unpredictability about the weather and unpredictability about the timing of birth. To implement this strategy we merge weather data from the University of Delaware's Center for Climatic Research with birth outcomes data from the India District Level Household Survey. We show that as predicted, rainfall shocks reduce the probability of an institutional birth (by about one percentage point). We show that, consistent with the hypothesized channel, this effect is present only in the month of birth and is observed only for women who live further away from a hospital. Using rainfall shocks as an instrument we find that women who have an institutional birth are 60 percentage points less likely to experience a post-delivery health complication. This effect is large and robust to various specification changes. To put the magnitude of this effect into context, we estimate that increasing the rate of institutional deliveries in India by 50% would decrease post-delivery complications by 11 percentage points and decrease maternal mortality by about 12%.