In federal democracies, parties often invest in local politics as a strategy to improve their performance on upcoming national elections. In this study, I use the concept of reverse coattails to investigate how winning local elections affect upper-level electoral dynamics in Brazil. Using a regression discontinuity design (RDD), I show that parties in Brazil boost their national performance, earning more votes on House elections in districts where their members control local offices. I discuss how access to “pork” controlled by co-partisan House members and mechanical information gains explain these effects. Additionally, I use a Bayesian LASSO algorithm to address data sparsity in RDD designs, and to demonstrate the existence of prolarge party bias on the coattail effects. By disentangling the various effects of winning local elections, this paper contributes to a greater understanding of how parties build electoral strength in fragmented democracies.