Direkt zum Inhalt

12. Juli 2017

SOEP Brown Bag Seminar

Dynamic binary panel data models in presence of autocorrelation in the error term. A Monte-Carlo study

Termin

12. Juli 2017
12:30 - 13:30

Ort

Eleanor-Dulles-Raum
DIW Berlin im Quartier 110
Room 5.2.010
Mohrenstraße 58
10117 Berlin

Sprecher*innen

Dr. Alexander Mosthaf (University of Mainz)

Dynamic panel data models are often applied to estimate the degree of state dependence. For instance, it has often been shown that there is state dependence in unemployment, i. e. unemployment in present causally increases the probability of unemployment in future. This paper uses Monte-
Carlo simulations to compare several dynamic linear and nonlinear panel data models for binary dependent variables. In particular, we focus on data generating processes where dynamics may stem from true state dependence or from autocorrelation in the error term (which may lead to the measurement of spurious state dependence). We employ estimators which explicitly model both sources of persistence and compare their performance with standard dynamic models. Furthermore, we test how these estimators perform in presence time-varying unobserved variables which follow time trends or which may be described by a moving-average model.

(joint with Thorsten Schank and Constantin Weiser)

Kontakt

keyboard_arrow_up