GARCH models with changes in variance: An approximation to risk measurements

Abstract

This study aims to model volatility as an approximation to an optimum measurement of stock market risk because of the importance of this concept for, among other things, the proper management of portfolios. Following the proposal of Lamoureux and Lastrapes (1990), the authors consider that the high degree of persistence detected in GARCH models arises from a poor specification of the equation of the variance due to not considering the possible deterministic changes in the unconditional variance of the financial series. To determine the point in time as well as the duration of these changes, the proposal made by Inclan and Tiao (1994) is used. As an empirical application, whether or not the consideration of the changes obtained influences the diagram of conditional heteroscedasticity presented by the spot and futures series on the IBEX-35 index is tested for the Spanish stock market. The results show that the consideration of these changes considerably decreases the degree of persistence for the IBEX-35 index while, for its futures contract, the GARCH diagram disappears.