Randomization Tests for Observational Studies with Staggered Treatment Adoption

Azeem M. Shaikh from the University of Chicago proposes a methodology for inference in observational studies with time-varying treatment adoption. The analysis assumes a Cox proportional hazards model for treatment timing and studies randomization tests for a null hypothesis of no treatment effect. The methodology includes a simulation study and an empirical application using synthetic control-based test statistics and tobacco legislation data.

Use Cases

• Testing the null hypothesis of no treatment effect based on the described randomization test framework.
• Simulating the performance of inference methods under staggered treatment adoption scenarios.
• Applying synthetic control-based test statistics to real-world policy data, as demonstrated with tobacco legislation.
• Assessing robustness of causal inference methods to misspecification of the treatment adoption model.

Strengths

• Methodology is grounded in a formal Cox proportional hazards model for treatment timing.
• Includes both theoretical finite-sample validity proofs and a practical simulation study.
• Empirical application uses a known dataset from Abadie, Diamond and Hainmueller.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and file formats are unknown, which may limit suitability assessment.

Provenance

Source

University of Chicago

Collection Method

Methodological research paper with simulation and empirical application.