Synthetic control is used when a policy affects one company, region or other unit but no single untreated unit offers a convincing comparison. Researchers assign weights to several unaffected units, producing a combined comparison whose history resembles the affected unit before the policy started.
After the change, the difference between the affected unit and its synthetic counterpart is treated as an estimate of the policy's effect. For a return-to-office mandate, researchers might build a comparison using firms with similar pre-policy patterns in turnover, hiring or employee reviews.
The method depends on having a useful pool of untreated units and a good fit before treatment. It may fail if the policy also affects comparison firms, if outcomes are too noisy or if an unrelated shock hits the treated unit at the same time.
Check the pre-policy fit, which units received weight, whether weights were constrained, and whether placebo analyses produce similarly large gaps elsewhere. A synthetic control creates a defensible counterfactual only when the data and assumptions support it.