Affordable Access

How to Increase Accuracy of Volatility Forecasts Based on GARCH Models



Microsoft Word - 14_Fiszeder.docx © Co py rig ht b y T he N ico lau s C op er nic us U niv er sit y S cie nt ifi c P ub lis hin g H ou se DYNAMIC ECONOMETRIC MODELS Vol. 8 – Nicolaus Copernicus University – Toruń – 2008 Piotr Fiszeder Nicolaus Copernicus University in Toruń How to Increase Accuracy of Volatility Forecasts Based on GARCH Models 1. Introduction There is a large literature on volatility forecasting (see Poon and Granger, 2003), but nevertheless it is difficult to extract a coherent set of prescriptions concerning the most appropriate empirical procedure for tackling this issue. The results of empirical analyses are unclear and often even contradictory. Various conditional variance specifications within the parametric GARCH class of mod- els were proposed in the literature, but there is no consensus on the relative quality of out-of-sample forecasts of those formulations. Analyses with GARCH models for the Polish stock market were performed among others by Piontek (2003), Doman and Doman (2004), Fiszeder (2004a, 2004b, 2005), Osiewalski, Pajor and Pipień (2004) and Pipień (2006). Only in the investiga- tions of Doman and Doman and Fiszeder were intraday data used for evaluation of forecasts quality. In this paper a significantly wider class of GARCH models, especially models extended with additional information, was used and the re- search period was wider, which could significantly influence results. The main purpose of this study is to compare a performance of the different specifications of GARCH models for predicting volatility. Additional information used in construction of the GARCH model or in estimation of its parameters does not always lead to an increase in accuracy of volatility forecasts. The plan for the rest of the paper is as follows. Section 2 outlines the com- peting methods used in the analysis and the measures used to assess the per- formance of the candidate models. In section 3 accuracy of v

There are no comments yet on this publication. Be the first to share your thoughts.


Seen <100 times