CLASSICAL AND BAYESIAN SWITCHING VOLATILITY MODELS FOR ANALYSING STOCK RETURNS IN GHANA

Abstract

The focus of this study was to model and forecast stock returns of Ghana Commercial Bank on the Ghana Stock Exchange using classical and Bayesian switching volatility models. Due to the presence of stylised facts in stock returns, this study finds it imperative to identify an appropriate risk model that best describes these features. The data utilised in this study are the stock prices of Ghana Commercial Bank transformed into monthly averages of daily closing prices covering 138 months. The study applied the two-state Markov-Switching GARCH models in deciding on the appropriate models to forecast the stock returns under the classical and Bayesian perspectives. In choosing the substantive models, selection criteria’s such as log-likelihood, Akaike Information Criterion, Bayesian Information Criterion were considered under the classical estimation and Deviance Information Criteria was considered under the Bayesian estimation. Based on the selection criteria’s under both estimations, E-GARCH variance specification with skewed student-t conditional distribution (innovation) was found appropriate modelling the stock returns. The estimates under both approaches find the first regime to possess the features of “turbulent market conditions” while regime two exhibit “tranquil market conditions”. However, comparative risk analysis finds the Bayesian perspective to generally perform better in estimating VaR and ES at both the 1% and 5% respectively as compared to the classical perspective. Investors should invest in Ghana Commercial Bank due to the good returns associated with the stocks and where there is the existence of “turbulent market conditions”, the recovery rate is shorter for these stocks.