My dissertation focuses on explaining the incumbency disadvantage: the counter-intuitive phenomenon in which incumbents are systematically less likely to win elections than similar non-incumbents. This anti-incumbent bias is prominent in many developing democracies, in stark contrast to the well-known incumbency advantage in the U.S. and other mature democracies. My three-paper dissertation examines an important potential cause: corruption. I show theoretically and empirically that that the incumbency disadvantage is prevalent when the opportunity cost of corruption is low and incumbents’ gains from corruption increase during their tenure, because of learning on the job or the time it takes to set up rent-extraction networks. These factors may force voters to "shuffle the deck" frequently by throwing incumbents out of office in order to prevent a large corruption increase, even if the challengers are deemed corrupt as well.
In this paper, forthcoming in the Journal of Theoretical Politics, I carefully formalize the conditions in which corruption leads to incumbency disadvantage. I show that the incumbency disadvantage becomes more likely as: the cost to being corrupt decreases, increasing incentives to engage in corruption; as the quality of the pool of potential incumbents deteriorates, increasing the probability that any incumbent is corrupt; and when the returns from corruption increase over the course of an incumbent's tenure (e.g. because of learning on the job or the time it takes to develop a rent-extraction network), forcing voters to minimize corruption increase by replacing incumbents frequently, even if challengers are also perceived as corrupt. I further show that increasing corruption can make the incumbency disadvantage quite persistent, even if incumbents over time become more competent as well as corrupt; if their preferences are largely aligned with the voters'; and sometimes even when the candidates' integrity is improving.
The remainder of the dissertation is devoted to testing these predictions. My empirical focus is mayoral elections in Romania. Romania offers rich data to measure corruption and precisely estimate the incumbency disadvantage, and its unusual institutional features facilitate a research design that uncovers the causal effect of corruption and increasing rate of corruption on the incumbency disadvantage.
In this paper, I test the main predictions of my theory on Romanian mayoral elections, using fine-grained measures of local corruption, discussed in the remaining dissertation chapter (below). To tackle the challenge of isolating the effect of corruption from other potential causes of the incumbency disadvantage correlated with corruption, I exploit an exogenous variation in corruption afforded by an unusual rule in Romania that mayoral salaries sharply change at thresholds determined solely by a town's population size. Other scholars have shown, and I confirm, that higher salary can lower corruption by increasing the opportunity cost to being corrupt and attracting more upstanding candidates for office. These channels are closely related to my theoretical predictions. I combine this identification strategy with another: the regression discontinuity design that compares electoral winners and losers in close elections to detect and isolate incumbency advantages. In concert, these two strategies produce strong support for my theoretical predictions. There is a large incumbency disadvantage in Romanian mayoral elections, and this disadvantage is much higher in places where the incidence of corruption is higher. I also find strong evidence that Romanian mayors on average become more corrupt the longer they stay in office and that thus more experienced incumbents are more strongly disadvantaged in high-corruption localities.
This paper grapples with a fundamental problem of studying corruption empirically: measurement. Because corruption is covert, it is difficult to detect. Most available corruption measures are based on surveys of individuals' perceptions and typically aggregated at the country level. My contribution is to create village-level and city-level measures based on objective rather than survey data. With the help of research grants from the National Science Foundation and New York University, I coded more than 4,000 asset declarations of mayoral candidates to detect abnormal wealth accumulation among incumbent mayors; collected over three million public procurement contracts to identify tenders with suspiciously high prices, uncompetitive bids and opaque procedures; constructed a measure of leakage and waste in infrastructure spending; and geo-coded close to 3,000 corruption cases prosecuted by Romanian anti-corruption agencies.
End of page. Get back to top