erc/metu
INTERNATIONAL CONFERENCE IN ECONOMICS  IV
September 13-16, 2000, Ankara

A Study on the Finite Sample Bias of an Unstable VAR with a Drift

Michalis P. Stamatogiannis (University of York, UK )

Abstract

We study a specific case of the bias of the maximum likelihood estimator (MLE) in a finite sample VAR. Our analysis is based on the paper of Abadir, Hadri and Tzavalis (1999) who study the implications of varying the dimensions of VARs, on the biases of the MLE and least squares estimator (LSE). They give a formula which relates the multivariate bias to the univariate one. We investigate the evolution of the finite sample bias in a specific case, in order to see if the analysis of Abadir et al. is robust. The actual data generation process of the series is a simple unit root. The estimated VAR model contains a drift term. Using Monte Carlo simulations we find that the bias is higher in our case. The Abadir formula does not give the bias in this case. However, we find an equation which has a similar functional form.

Economic Research Center
Middle East Technical University
06531 Ankara Turkey
Phone: + 90 312 210 3044, 210 2003
Fax: +90 312 210 1244
e-mail: metuerc@metu.edu.tr