Authors & affiliations

Cristhian Jaramillo-Huaman and Javiera Cartagena-Farias, Care Policy and Evaluation Centre (CPEC), London School of Economics and Politial Science.

Introduction

Synthetic Control Methods are a quasi-experimental method used to estimate the impact of a policy or intervention (both terms are used interchangeably in this text). The method compares the outcomes of the unit exposed to the intervention with those of a group of units that were not affected by it, also known as the donor pool. The main goal is to estimate a counterfactual – i.e., what would have happened to the treated unit if the intervention had not taken place (Abadie, 2021). These methods are particularly useful to evaluate policy interventions implemented at a national level (e.g., one country adopting a specific type of care model).

Rather than selecting just one comparison unit, synthetic control methods create a weighted combination of units from the donor pool. This weighted combination forms a “synthetic” version of the treated unit that closely matches its pre-intervention characteristics and trends. By comparing the actual post-intervention outcomes of the treated unit to the outcomes of this synthetic control, we can estimate the causal effect of the intervention. The difference between the two outcomes reflects the estimated impact of the policy or intervention and helps assess how effective it was.

One of the key strengths of synthetic control approaches is its transparency: researchers can see exactly which donor units contribute to the synthetic control and how much weight each one receives. This makes the method particularly useful for cases where randomised experiments are not feasible, and where a credible comparison group needs to be built from observational data (Abadie et al., 2011). In addition, there is no need for the ‘parallel trends assumption’ from the Difference-in-Differences approach to hold (Clarke et al., 2023), providing researchers with more flexibility for analysis. Unsurprisingly, Athey and Imbens (2017) described Synthetic Controls as “arguably the most important innovation in the policy evaluation literature in the last 15 years”. Despite its potential, this method has been underused, especially compared to other quasi-experimental designs (Bouttell et al., 2018).

Description

A hypothetical example

We use a hypothetical example to illustrate how the Synthetic Control Method works. Figure 1 presents a synthetic control analysis estimating the potential impact of the 2014 Care Act on life expectancy among individuals aged 65 and over in England. The solid line represents the hypothetically observed life expectancy in England, while the dashed line shows the estimated outcome for a fully hypothetical synthetic control group. This synthetic control could potentially be constructed using a weighted combination of data from other European countries where no similar legislative framework was implemented. To ensure a good pre-intervention match, the synthetic control group should be calibrated using variables such as the proportion of the population aged 65 and over, the proportion receiving care-related benefits, and median gross earnings, among others. In this hypothetical scenario, the synthetic control would closely mirror England’s life expectancy prior to 2014. However, following the implementation of the Care Act, life expectancy in England would appear to diverge upwards relative to the synthetic control, which, if this were a real analysis, would have suggested a possible positive effect of the policy. It is important to note that this entire illustration is based on hypothetical data and should not be interpreted as empirical evidence of the Care Act’s effectiveness.

Figure 1. Synthetic control estimation: hypothetical impact of the 2014 Care Act on life expectancy

Data and Contextual Requirements to use Synthetic Controls

A valid comparison group

It is essential to have a valid comparison group — that is, units that have not been exposed to the intervention at any point. If another unit has adopted a similar intervention, it should be excluded from the donor pool because its outcomes would be influenced in ways that undermine the comparison (Abadie, 2021).

Anticipation

Another important assumption is that units should not change their behaviour before the intervention takes place. If the analysed unit starts reacting in anticipation of the intervention, this can bias the results. In such cases, it may be necessary to backdate the intervention period and thus account for early effects. This assumption of no anticipation goes hand in hand with the assumption of no interference. In other words, the intervention should only affect the treated unit and not spill over to other units in the donor pool. If there is evidence of spillover, the affected units should be removed to preserve the validity of the comparison (Abadie, 2021).

Convex hull condition

The convex hull condition is another requirement.  It means that before the intervention, the treated unit’s outcomes should be possible to recreate as a combination of the donor unit’s outcomes. For example, if the treated unit had pre-intervention values that were much higher or lower than any donor unit, no combination of donor units would match it well, so the method would not work properly (Abadie, 2021).

Enough data

Finally, one of the most important practical conditions is the availability of data. Synthetic control requires enough pre-intervention data to build a reliable synthetic match, and enough post-intervention data to assess the effect over time. A longer time horizon strengthens the credibility of the findings (Abadie, 2021).

Data requirements

Because of the way synthetic control works, this method requires panel data with a long enough time horizon to provide reliable estimates, as it is important to have sufficient information for both the pre-intervention and post-intervention period. The pre-intervention data are crucial for constructing a synthetic control that closely mirrors the treated unit’s characteristics and trends before the intervention. Post-intervention data allows us to observe and measure the impact of the intervention over time. One challenge researchers may face is deciding how long the post-intervention period should be. This depends on how quickly the effect of the intervention is expected to materialise. If the policy’s impact builds gradually, a longer follow-up period will be needed to capture its full effect.

Example (in long-term care)

Synthetic control is applied across disciplines including political science, economics, social policy, and public health. In the field of long-term care, notable applications of this method include:

• Seamer et al. (2023) showed how emergency admission rates reduced after the introduction of an integrated care programme in England.
• Frochen, Rodnyansky, and Ailshire (2024) compared the pre and post effects of the ordinance on the number of large residential care facilities developed in Los Angeles.
• Xiang et al. (2025) demonstrated that the implementation of community healthcare integration policies decreased the fiscal expenditures in several cities in China.
• Xinliang et al. (2021) assessed how long-term care insurance in China boost women’s employment, income, and working hours by reducing their elderly care burden.

