Regression Discontinuity in Elections
Also known as: Close-election RD, Electoral regression discontinuity, Vote-share RD design, Incumbency-advantage RD
Regression discontinuity in elections is a quasi-experimental design that exploits the sharp winning threshold in electoral contests to estimate causal effects of holding office. Just above the threshold a candidate or party wins; just below, it loses. In very close races, which side ends up just over the line is plausibly as good as random, so comparing the later outcomes of bare winners and bare losers identifies the causal effect of winning — most famously the incumbency advantage — without confounding by candidate or district quality.
Key highlights
- Delivers credible causal estimates from observational electoral data by exploiting as-good-as-random outcomes in close races.
- Transparent and testable: balance, density, and placebo checks let analysts probe the identifying assumption.
- Requires weaker assumptions than most observational designs, approaching the credibility of a randomized experiment near the threshold.
- Validated at scale across many electoral settings, with broad evidence that the design generally holds.
Intuition
This section is available to Pro members. Upgrade to Pro
How it works
This section is available to Pro members. Upgrade to Pro
When to use it
Use regression discontinuity in elections when you want the causal effect of winning office or crossing an electoral threshold and you have many contests with vote margins near the cutoff and measurable downstream outcomes. It is the standard tool for estimating incumbency advantage and the effects of party control. It is less appropriate when close races are few, when candidates or parties can precisely manipulate the margin (sorting around the threshold), or when the effect of interest is for landslide rather than marginal contests, since the estimate is local to the threshold.
Strengths & limitations
- Delivers credible causal estimates from observational electoral data by exploiting as-good-as-random outcomes in close races.
- Transparent and testable: balance, density, and placebo checks let analysts probe the identifying assumption.
- Requires weaker assumptions than most observational designs, approaching the credibility of a randomized experiment near the threshold.
- Validated at scale across many electoral settings, with broad evidence that the design generally holds.
- Estimates a local effect at the threshold (for marginal races), which may not generalize to non-close contests.
- Requires many observations near the cutoff; sparse close races yield imprecise estimates.
- Vulnerable to manipulation if actors can precisely sort vote margins around the winning threshold.
- Sensitive to bandwidth and functional-form choices, which can affect the estimated discontinuity.
Common pitfalls
This section is available to Pro members. Upgrade to Pro
Applications
This section is available to Pro members. Upgrade to Pro
Frequently asked
Why are close elections treated like a randomized experiment?
In a very close race, the final vote margin turns on factors no candidate can precisely control — turnout shocks, weather, counting noise — so whether a candidate lands just above or just below the winning threshold is essentially random. Among these near-ties, bare winners and bare losers are statistically interchangeable on pre-determined characteristics, just as random assignment would make treatment and control groups comparable. The jump in later outcomes at the threshold therefore reflects the causal effect of winning rather than candidate or district quality.
What does the McCrary density test check?
The McCrary test examines whether the density of the running variable (the vote margin) is continuous at the threshold. A discontinuity in the density — for example, a pile-up of cases just above the winning cutoff — suggests that actors can precisely manipulate or sort around the threshold, which would violate the local-randomization assumption and bias the estimate. A smooth density supports the claim that close outcomes are as-good-as-random and the design is valid.
Is the close-election RD estimate generalizable to all elections?
Not directly. RD identifies a local average treatment effect for contests decided near the threshold — that is, for marginal, highly competitive races. The incumbency advantage or office effect in landslide contests, with very different candidates and districts, may differ. The estimate is most credible precisely where it is local, so researchers interpret it as the effect at the margin and are cautious about extrapolating it to non-close elections.
Sources
- 1.Lee, D. S. (2008). Randomized Experiments from Non-random Selection in U.S. House Elections. Journal of Econometrics, 142(2), 675–697.
- 2.Lee, D. S., & Lemieux, T. (2010). Regression Discontinuity Designs in Economics. Journal of Economic Literature, 48(2), 281–355.
- 3.Eggers, A. C., Fowler, A., Hainmueller, J., Hall, A. B., & Snyder, J. M. (2015). On the Validity of the Regression Discontinuity Design for Estimating Electoral Effects: New Evidence from Over 40,000 Close Races. American Journal of Political Science, 59(1), 259–274.
You have read it. What now?
Cite this page
ScholarGate. (2026, June 22). Regression Discontinuity in Elections. ScholarGate. https://scholargate.app/political-science/regression-discontinuity-in-elections