University of New South Wales

Testing for Structural Breaks in the Long-run Means of VARs

Email address: r.bewley@unsw.edu.au

Abstract:
Because of the nonlinear relationship between the constant terms in a VAR, on the one hand, and drift in the case of I(1) variables, or the means of the series in stationary VARs on the other, it is not possible to use standard tests for structural breaks directed specifically at individual long-run means. Given the importance of the estimated long-run means in forecasting performance (Bewley, 2000), the detection of structural breaks in these parameters is crucial in building VAR forecasting models. It is shown that direct tests of structural breaks in long-run means can easily be constructed for the Bewley (1979) transformation of a VAR or VEC and that these tests are more powerful than those directed only at the constant terms in the traditional VAR framework.

PDF file of paper: bewley_abstract.pdf

Session: Time Series Analysis

Time: Friday, 6 July, 8:45am - 10:15am

Room: B