1.3.7 Advanced Methods, Econometrics for Health | Masters in Health Economics
Why It Matters
These techniques provide the only credible way to isolate policy effects in complex health settings, directly influencing decisions on taxation, regulation, and funding.
Key Takeaways
- •Fixed effects control time‑invariant unobserved factors within units
- •Random effects are efficient but biased if correlated with regressors
- •Hausman test decides between fixed and random effects models
- •Synthetic control builds a weighted counterfactual for single‑treated units
- •Placebo tests validate synthetic control results in space and time
Summary
The lecture introduces advanced econometric tools—panel‑data estimators and the synthetic control method (SCM)—as essential for rigorous health‑policy evaluation. It contrasts fixed‑effects (FE) models, which purge all time‑invariant unobserved characteristics by focusing on within‑unit changes, with random‑effects (RE) models that assume those characteristics are uncorrelated with regressors and thus offer tighter confidence intervals.
Key insights include the mechanics of FE and RE, the Hausman (or “horseman”) test that statistically adjudicates which estimator is appropriate, and the limitations of each approach (e.g., FE cannot identify variables that never vary). The SCM is presented as a solution for single‑treated units: construct a synthetic counterpart from a weighted donor pool that mirrors pre‑policy trends, then compare post‑policy outcomes. The California tobacco‑tax case illustrates how SCM revealed a sharper decline in smoking than the synthetic control, confirming policy effectiveness.
Notable examples reinforce the concepts: the instructor likens FE to a “time machine” that removes static bias, and describes building a “synthetic UK” to assess a novel oncology‑price cap. Placebo tests—both in‑space (applying SCM to untreated donors) and in‑time (shifting the intervention date)—serve as robustness checks, ensuring the observed effect is not an artifact.
For health economists, mastering these methods means moving beyond naïve averages to credible causal inference when randomized trials are infeasible. Selecting the right estimator, validating assumptions, and employing SCM where appropriate can shape evidence‑based policy, guide resource allocation, and ultimately improve population health outcomes.
Comments
Want to join the conversation?
Loading comments...