Throughout much of mankind's experience with elections, vote brokers – local elites who direct the voting decisions of a subset of the electorate – have been able to make or break political careers. In various polities, brokers have thrived in spite of the secret ballot, a surprising outcome given that vote secrecy would ostensibly allow citizens to pocket the inducements offered by such individuals and vote their consciences anyway. To address this puzzle, we develop a framework for understanding the persistence and demise of vote brokerage under the secret ballot. In our model, a broker contracts with voters using an outcome contingent contract: some fixed benefit is promised to all voters sharing one of several observable profiles should the broker's candidate win the election. Using this framework, we demonstrate that the existence of brokerage depends on the size of the electorate contained within the jurisdiction controlled by the broker, with large jurisdiction sizes tending to drive brokerage out of existence. Moreover, we detail the manner in which the strategies employed by brokers depend on their economic power, the size of social groups, and ideological polarization. Empirical evidence from Minas Gerais, Brazil is used to evaluate the performance of the model.