Beyond the traditional synthetic control approach we can find:

• Generalised Synthetic Control (GSC): Extends synthetic control to settings with multiple treated units and time-varying unobserved confounders (Xu, 2017).
• Augmented Synthetic Control (ASC): Combines synthetic control with outcome regression to improve fit and efficiency (Ben-Michael et al., 2021).
• Bayesian Synthetic Control: Introduces a probabilistic framework to account for uncertainty in weights and outcomes (Kim et al., 2020).
• Synthetic Difference-in-Differences (DiD): Combines DiD with a synthetic group approach (Arkhangelsky et al., 2021).
• Individual synthetic control: Extends the synthetic control method by allowing the estimation of individuals and average treatment effects when multiple units receive treatment at different times (Vagni & Breen, 2021).
• Individual synthetic control : Used by Petrillo et al. (2025) to analyse the casual estimates of the income penalty faced by informal carers.

References

Abadie, A. (2021). Using Synthetic Controls: Feasibility, Data Requirements, and Methodological Aspects. Journal of Economic Literature, 59(2), 391–425. https://doi.org/10.1257/jel.20191450

Abadie, A., Diamond, A., & Hainmueller, J. (2011). Synth : An R Package for Synthetic Control Methods in Comparative Case Studies. Journal of Statistical Software, 42(13). https://doi.org/10.18637/jss.v042.i13

Arkhangelsky, D., Athey, S., Hirshberg, D. A., Imbens, G. W., & Wager, S. (2021). Synthetic Difference-in-Differences. American Economic Review, 111(12), 4088–4118. https://doi.org/10.1257/aer.20190159

Athey, S., & Imbens, G. W. (2017). The State of Applied Econometrics: Causality and Policy Evaluation. Journal of Economic Perspectives, 31(2), 3–32. https://doi.org/10.1257/jep.31.2.3

Ben-Michael, E., Feller, A., & Rothstein, J. (2021). The Augmented Synthetic Control Method. Journal of the American Statistical Association, 116(536), 1789–1803. https://doi.org/10.1080/01621459.2021.1929245

Bouttell, J., Craig, P., Lewsey, J., Robinson, M., & Popham, F. (2018). Synthetic control methodology as a tool for evaluating population-level health interventions. Journal of Epidemiology and Community Health, 72(8), 673–678. https://doi.org/10.1136/jech-2017-210106

Clarke, G. M., Steventon, A., & O’Neill, S. (2023). A comparison of synthetic control approaches for the evaluation of policy interventions using observational data: Evaluating the impact of redesigning urgent and emergency care in Northumberland. Health Services Research, 58(2), 445–457. https://doi.org/10.1111/1475-6773.14126

Frochen, S., Rodnyansky, S., & Ailshire, J. (2024). The Eldercare Facility Ordinance of Los Angeles: A Synthetic Control Analysis of Residential Care Development and Growth. Journal of Planning Education and Research, 44(3), 1529–1541. https://doi.org/10.1177/0739456X221091432

Kim, S., Lee, C., & Gupta, S. (2020). Bayesian Synthetic Control Methods. Journal of Marketing Research, 57(5), 831–852. https://doi.org/10.1177/0022243720936230

Petrillo, M., Valdenegro, D., Rahal, C., Zhang, Y., Pryce, G., & Bennett, M. (2025). Estimating the Cost of Informal Care with a Novel Two-Stage Approach to Individual Synthetic Control. https://doi.org/10.2139/ssrn.5255642

Seamer, P., Lloyd, T., Conti, S., & O’Neill, S. (2023). The Long-Term Impacts of an Integrated Care Programme on Hospital Utilisation among Older Adults in the South of England: A Synthetic Control Study. International Journal of Integrated Care, 23(3). https://doi.org/10.5334/ijic.6475

Vagni, G., & Breen, R. (2021). Earnings and Income Penalties for Motherhood: Estimates for British Women Using the Individual Synthetic Control Method. European Sociological Review, 37(5), 834–848. https://doi.org/10.1093/esr/jcab014

Xiang, Y., Zhong, G., Lei, X., & Liu, L. (2025). Community health care integration and fiscal expenditures: evidence from a synthetic control approach. Frontiers in Public Health, 13. https://doi.org/10.3389/fpubh.2025.1524984

Xinliang, Y., Yaxuan, Z., Xiaohan, F., Qian, L., & Wenguang, Y. (2021). Long-term Care Insurance, Female Employment and Equal Labor Rights: Based on the Overlapping Generation Model and the Synthetic Control Method. Journal of Finance and Economics, 47(10), 95–109. https://doi.org/10.16538/j.cnki.jfe.20210813.302

Xu, Y. (2017). Generalized Synthetic Control Method: Causal Inference with Interactive Fixed Effects Models. Political Analysis, 25(1), 57–76. https://doi.org/10.1017/pan.2016.2

Suggested Citation

Jaramillo-Huaman, C. and Cartagena-Farias, J. (2025) Synthetic Control Methods. GOLTC Methods Guide series, 5. Global Observatory of Long-Term Care, Care Policy and Evaluation Centre, London School of Economics and Political Science. https://goltc.org/publications/synthetic-control-methods/