Depth-Weighted Estimation of Heterogeneous Agent Panel Data Models∗ Yoonseok Lee† Donggyu Sul‡ Syracuse University University of Texas at Dallas June 2020 Abstract We develop robust estimation of panel data models, which is robust to various types of outlying behavior of potentially heterogeneous agents. We estimate parameters from individual-specific time-series and average them using data-dependent weights. In partic- ular, we use the notion of data depth to obtain order statistics among the heterogeneous parameter estimates, and develop the depth-weighted mean-group estimator in the form of an L-estimator. We study the asymptotic properties of the new estimator for both homogeneous and heterogeneous panel cases, focusing on the Mahalanobis and the pro- jection depths. We examine relative purchasing power parity using this estimator and cannot find empirical evidence for it. Keywords: Panel data, Depth, Robust estimator, Heterogeneous agents, Mean group estimator. JEL Classifications: C23, C33. ∗First draft: September 2017. The authors thank to Robert Serfling and participants at numerous sem- inar/conference presentations for very helpful comments. Lee acknowledges support from Appleby-Mosher Research Fund, Maxwell School, Syracuse University. †Corresponding author. Address: Department of Economics and Center for Policy Research, Syracuse University, 426 Eggers Hall, Syracuse, NY 13244. E-mail:
[email protected] ‡Address: Department of Economics, University of Texas at Dallas, 800 W. Campbell Road, Richardson, TX 75080. E-mail:
[email protected] 1Introduction A robust estimator is a statistic that is less influenced by outliers. Many robust estimators are available for regression models, where the robustness is toward outliers in the regression error.