Comparing Two-way Fixed Effects and New Estimators for Difference-in-Differences: A Simulation Study and Empirical...

Area of Science:

• Epidemiology
• Health Policy Research
• Econometrics

Background:

• Two-way fixed effects (TWFE) are commonly used for policy evaluation with staggered adoption.
• However, TWFE can yield biased estimates with heterogeneous or dynamic treatment effects.
• Alternative methods have been proposed to address these limitations.

Purpose of the Study:

• To compare the performance of TWFE with alternative methods for estimating average treatment effect on the treated (ATT) under staggered policy adoption.
• To evaluate the impact of Medicaid expansion on preterm birth using these different methods.
• To assess estimator performance through simulations mimicking empirical data.

Main Methods:

• Compared TWFE, group-time ATT, cohort ATT, and target-trial estimators.
• Applied methods to analyze Medicaid expansion's effect on preterm birth using national birth records.
• Conducted simulations with constant, heterogeneous, and dynamic effects to evaluate bias, MSE, and CI coverage.

Main Results:

• TWFE estimated an increased risk of preterm birth (RD, 0.12).
• Alternative methods (group-time ATT, cohort ATT, target-trial) estimated protective or null effects (RDs ranging from -0.02 to -0.16).
• TWFE performed poorly with heterogeneous and dynamic effects in simulations, unlike alternative methods.

Conclusions:

• Newer methods are superior to TWFE for staggered policy adoption when effects are heterogeneous or dynamic.
• Provided simulation and analysis code to encourage adoption of these improved methods in epidemiology.