“…Unlike a conventional two‐way fixed effects model, counterfactual estimators take observations under the treatment condition (in our context,
) as missing data and directly estimate their counterfactuals: the readmission rates for the treated hospitals if they have not experienced vertical integration. Essentially, counterfactual estimators convert the original causal identification problem into a “prediction” problem: Using the observations in untreated conditions (in our context,
) to predict the missing outcomes of treated observations in counterfactual conditions (Liu et al.,
2021; Pan & Qiu,
2022). This approach addresses unobserved, time‐varying confounders using a latent factor approach.…”