Where did the GARCH Models Perform Best in Terms of Volatility Forecasting? Equity vs. Commodities Markets

Iulian Lolea

《Academic Journal of Economic Studies》 3卷3期 (2017/09) Pp. 79-86

摘要

This article aims to compare the performance of GARCH models in terms of volatility forecasting for two asset classes: equity and commodities. The idea behind this research was that GARCH models may perform differently depending on the asset class for which they are used. A comparison based on performance of GARCH, EGARCH and GJR-GARCH for the Romanian equity market, Polish equity market, gold market and Brent Crude Oil market has been done. The results were in line with initial expectations. Both, in-sample and out-of-sample analysis, highlighted the over performance of GARCH models for the equity market compared to the commodity market. Moreover, the performance of GARCH models in terms of volatility forecasting for the gold market decreases as the forecast horizon increases. Thus, it has been proved that there is a bias of classical GARCH models to perform better for equity markets, compared to commodity markets, but future researches can be directed towards adapting GARCH models to the commodity market. A possible solution would be to implement models that allow regime switching as the Markov Switching GARCH (MRS-GARCH) models do.

關鍵字

Volatility forecasting ； commodities ； equity markets ； statistical loss functions ； out-of-sample