Study Tackles Both Serial and Spatial Correlation in Panel Data Difference-in-Difference Models

Assistant Professor Karen Yan
About This Project

headshot of Assistant Professor Karen YanBy: Aselia Urmanbetova

In joint research with Yu Sun published in the prestigious Journal of Econometrics in 2019, Assistant Professor Karen Yan and her co-author address the issue of inference in the popular Difference-in-Differences model (DD).

The DD model is increasingly popular in policy analyses that can easily identify a discrete policy change within panel data sets. To conduct the DD analysis, groups affected by the policy change (i.e., treatment groups) and groups not affected by the policy change (i.e., control groups) are clearly identified, and the difference in their outcomes before and after the policy change is taken as a measure of the impact of the policy.

Yan’s research focuses on the issues relating to the standard error of the DD estimate and the distribution of the test statistics, which are used in conducting hypothesis testing — an essential procedure in the empirical analysis. Many DD analyses involve panel data with cross-sectional and serial correlations. Yan and her co-author develop a new DD standard error estimator that allows for arbitrary cross-sectional and serial correlations. They also apply a fixed-b approach (a technique for deriving the asymptotic distributions of test statistics) to obtain the distribution of the test statistic.

This breakthrough methodology is important as it enables empirical economists to test the significance of their DD model estimates much more accurately than in the previous methods.

Explore more research by faculty in the School of Economics on our featured project page.