The authors thank associate editor Anders Rahbek, two anonymous referees, Javier Gómez-Biscarri, Uwe Hassler, Enrique Sentana, and seminar participants at the Humboldt University of Berlin, and conference participants of the 2006 Annual ESRC Conference (Bristol), the 2007 EFMA Conference (Vienna), the 31st SAE Conference (Oviedo), XIV Finance Forum (Castellón) and the 2nd Congress of the Italian Empirical Economics and Econometrics Society (Rimini) for comments and suggestions. Any remaining errors are the authors’ own responsibility. Financial support from POCTI/FEDER (grant ref. PTDC/ECO/64595/2006) and the Spanish Deparment of Education and Science (grant ref. ECO2008-02599/ECON) is gratefully acknowledged. A previous version of this article is available in the working paper series edited by Fundacion de Cajas de Ahorros (FUNCAS), WP-247.
The Effects of Additive Outliers and Measurement Errors when Testing for Structural Breaks in Variance*
Version of Record online: 16 FEB 2011
© Blackwell Publishing Ltd and the Department of Economics, University of Oxford, 2011
Oxford Bulletin of Economics and Statistics
Volume 73, Issue 4, pages 449–468, August 2011
How to Cite
Rodrigues, P. M. M. and Rubia, A. (2011), The Effects of Additive Outliers and Measurement Errors when Testing for Structural Breaks in Variance. Oxford Bulletin of Economics and Statistics, 73: 449–468. doi: 10.1111/j.1468-0084.2010.00621.x
- Issue online: 20 JUN 2011
- Version of Record online: 16 FEB 2011
- Final Manuscript Received: October 2010
This article discusses the asymptotic and finite-sample properties of the CUSUM tests for detecting structural breaks in volatility when the data are perturbed with (additive) outliers and/or measurement errors. The special focus is on the parametric and non-parametric tests in Inclán and Tiao (1994) and Kokoszka and Leipus (2000). Whereas the asymptotic distribution of the former can be largely affected, the distribution of the latter remains invariant and renders consistent break-point estimates. In small samples, however, large additive outliers are able to generate sizeable distortions in both tests, which explains some of the contradictory findings in previous literature.