Although re-election prospects can improve policy implementation by incumbents, they can also create incentives for politically-motivated targeting, which might jeopardize distributional efficiency. While existing empirical tests typically focus on these potential countervailing incentives in isolation, this paper analyzes their net effect in the context of Bolsa Família (BF) in Brazil, using a regression discontinuity design and data on seven million households from the program's registry. The evidence supports political targeting over accountability: mayors with re-election incentives are four times more likely to include nonpoor, ineligible households in the policy. These results cannot be explained by higher effort in indiscriminate program expansion. On the contrary, evidence from both a survey with 11,000+ households and the heterogeneity in the estimates suggests that they reflect a breakdown on the information channels that lead the excluded poor to hold local administrations accountable. On the other hand, the included nonpoor are more likely to support incumbents in their re-election attempts, as they fear losing the benefit should the administration change. Finally, anomalous income reporting patterns also show that this electorally-driven targeting is more common for households enrolled by public servants politically connected to the mayor.