In 2007 the state of Andhra Pradesh in southern India began rolling out Aarogyasri health insurance to reduce catastrophic health expenditures in households 'below the poverty line'. We exploit variation in program roll-out over time and districts to evaluate the impacts of the scheme using difference-in-differences. Our results suggest that within the first nine months of implementation Phase I of Aarogyasri significantly reduced out-of-pocket inpatient expenditures and, to a lesser extent, outpatient expenditures. These results are robust to checks using quantile regression and matching methods. No clear effects on catastrophic health expenditures or medical impoverishment are seen. Aarogyasri is not benefiting scheduled caste and scheduled tribe households as much as the rest of the population.