Estimación QMV de modelos GARCH vs estimación no paramétrica de la volatilidad local en sieres temporales financieras con outliers
- Fernández Arufe, Josefa E. (dir.)
- Rojo García, José Luis (dir.)
- Moyano Pesquera, Pedro Benito (coord.)
- Somarriba Arechavala, Noelia (coord.)
Publisher: Asociación Española de Economía Aplicada, ASEPELT
ISBN: 84-96477-93-2
Year of publication: 2007
Volume Title: Área VII : Métodos cuantitativos
Volume: 7
Pages: 585-609
Congress: ASEPELT España. Reunión anual (21. 2007. Valladolid)
Type: Conference paper
Abstract
In this paper, we study the robustness behavior of a class of non-parametric estimators of local volatility due to Lally, Randal and Thomson (2001, 2004) in the context of leptokurtic distributions. The nonparametric estimator is a local robust scale estimator based on finite moving-averages of the squared deviations of a time series from its local level, possibly time-variant. Our main object is to get new evidente about the non-robustness of the quasi-maximum likelihood estimator of a GARCH model, and to study the robustness of this non-parametric estimator throught the boundness of the contamination bias due to the contamination of the underlying volatility process by volatility outliers. Theoretical results are complemented with empirical results of the non-parametric estimation of the volatility process for several financial time series