Do you know your biases? A Monte Carlo analysis of dynamic panel data estimators

Vadim Kufenko, Klaus Prettner

Publication: Working/Discussion PaperWU Working Paper

Abstract

We assess the performance of widely-used dynamic panel data estimators based on Monte Carlo simulations of a dynamic economic process. Knowing the true underlying coefficient of the autoregressive term, we show that most estimators exhibit a severe bias even in the absence of measurement errors, omitted variables, and endogeneity issues. We analyze how the bias changes with the sample size, the autoregressive coefficient, and the estimation options. Based on our insights, we recommend i) carefully choosing appropriate estimators given the underlying structure of the data and ii) scrutinizing the estimation results based on the insights of simulation studies.
Original languageEnglish
Place of PublicationVienna
PublisherWU Vienna University of Economics and Business
Publication statusPublished - 1 Sep 2021

Publication series

NameDepartment of Economics Working Paper Series
No.316

WU Working Paper Series

  • Department of Economics Working Paper Series

Cite this