Spatial Event Study Design
Also known as: spatial event study, geographic event study, spatial dynamic DiD, place-based event study
Spatial event study design estimates the dynamic causal effects of a geographically concentrated shock or policy by plotting how outcomes in affected locations evolve relative to unaffected locations across time periods, while explicitly accounting for spatial spillovers and autocorrelation across geographic units. It is widely used in regional and urban economics to evaluate place-based policies, trade shocks, and local labour market interventions.
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 spatial event study design when you have panel data on geographic units, a clearly dated shock or policy that affects units with varying intensity based on geography or local industry mix, and you want to estimate dynamic (period-by-period) causal effects rather than a single average. It is particularly appropriate when spillovers across adjacent units are plausible — trade shocks, environmental regulations, infrastructure investment, or disease spread. Do not use it when geographic units are largely independent (spillovers are negligible and spatial autocorrelation is absent), when you have only a single pre- and post-period (use standard DiD instead), or when the shock timing is endogenous to local conditions, which undermines the identification strategy.
Strengths & limitations
- Provides a full dynamic picture of treatment effects, revealing whether impacts are immediate, delayed, or persistent rather than a single average effect.
- Explicitly accounts for spatial interdependence between geographic units, yielding valid standard errors and correcting for spillover-induced bias.
- The pre-trend plot offers a transparent visual test of the parallel-trends assumption, improving credibility and reproducibility.
- Accommodates continuous treatment intensity (e.g., local industry exposure shares) rather than requiring a binary treated/untreated classification.
- Well-suited to large administrative datasets where geographic identifiers are available and the cross-sectional dimension is rich.
- Requires constructing and justifying a spatial weights matrix, which involves researcher discretion and sensitivity to the chosen distance metric or contiguity definition.
- Identification relies on exogenous variation in geographic exposure; if affected units systematically differ from unaffected ones in pre-existing trends, estimates are biased.
- With many time periods and geographic units, the model can become computationally intensive, especially when estimating spatial error or spatial lag corrections.
- Spillover corrections add complexity and may be sensitive to the chosen bandwidth for Conley standard errors.
- Staggered treatment timing across geographic units requires additional estimators (e.g., heterogeneity-robust DiD) to avoid bias from two-way fixed effects in the multi-period case.
Frequently asked
How does spatial event study differ from a standard event study?
A standard event study treats units as independent and estimates a common treatment effect trajectory. The spatial variant explicitly models dependence across geographic units through a spatial weights matrix, adjusts standard errors for spatial autocorrelation, and often uses a continuous geographic exposure measure rather than a binary treatment indicator.
What spatial weights matrix should I use?
Common choices are binary contiguity (neighbours share a border), inverse distance, or economic distance (e.g., bilateral trade flows). There is no universally correct choice; the key is theoretical justification based on the channel through which spillovers operate, and sensitivity checks using alternative matrices.
Can I use this method with staggered treatment timing?
Yes, but standard two-way fixed effects event study estimators can be biased when treatment timing varies across units. Heterogeneity-robust estimators such as those proposed by Callaway and Sant'Anna (2021) or Sun and Abraham (2021) should be used in the staggered spatial setting.
What sample size is needed?
There is no fixed threshold, but reliable spatial autocorrelation tests and stable fixed-effects estimates generally require at least 50–100 geographic units and several pre- and post-event periods. Very small cross-sections make spatial corrections unstable.
How do I report results?
Report the full event-study coefficient plot (beta_k against time relative to event) with confidence intervals, the pre-trend test statistic, the spatial autocorrelation test (Moran's I on residuals), the chosen weights matrix and clustering approach, and robustness checks with alternative weight specifications.
Sources
- Autor, D. H., Dorn, D., & Hanson, G. H. (2013). The China Syndrome: Local Labor Market Effects of Import Competition in the United States. American Economic Review, 103(6), 2121-2168. DOI: 10.1257/aer.103.6.2121 ↗
- Kline, P. (2012). The Impact of Juvenile Curfew Laws on Arrests of Youth and Adults. American Law and Economics Review, 14(1), 44-67. link ↗
How to cite this page
ScholarGate. (2026, June 3). Spatial Event Study Design. ScholarGate. https://scholargate.app/en/causal-inference/spatial-event-study-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
- Dynamic Difference-in-DifferencesCausal inference↔ compare
- Panel Event StudyCausal inference↔ compare
- Spatial Difference-in-DifferencesCausal inference↔ compare
- Spatial Regression Discontinuity DesignCausal inference↔ compare