Synthetic Control Method (SCM)
Synthetic Control Method · Also known as: synthetic control method, SCM, synthetic counterfactual, Sentetik Kontrol Yöntemi (SCM)
The Synthetic Control Method, introduced by Abadie, Diamond and Hainmueller in 2010, builds a weighted counterfactual for a single treated unit from a pool of untreated donor units. It is widely regarded as the gold standard for evaluating large policy interventions, natural experiments, and N=1 case studies where no obvious comparison unit exists.
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.
+2 more
When to use it
Use SCM when a single unit is exposed to an intervention and you observe it alongside a pool of comparable untreated donor units across a panel or time series. It requires a sufficiently long pre-treatment window (at least about 20 periods) so the synthetic unit can be matched to the treated unit's history. Key assumptions are: a long enough pre-treatment period for good fit, the presence of similar units in the donor pool, no interference (donor units must be unaffected by the treatment), and weights estimated under the convex-combination constraint. It is less suitable when the donor pool is too small or the pre-treatment fit is poor.
Strengths & limitations
- Provides a transparent, data-driven counterfactual for N=1 situations where no single comparison unit is available.
- Considered the gold standard for evaluating large policy interventions, natural experiments, and comparative case studies.
- The convexity constraint guards against extrapolation: the synthetic unit stays within the range of the observed donor units.
- Needs a long pre-treatment period; with fewer than about 20 periods the weighted combination cannot reproduce the treated unit's pre-period history.
- A poor pre-treatment fit (high RMSPE) makes the synthetic counterfactual unreliable.
- Requires a donor pool of genuinely comparable units that are not themselves affected by the treatment (no interference).
Frequently asked
How is the synthetic control different from a simple control group?
Instead of picking one comparison unit, SCM constructs a weighted blend of several untreated donor units so that the combination matches the treated unit's pre-treatment characteristics and outcome path. This data-driven counterfactual is usually closer to the treated unit than any single comparison unit would be.
What does the pre-treatment fit tell me?
It is the core diagnostic. A close pre-treatment fit (low root-mean-square prediction error, RMSPE) means the synthetic unit reproduces the treated unit's history well, so the post-treatment gap can be read as a treatment effect. A poor fit signals that the counterfactual is unreliable.
How many periods and donor units do I need?
You need a reasonably long pre-treatment window (at least about 20 periods) and a pool of genuinely comparable untreated units. Too few periods or too small a donor pool means the convex combination cannot match the treated unit, and the method should give way to an interrupted time series design.
What should I do if the donor pool is too small?
When the donor pool cannot reproduce the treated unit's pre-treatment trajectory, switch to an interrupted time series analysis, which estimates the intervention effect from the treated unit's own before-and-after series rather than from a weighted comparison.
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 1). Synthetic Control Method. ScholarGate. https://scholargate.app/en/causal-inference/synthetic-control
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.
- Interrupted Time SeriesCausal inference↔ compare
- Matching MethodsCausal inference↔ compare
- Panel Fixed EffectsEconometrics↔ compare
- Regression DiscontinuityCausal inference↔ compare
- Two-Stage Least Squares (2SLS)Causal inference↔ compare