In federal democracies, parties often invest in local politics for national gains. This strategy is crucial, particularly in noisy democracies where politicians need to find information shortcuts to attract inattentive voters. I define this strategy as political boosting and investigate its dynamics in Brazil, a textbook example of a fragmented party system. Using a regression discontinuity, I show that parties in Brazil boost their national performance earning more votes on House elections in districts where their members control the local office. I discuss how higher access to pork controlled by House members and information gains from local incumbency explain the effects. Using a Bayesian LASSO algorithm to address data sparsity in RD designs, I further show the existence of pro-large party bias. By disentangling the effects of boosting after winning local non-concurrent elections, the paper explains the general logic by which parties build electoral strength in fragmented democracies.