Spatial Causal Impact Analysis
Also known as: spatial causal inference, geo-causal analysis, spatial treatment effect estimation, spatial impact evaluation
Spatial causal impact analysis estimates the causal effect of a spatially-targeted intervention — a policy, shock, or treatment applied to particular locations — while explicitly accounting for geographic spillovers between treated and untreated units. By combining quasi-experimental designs such as difference-in-differences or regression discontinuity with spatial econometric models, it separates the direct local effect of a treatment from indirect effects that diffuse to neighbouring areas.
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 causal impact analysis when your intervention is geographically targeted (a zoning law, environmental regulation, infrastructure project, or local economic shock) and you suspect that treated locations affect nearby untreated locations. You need geo-referenced outcome and treatment data for multiple units (ideally 50 or more), and at least basic proximity information to build a weights matrix. It is especially valuable when standard DiD or RDD assumptions are undermined by spatial spillovers — for example, when firms near a policy boundary respond to the policy even though they fall just outside the treated zone. Do not use it when units are genuinely independent across space (very distant, no plausible transmission mechanism), or when your dataset lacks reliable location information.
Strengths & limitations
- Explicitly models geographic spillovers, preventing the under- or over-estimation of treatment effects that occurs when spatial interaction is ignored.
- Decomposes the total treatment effect into interpretable direct and indirect (spillover) components, which is essential for area-based policy evaluation.
- Compatible with standard quasi-experimental designs (DiD, RDD, synthetic control), extending their validity to spatially dependent data.
- Flexible weights matrix specification allows the analyst to encode domain-specific transmission mechanisms (contiguity, road distance, trade flows).
- Provides a more credible test of SUTVA violations and quantifies how far treatment effects propagate geographically.
- The choice of spatial weights matrix W is subjective; results can be sensitive to whether contiguity, distance-decay, or network weights are used.
- Requires geo-referenced data with at least moderate spatial coverage; sparse or unevenly distributed units weaken the spatial structure.
- Identification of causal effects is more demanding than in aspatial designs: spillovers can confound parallel-trends tests and complicate the definition of clean control groups.
- Computationally intensive for large spatial panels; maximum likelihood estimation of spatial lag models scales poorly with many units.
- Spatial models are harder to communicate to non-specialist audiences than standard regression or DiD tables.
Frequently asked
What is SUTVA and why does it matter here?
SUTVA — the Stable Unit Treatment Value Assumption — requires that a unit's outcome depends only on its own treatment, not on the treatment of others. Geography routinely violates this: policies, shocks, and interventions spill over to neighbours. Spatial causal impact analysis relaxes SUTVA by explicitly modelling these spillovers instead of assuming they away.
How do I choose the spatial weights matrix?
The weights matrix should reflect the plausible transmission mechanism: contiguity (shared border) for administrative spillovers, inverse distance or road-distance for economic or commuting spillovers, network adjacency for trade or supply-chain effects. Always run a sensitivity analysis over at least two or three plausible specifications and report whether results hold across them.
Can I combine spatial causal impact analysis with difference-in-differences?
Yes — spatial DiD is the most common combination. You include the standard DiD interaction term in a spatial model, allowing for spatially-lagged outcomes or covariates. The key is to verify the parallel-trends assumption using pre-treatment periods while correcting standard errors for spatial autocorrelation.
What is the difference between the direct and indirect effects?
The direct effect is the change in a treated unit's own outcome due to its own treatment. The indirect (spillover) effect is the change in a neighbouring unit's outcome due to the treated unit's treatment. In spatial lag models both are derived from the matrix inverse (I − ρW)⁻¹, meaning they depend on the entire network structure, not just immediate neighbours.
How large a sample do I need?
As a rough guide, at least 50 spatial units are needed to estimate the spatial lag parameter reliably. With fewer units the weights matrix becomes too sparse and the spatial parameters are imprecisely estimated. Panel depth (multiple time periods) partially compensates for a small number of locations.
Sources
- Delgado, M. S., & Florax, R. J. G. M. (2015). Difference-in-differences techniques for spatial data: Local autocorrelation and spatial interaction. Economics Letters, 137, 123-126. DOI: 10.1016/j.econlet.2015.10.035 ↗
- Halleck Vega, S., & Elhorst, J. P. (2015). The SLX Model. Journal of Regional Science, 55(3), 339-363. DOI: 10.1111/jors.12188 ↗
How to cite this page
ScholarGate. (2026, June 3). Spatial Causal Impact Analysis. ScholarGate. https://scholargate.app/en/causal-inference/spatial-causal-impact-analysis
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
- Geographically Weighted RegressionSpatial analysis↔ compare
- Propensity Score MatchingResearch Statistics↔ compare
- Synthetic Control MethodCausal inference↔ compare