Difference-in-Discontinuities Design
Also known as: diff-in-disc, DiD-RDD, Süreksizliklerde Fark (Difference-in-Discontinuities)
Difference-in-Discontinuities is a hybrid quasi-experimental design that fuses regression discontinuity (RDD) with difference-in-differences (DID), introduced by Grembi, Nannicini and Troiano (2016). It compares the discontinuity at the same cutoff value across two periods to isolate a causal effect.
Key highlights
- Removes confounders that act at the same cutoff in both periods, which a single RDD cannot separate.
- Combines the local credibility of RDD with the differencing logic of DID.
- The continuity assumption can be tested period by period before differencing.
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 this design when an RDD threshold is shared across two periods but is contaminated by simultaneous policy changes or candidate confounders tied to the same cutoff. It suits panel or time-series data with at least about 200 observations so that each period has enough mass near the threshold for a reliable separate RDD fit. It relies on the same cutoff holding in both the treatment and comparison periods, the continuity assumption being testable for each period separately, and no change in behavioural sorting (manipulation) at the cutoff between periods.
Strengths & limitations
- Removes confounders that act at the same cutoff in both periods, which a single RDD cannot separate.
- Combines the local credibility of RDD with the differencing logic of DID.
- The continuity assumption can be tested period by period before differencing.
- Requires a large sample (at least about 200) because each period needs its own reliable RDD estimate near the cutoff.
- If the cross-period continuity assumption fails, the estimate is untrustworthy and standard RDD is preferred.
- Sensitive to changes in manipulation or sorting at the cutoff between periods.
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
How is this different from a plain regression discontinuity?
A plain RDD estimates one jump at the cutoff, which may bundle together every rule tied to that threshold. Difference-in-discontinuities estimates the jump in two periods and subtracts them, cancelling any confounder common to both periods so only the treatment that changed remains.
Why do I need at least about 200 observations?
Each period gets its own separate RDD fit using only points near the cutoff. With fewer than about 200 observations there is too little mass around the threshold per period, so the separate estimates become unreliable and standard RDD is preferred.
What assumption replaces the single-period continuity assumption?
Continuity must hold and be tested for each period separately, the same cutoff must apply in both periods, and behavioural sorting (manipulation) at the cutoff must not change between periods. If cross-period continuity fails, fall back to standard RDD.
How is the effect actually computed?
A weighted local regression is fitted to each period to estimate its discontinuity at the cutoff, then the difference-in-discontinuities estimate is the after-period jump minus the before-period jump.
Sources
- 1.Grembi, V., Nannicini, T. & Troiano, U. (2016). Do Fiscal Rules Matter? A Difference-in-Discontinuities Design. American Economic Journal: Applied Economics, 8(3), 1-30.
- 2.Cattaneo, M. D., Idrobo, N. & Titiunik, R. (2020). A Practical Introduction to Regression Discontinuity Designs: Foundations. Cambridge University Press.ISBN 978-1108710206
You have read it. What now?
Cite this page
ScholarGate. (2026, June 1). Difference-in-Discontinuities. ScholarGate. https://scholargate.app/causal-inference/difference-in-discontinuities