Causes and Consequences of Elections in Nondemocracies

In this dissertation, I offer an answer to one of the most important questions about authoritarian politics today: why do dictatorships hold elections? In order to answer this broad question, I study the causes and consequences of elections as well as the role of elections in nondemocratic settings. First, I develop a theory about the causes of elections in dictatorships, which is based on the different threats that dictatorships face and the different goals that they have in order to lessen or avoid these threats. I argue that dictatorships opt for elections for the effective executive if they need to avoid violent removal. In contrast, dictatorships begin elections for a national legislature if they seek to maintain the unity and cohesion of elites in the ruling circle and/or to coopt elites from outside of the regime. Second, I present a theory about the consequences of elections in dictatorships. I contend that two seemingly competing effects of elections are mutually complementary. Individual elections can create a momentum for regime change, leading to the collapse of dictatorships and democratic transitions. At the same time, once dictatorships survive elections, election results convey useful information for the purpose of cooptation and send a signal that deters future challenges to the regime. Tests of my theory on a sample of dictatorships after World War II show robust support for my theory about the causes and consequences of elections. Finally, I revisit the information collection role of elections in nondemocratic settings. I theorize that elections can be either informative or less informative depending on the strategic decisions that major opposition parties make. I develop a formal theory to describe this causal mechanism. An important implication of my theory is that informative elections are associated with post-electoral redistribution of goods and patronage while less informative elections in which major opposition parties boycott elections are not. I test this implication by using original data collected from Serbia in 1990s and present results that are consistent with my theory.