Panel Data Synthetic Control Method
Synthetic Control Method for Panel Data · Also known as: SCM panel, panel synthetic control, synthetic control estimator, comparative case study
The panel data synthetic control method estimates the causal effect of an intervention on a single treated unit by constructing a data-driven weighted combination of untreated units — a synthetic control — that best reproduces the treated unit's pre-treatment outcome trajectory. The post-treatment gap between the treated unit and its synthetic counterpart is the estimated treatment effect.
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.
+1 more
When to use it
Use the panel data synthetic control method when: (1) a single aggregate unit (country, region, firm) receives a treatment while several comparable untreated units serve as donors; (2) a long panel of pre-treatment periods is available to construct and validate the synthetic match; (3) you need a transparent, data-driven counterfactual rather than an ad hoc comparison group. Do not use it when there are too few pre-treatment periods to achieve a good fit, when the treated unit is an outlier that cannot be approximated by any convex combination of donors, when there are many treated units (difference-in-differences is better suited), or when extrapolation outside the support of the donor pool is required.
Strengths & limitations
- Produces an explicit, transparent counterfactual whose pre-treatment fit can be inspected and reported, making the causal assumption visibly assessable.
- Does not require the parallel-trends assumption; instead it requires that the pre-treatment fit is close, which is verifiable in the data.
- Exploits the full panel time series — long pre-treatment histories improve fit and credibility of the synthetic match.
- Permutation-based inference avoids distributional assumptions and is valid even with a small number of units.
- Weights are non-negative and sum to one, preventing extrapolation and providing an interpretable convex combination of real units.
- Designed for a single treated unit; extensions to multiple treated units are less straightforward than standard DiD.
- A sufficiently long pre-treatment panel is necessary; short pre-treatment periods undermine the fit and validity of the synthetic control.
- The donor pool must contain units that can actually approximate the treated unit; if the treated unit is an outlier, the method has no valid comparison.
- Permutation inference has limited power when the donor pool is small (fewer than about 20 units), because the placebo distribution is coarse.
- Sensitive to predictor selection and the balance matrix V; results can vary across reasonable specification choices.
Frequently asked
How many pre-treatment periods do I need?
There is no strict minimum, but a longer pre-treatment panel substantially improves the quality and credibility of the synthetic match. In practice, researchers typically aim for at least 10 to 20 pre-treatment periods; fewer periods risk overfitting and a misleading appearance of good fit.
How is inference done without standard p-values?
Inference is permutation-based. You iteratively apply the synthetic control method to each donor unit as if it were the treated unit, producing a distribution of placebo gaps. The fraction of placebos with a post-treatment gap at least as large as the treated unit's gap serves as the p-value analog.
What if my treated unit cannot be well-approximated by the donor pool?
Poor pre-treatment fit is a signal that the donor pool cannot construct a valid counterfactual. You should expand the donor pool if possible, or acknowledge that the synthetic control approach is not feasible for this case and consider alternative methods such as difference-in-differences.
Can I use this method with multiple treated units?
The standard framework is designed for a single treated unit. Extensions such as the generalized synthetic control (Xu 2017) or the matrix completion method handle multiple treated units, but they introduce additional assumptions and estimation complexity.
How does this differ from standard difference-in-differences?
DiD averages outcomes across all control units with equal (or regression-based) weights and relies on the parallel-trends assumption. The synthetic control explicitly constructs optimal, data-driven weights to reproduce the treated unit's pre-treatment trajectory, making the comparison transparent and not requiring parallel trends.
Sources
- Abadie, A., Diamond, A., & Hainmueller, J. (2010). Synthetic Control Methods for Comparative Case Studies: Estimating the Effect of California's Tobacco Control Program. Journal of the American Statistical Association, 105(490), 493-505. DOI: 10.1198/jasa.2009.ap08746 ↗
- Abadie, A. (2021). Using Synthetic Controls: Feasibility, Data Requirements, and Methodological Aspects. Journal of Economic Literature, 59(2), 391-425. DOI: 10.1257/jel.20191450 ↗
How to cite this page
ScholarGate. (2026, June 3). Synthetic Control Method for Panel Data. ScholarGate. https://scholargate.app/en/causal-inference/panel-data-synthetic-control-method
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
- Matching EstimatorCausal inference↔ compare
- Panel Data Difference-in-DifferencesCausal inference↔ compare
- Panel Fixed EffectsEconometrics↔ compare
- Synthetic Control MethodCausal inference↔ compare