Panel Data Regression Discontinuity Design
Also known as: Panel RD, Panel RDD, Longitudinal Regression Discontinuity, Fixed-Effects RDD
Panel data regression discontinuity design (Panel RDD) combines the sharp local identification of a regression discontinuity with the within-unit variation available in repeated-observation panel data. Units are observed across multiple periods, and treatment is assigned based on whether a running variable crosses a known threshold. By leveraging both the discontinuity and panel structure, researchers can control for unobserved unit-level heterogeneity while estimating a causal treatment effect near the threshold.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Use panel RDD when treatment is assigned by a known threshold on a continuous score, units are observed in multiple periods, and you want to combine the local randomisation of RDD with within-unit fixed-effect control. Ideal settings include repeated means-testing eligibility, annual exam-based programme admission, and regulatory threshold monitoring over time. Do not use it if units can manipulate their running-variable score, if the threshold changes arbitrarily across periods making a common cutoff undefined, if only a single period is available (use standard RDD instead), or if the panel is very short and the number of observations near the cutoff is small.
Strengths & limitations
- Combines two sources of identification — local randomisation at the cutoff and within-unit variation — making causal estimates more robust than cross-sectional RDD alone.
- Unit fixed effects absorb time-invariant unobserved heterogeneity, reducing omitted-variable bias beyond what a single-period RDD achieves.
- The threshold assignment rule is transparent and externally verifiable, supporting credible communication of causal claims.
- Panel structure can justify tighter bandwidths because repeated observations per unit increase effective sample size near the cutoff.
- Time fixed effects additionally control for aggregate shocks common to all units, further limiting confounding.
- The causal estimate is local — it applies only to units near the threshold, not to all treated units, limiting external validity.
- Manipulation of the running variable invalidates the design; density tests must be passed before results can be trusted.
- Requires sufficient observations within the chosen bandwidth across time; very sparse panels near the cutoff yield imprecise estimates.
- If the threshold or assignment rule changes across periods, a unified panel RDD estimate may conflate different mechanisms.
- Choosing functional form for f(x_it) on each side of the cutoff is consequential; polynomial overfitting can distort results.
Frequently asked
How does panel RDD differ from standard RDD?
Standard RDD uses a single cross-section (or pooled cross-sections) near the threshold. Panel RDD adds unit fixed effects by following the same units over multiple periods, removing time-invariant unobserved confounders and increasing precision, at the cost of requiring longitudinal data.
Can I pool all periods into one regression with fixed effects?
Yes — the canonical approach is to pool all periods and include unit and time fixed effects alongside the running variable function and treatment indicator. If the threshold or assignment rule varies by period, interact the treatment effect with time dummies to allow effect heterogeneity.
What bandwidth should I use?
Data-driven selectors such as the Calonico-Cattaneo-Titiunik (CCT) optimal bandwidth are recommended. In a panel setting the effective sample size near the cutoff is larger because each unit contributes multiple observations, which may allow a narrower bandwidth with adequate precision.
How do I test for manipulation of the running variable in a panel?
Apply McCrary's (2008) density discontinuity test period by period and on the pooled sample. If units can manipulate their score in some periods but not others, the design may be valid only for the non-manipulation periods.
What if the threshold changes across periods?
If the cutoff value changes over time, normalise the running variable in each period relative to that period's threshold (i.e., use x_it minus c_t). This re-centering allows pooling across periods with different nominal cutoffs while preserving the local-randomisation logic.
Sources
- Lee, D. S., & Lemieux, T. (2010). Regression Discontinuity Designs in Economics. Journal of Economic Literature, 48(2), 281-355. DOI: 10.1257/jel.48.2.281 ↗
- Hahn, J., Todd, P., & Van der Klaauw, W. (2001). Identification and Estimation of Treatment Effects with a Regression-Discontinuity Design. Econometrica, 69(1), 201-209. DOI: 10.1111/1468-0262.00183 ↗
How to cite this page
ScholarGate. (2026, June 3). Panel Data Regression Discontinuity Design. ScholarGate. https://scholargate.app/en/causal-inference/panel-data-regression-discontinuity-design
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Difference-in-DifferencesEconometrics↔ compare
- Fuzzy Regression DiscontinuityCausal inference↔ compare
- Panel Data Difference-in-DifferencesCausal inference↔ compare
- Panel Data Interrupted Time SeriesCausal inference↔ compare
- Panel Fixed EffectsEconometrics↔ compare