Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
134 tokens/sec
GPT-4o
10 tokens/sec
Gemini 2.5 Pro Pro
47 tokens/sec
o3 Pro
4 tokens/sec
GPT-4.1 Pro
38 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

Semiparametric Bayesian Difference-in-Differences (2412.04605v3)

Published 5 Dec 2024 in econ.EM and stat.ML

Abstract: This paper studies semiparametric Bayesian inference for the average treatment effect on the treated (ATT) within the difference-in-differences (DiD) research design. We propose two new Bayesian methods with frequentist validity. The first one places a standard Gaussian process prior on the conditional mean function of the control group. The second method is a double robust Bayesian procedure that adjusts the prior distribution of the conditional mean function and subsequently corrects the posterior distribution of the resulting ATT. We prove new semiparametric Bernstein-von Mises (BvM) theorems for both proposals. Monte Carlo simulations and an empirical application demonstrate that the proposed Bayesian DiD methods exhibit strong finite-sample performance compared to existing frequentist methods. We also present extensions of the canonical DiD approach, incorporating both the staggered design and the repeated cross-sectional design.

Summary

  • The paper introduces a semiparametric Bayesian estimator for ATT using a Gaussian process prior and establishes a Bernstein-von Mises theorem for frequentist equivalence.
  • The paper develops a double robust Bayesian procedure that integrates propensity score adjustments to correct bias under limited overlap conditions.
  • The paper validates these methods through Monte Carlo simulations and an empirical reanalysis of a minimum wage study to demonstrate practical robustness.

Overview of Semiparametric Bayesian Difference-in-Differences

The paper "Semiparametric Bayesian Difference-in-Differences" by Christoph Breunig, Ruixuan Liu, and Zhengfei Yu presents innovative methodologies for semiparametric Bayesian inference within the difference-in-differences (DiD) framework for estimating average treatment effects on the treated (ATT). This work introduces two novel Bayesian approaches with frequentist validity: one utilizing standard Gaussian process priors, and another double robust approach featuring both prior and posterior corrections.

Methodological Contributions

  1. Standard Gaussian Process Priors:
    • The first proposed method employs a standard Gaussian process prior over the conditional mean function for the control group, yielding a semiparametric Bayesian estimator for ATT. Notably, the authors demonstrate the asymptotic equivalence of this Bayesian estimator with an efficient frequentist estimator by establishing a semiparametric Bernstein-von Mises (BvM) theorem. This result provides a Bayesian analogue to the outcome regression approach commonly used in frequentist settings, facilitating uncertainty quantification without estimating propensity scores, thus showing robustness to overlap violations.
  2. Double Robust Bayesian Procedure:
    • The second method extends the robustness of Bayesian inference by integrating propensity score based adjustments. This approach entails modifying the prior with an influence function based on propensity scores, alongside posterior corrections to account for bias under double robust smoothness conditions. The authors establish a semiparametric BvM theorem in this setting, warranting reliable Bayesian inference even when the overlap assumption nearly fails.

Numerical Results and Practical Implications

Monte Carlo simulations validate the strong finite sample performance of these Bayesian methods, comparing favorably against existing frequentist DiD techniques such as doubly robust estimators and machine learning-based approaches. Specifically, while the standard Gaussian process prior method yields efficient inference when dimensional complexity is moderate and overlap holds, the double robust method excels under propensity score misspecification or highly complex models.

Additionally, the paper illustrates the applicability of these Bayesian methods through an empirical case paper revisiting Card and Krueger's analysis of minimum wage impacts on employment, confirming the robustness of estimates obtained via Bayesian DiD methods.

Future Directions

This research opens several avenues for further exploration in the field of causal inference using Bayesian techniques. Future studies could expand these methodologies to handle even more intricate datasets, such as those with time-varying treatments or dynamic panel settings. Moreover, the integration of these techniques with advanced machine learning models could enhance predictive capabilities and computational efficiency, offering a richer inferential framework for contemporary econometric analysis.

Overall, this work significantly enrichens the toolkit available for Bayesian causal inference, providing robust and flexible methods suited to a broad range of practical applications in policy evaluation and beyond.

X Twitter Logo Streamline Icon: https://streamlinehq.com