Spatial Counterfactual Impact Evaluation (SCIE)
Spatial Counterfactual Impact Evaluation · Also known as: SCIE, spatial CIE, place-based counterfactual evaluation, regional counterfactual analysis
Spatial Counterfactual Impact Evaluation (SCIE) is a family of quasi-experimental methods that estimate the causal effect of geographically targeted policies — such as EU Cohesion Funds, enterprise zones, or place-based subsidies — by constructing a spatial counterfactual: what outcomes the treated region would have experienced without the intervention, inferred from comparable untreated regions or from discontinuities at policy boundaries.
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 SCIE when a policy or intervention is assigned to geographic units (regions, municipalities, zones) and you need a causal estimate of its impact on economic, social, or environmental outcomes. It is well suited to EU structural fund evaluations, enterprise zone studies, and boundary-based natural experiments. The method requires credible geographic variation in treatment assignment — a meaningful boundary, discontinuity, or comparable control areas — and spatial outcome data for both treated and untreated units. Do not use it when treatment is purely random (an RCT is stronger), when there are no comparable spatial controls, when the policy area is so large that no untreated neighbours remain, or when outcomes are not geographically referenced.
Strengths & limitations
- Identifies causal effects of place-based policies from observational spatial data by exploiting geographic assignment rules.
- Handles the fact that regions self-select into policy programmes by matching on spatial pre-treatment characteristics or exploiting sharp eligibility thresholds.
- Can explicitly model and correct for spatial spillovers and general-equilibrium displacement effects that standard DiD ignores.
- Flexible framework: compatible with RD, DiD, synthetic control, and matching depending on the assignment mechanism.
- Particularly well suited to evaluating EU Cohesion Policy and other large-scale place-based interventions where experimental data are unavailable.
- Identifying assumptions (parallel spatial trends, no sorting around boundaries) are untestable in strict terms and may fail near policy thresholds where firms or households relocate in anticipation.
- Spatial spillovers violate SUTVA; if untreated control areas benefit or lose from the policy indirectly, the counterfactual is contaminated.
- Data demands are high: long pre-treatment panels at fine geographic resolution are needed to verify trends and conduct placebo tests.
- Boundary-based designs lose external validity — effects estimated at the threshold may not generalise to the interior of treated areas.
- Implementation complexity is substantial; no single off-the-shelf estimator exists, and choices of bandwidth, matching covariates, and spillover buffers all affect results.
Frequently asked
How is this different from a standard difference-in-differences?
Standard DiD compares a treated and a control group before and after an intervention but assumes no spatial spillovers and ignores the geographic structure of treatment. SCIE explicitly exploits geographic assignment rules (boundaries, thresholds), accounts for spatial autocorrelation in errors, and models or corrects for spillovers across the treatment boundary — making it appropriate when units are spatially connected regions rather than independent observations.
What is the SUTVA problem in spatial settings?
SUTVA (Stable Unit Treatment Value Assumption) requires that one unit's treatment does not affect another unit's outcome. In spatial settings this is routinely violated: a subsidy to one region may attract investment away from a neighbouring untreated region (displacement), or generate supply-chain spillovers. SCIE addresses this by using buffer zones, donut designs, or spatial-lag models to detect and correct for such contamination.
When is a regression discontinuity approach preferred over matching?
Spatial RD is preferred when there is a sharp, administratively determined eligibility threshold (e.g., GDP per capita below a fixed EU benchmark) and you can observe units just on both sides of it. Matching is preferred when no such threshold exists but you can find a pool of untreated areas with similar pre-treatment characteristics. RD gives stronger local identification; matching offers broader coverage but requires stronger covariate balance assumptions.
What data resolution do I need?
The finest feasible geographic resolution consistent with the policy assignment unit is ideal. EU cohesion studies typically use NUTS2 or NUTS3 regions; firm-level studies use municipality or postcode data. Coarser aggregation can mask heterogeneous effects and make boundary designs infeasible. A pre-treatment panel of at least three to five years is recommended for trend checks.
How do I handle spatial autocorrelation in inference?
Cluster standard errors at the regional or NUTS level to account for within-cluster correlation. When clusters are large or few, spatial HAC (Conley) standard errors that down-weight observations by geographic distance are a robust alternative. Always report both and note if conclusions change.
Sources
- Cerqua, A., & Pellegrini, G. (2014). Do subsidies to private capital boost firms' growth? A multiple regression discontinuity design approach. Journal of Public Economics, 109, 114-126. DOI: 10.1016/j.jpubeco.2013.11.005 ↗
- Pellegrini, G., Terribile, F., Tarola, O., Muccigrosso, T., & Busillo, F. (2013). Measuring the effects of European Regional Policy on economic growth: A regression discontinuity approach. Papers in Regional Science, 92(1), 217-233. DOI: 10.1111/j.1435-5957.2012.00459.x ↗
How to cite this page
ScholarGate. (2026, June 3). Spatial Counterfactual Impact Evaluation. ScholarGate. https://scholargate.app/en/causal-inference/spatial-counterfactual-impact-evaluation
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
- Instrumental Variables in Health ResearchHealth Economics↔ compare
- Propensity Score MatchingResearch Statistics↔ compare
- Synthetic Control MethodCausal inference↔ compare