Geographic Regression Discontinuity
Also known as: Spatial RD, Geographic RDD
Geographic Regression Discontinuity (GRD) is a quasi-experimental design that exploits sharp geographic boundaries—borders, policy boundaries, or natural features—to estimate causal effects. Introduced by Dell (2010) and others, it compares outcomes on either side of a boundary where treatment changes abruptly, leveraging the idea that units on opposite sides of a border are otherwise similar. This approach yields credible causal estimates for spatially localized policies, institutional changes, and natural phenomena.
Key highlights
- Provides credible causal estimates by leveraging exogenous boundary locations
- Transparent identification strategy easy to communicate to stakeholders
- Combines the credibility of randomized trials with observational data
- Allows targeted study of spatially localized interventions
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 GRD when a sharp geographic boundary creates a quasi-random assignment of treatment. Examples include country borders (different policies), city limits (tax jurisdictions), or natural boundaries (climate zones). It is particularly useful when randomization is infeasible but boundaries are well-defined. Assume that sorting across the boundary is minimal and that unobserved characteristics do not jump discontinuously.
Strengths & limitations
- Provides credible causal estimates by leveraging exogenous boundary locations
- Transparent identification strategy easy to communicate to stakeholders
- Combines the credibility of randomized trials with observational data
- Allows targeted study of spatially localized interventions
- Requires precise geographic data and correct boundary identification
- Estimates are local to the boundary; generalization to other regions is difficult
- Sorting or migration around the boundary can violate the identification assumption
- Small sample sizes in the bandwidth region can reduce precision
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 do I choose the bandwidth (distance from the boundary)?
Use data-driven methods like the bandwidth selector of Imbens and Lemieux (2008) or conduct sensitivity analysis varying the bandwidth. A practical approach: start with a bandwidth that includes roughly 30–100 units on each side, then test robustness to bandwidth halving and doubling.
What if treatment is not perfectly sharp at the boundary?
If treatment is fuzzy (compliance less than perfect), use fuzzy RD methodology, which instruments treatment with the geographic boundary. This yields the local average treatment effect (LATE) for those complying with the boundary.
How do I test whether my boundary assumption is valid?
Check whether baseline covariates (pre-treatment characteristics) are continuous at the boundary. If they jump discontinuously, this suggests sorting or hidden confounders violate your assumption. Plot outcome versus distance from the boundary to visually inspect smoothness.
Can I use GRD with multiple boundaries simultaneously?
Yes. Multi-dimensional RD exploits multiple boundaries (e.g., latitude and longitude). This requires careful specification to avoid over-differencing and can improve precision, but interpretation becomes complex.
Sources
- 1.Dell, M. (2018). The persistent effects of Peru's mining mita. Econometrica, 78(6), 1863-1911.
- 2.Imbens, G. W., & Lemieux, T. (2008). Regression discontinuity designs: A guide to practice. Journal of Econometrics, 142(2), 615-635.
You have read it. What now?
Cite this page
ScholarGate. (2026, June 3). Geographic Regression Discontinuity. ScholarGate. https://scholargate.app/econometrics/geographic-regression-discontinuity