The Electoral Advantage to Incumbency and Voters' Valuation of Politicians' Experience: A Regression Discontinuity Analysis of Close Elections

Using data on elections to the United States House of Representatives (1946-1998), this paper exploits a quasi-experiment generated by the electoral system in order to determine if political incumbency provides an electoral advantage – an implicit first-order prediction of most principalagent theories of politician and voter behavior. Candidates who just barely won an election (barely became the incumbent) are likely to be ex ante comparable in all other ways to candidates who barely lost, and so their differential electoral outcomes in the next election should represent a true incumbency advantage. The regression discontinuity analysis provides striking evidence that incumbency has a significant causal effect of raising the probability of subsequent electoral success –by about 0.40 to 0.45. Simulations – using estimates from a structural model of individual voting behavior – imply that about two-thirds of the apparent electoral success of incumbents can be attributed to voters’ valuation of politicians’ experience. The quasiexperimental analysis also suggest that heuristic “fixed effects” and “instrumental variable” modeling approaches would have led to misleading inferences in this context. * Matthew Butler provided outstanding research assistance. I thank John DiNardo for numerous invaluable discussions, and Josh Angrist, Jeff Kling, Jack Porter, Larry Katz, Ted Miguel, and Ed Glaeser for detailed comments on an earlier draft. I also thank seminar participants at Harvard, Brown, UIUC, and Berkeley for their additional useful suggestions.