Does Regression Discontinuity Design Work? Evidence from Random Election Outcomes

We use data for 198121 candidates and 1351 random election outcomes to estimate the effect of incumbency status on future electoral success. We find no evidence of incumbency advantage using data on randomized elections. In contrast, regression discontinuity design, using optimal bandwidths, produces a positive and significant incumbency effect. Using even narrower bandwidths aligns the results with those obtained using the randomized elections. So does the bias-correction of Calonico et al. (forthcoming). Standard validity tests are not useful in detecting the problems with the optimal bandwidths. The appropriate bandwidth seems narrower in larger elections and is thus context specific.