A1268
Title: Robust estimation of volatility models in the presence of additive outliers
Authors: Luiz Koodi Hotta - University of Campinas (Brazil) [presenting]
Jean Sabino Diniz - University of Campinas (Brazil)
Eduardo Gabriel Pinheiro - University of Campinas (Brazil)
Carlos Trucios - University of Campinas (Brazil)
Abstract: Estimation and prediction of volatility in univariate and multivariate financial time series are of crucial importance. One of the features of data from finance is the variety of types of series and applications, leading to several models proposed in the literature. Another important feature is the presence of outliers, especially additive outliers. We discuss the estimation of the volatility using several models, from the simple univariate GARCH model to the high dimensional cDCC model. For each of the entertained models, we first present the effects of the outliers on the estimates from traditional non-robust methods. Then, we propose a robust estimator and compare the performance of the traditional and robust methods. The comparison considers different frequencies and sizes of the outliers.
