Incumbency Effects in German and British Elections: A Quasi- Experimental Approach

Following the recent turn towards quasi-experimental approaches in the US literature on the incumbency advantage (Lee, 2001; Lee, forthcoming), we employ a Regression Discontinuity Design (RDD) to identify the causal effects of party incumbency in British and German post-World War II elections. The RDD framework exploits the randomized variation in incumbency status that occurs when a district race is close. Based on the assumption that parties do not exert perfect control over their observed vote shares, incumbents that barely won a race should be similar in their distribution of observed and unobserved confounders to non-incumbents that barely lost. This provides us with a naturally occurring counterfactual exploitable for causal inference under a weaker set of assumptions than conventional regression designs commonly used in the incumbency literature. In both British and German federal elections, we find that party incumbency has a signifcant positive impact on vote shares and the probability of winning in marginal districts, the sub- population of interest for which incumbency advantage is likely to make a difference. This stands in contrast to previous more ambiguous findings.