Modeling volatility of illiquid stocks

Abstract

Modeling the volatilities of financial assets has gained increasing attention in the recent years. However less attention has been paid to modeling the volatilities of relatively illiquid stocks. This paper evaluates the performance of alternative models in forecasting the volatilities of illiquid stocks traded on the Australian Stock Exchange. Our purpose is twofold. Firstly, Classes of GARCH models with daily range as regressor are assessed using daily data. Both in-sample estimates and out-of-sample forecasts show that the volatility forecasts from GARCH type models are enhanced by incorporating daily range as regressors. A simple GARCH (1,1) model with range as regressors is at least as good as more complex GARCH models with range as regressors. Secondly, the information content of implied volatilities is examined using monthly data. The forecasts based on implied volatilities are compared with those based on GARCH type models. While all models tend to over-predict the volatilities when the monthly data is used, the implied volatility model performs much worse than the GARCH type models and it seems clear that implied volatility models do not predict volatilities at all well for illiquid stocks.