Whether social transfers should be targeted or universal is an unsolved debate that is particularly relevant for the implementation of social protection schemes in developing countries. While the limited availability of public resources encourages targeting, the difficulty in identifying the poor promotes a universal allocation of benefits. To address this question, this study examines the targeting performance of, and access to, a social welfare scheme for an increasingly vulnerable group – India’s elderly poor. The results show that during a time period of social pension reforms, exclusion and inclusion errors were successfully reduced but exclusion of the elderly poor continues to be extremely high. Comparing the existing targeting approach to a random allocation, I show that the benefits of targeting are limited. The reforms aimed at increasing the transparency of social pension allocation were indeed achieved, such that possession of the Below Poverty Line ration card has become the primary determinant of access to social pensions. However, this focus on the ration card has its own weaknesses. Nonpoor individuals exploit the unwarranted possession of this ration card and results indicate that after the reforms individuals with direct connections to local government officials are more likely to access social pension benefits. The current targeting approach seems to be beneficial for well-connected and well-informed individuals while many poor elderly lacking connections or information lag behind.