Why do some political machines choose to pursue a clientelist strategy whereas other parties deploy non-clientelist strategies? And why do some clientelist parties succeed in mobilizing low-income voters while others fail? In a world in which political machines flourish in many new democracies, understanding the conditions under which such machines choose to pursue a clientelist strategy and succeed or fail to win votes with clientelist inducements is of critical importance to political scientists and policy makers. Drawing on a detailed study of strategic choice and variation in the performance of political machines in Argentina, my dissertation will build a theory about machine politics in new democracies. I hypothesize that it is the combination of the amount of state resources to which the machine has access, the efficiency with which it distributes these resources, and its ability to monitor voters and make good inference about how they voted that explains the success or failure of clientelist parties. I will test my hypothesis by making comparisons across political parties, across regions in Argentina, and, through survey research, across individual voters, in order to identify the factors that determine parties' strategic choice and their failure or success. Although my dissertation focuses on Latin America and, more specifically, on Argentina, it will provide theoretical and methodological tools with which to study political machines in other new democracies.