In this paper, we estimate the effect of social transfers on disposable income inequality for a large panel of OECD countries. We account for the potential simultaneity bias exploiting an instrumental variable derived using a theoretical model, which identifies the main driver of social transfers from the interaction between electoral systems and partisanship of coalitions winning elections. The empirical analysis shows that a 1% increase in social transfers reduces income inequality by half a percentage point. This analysis also reveals significant heterogeneous results conditional on the level of corruption and appears to be robust to different components of expenditure, alternative model specifications and falsification tests.