Estimating Value-At-Risk and Expected Shortfall on Emerging Markets: Evidence from Ghana Stock Exchange

Abstract

An important component of Value-at-Risk (VaR) and Expected Shortfall (ES) estimation is using robust volatility models, thus the need to test the relative performance of a wide range of volatility forecasting models to ascertain the one that gives the most adequate VaR and ES forecast. Risk management has become very necessary for institutions in all sectors of an economy. Stock markets represent almost all sectors of an economy; the Ghanaian stock market being no exception. With listed companies from manufacturing, financial, agricultural, mining, oil and gas and services industry, the Ghana Stock Exchange (GSE) is a good place to test risk measures basically developed for advanced and liquid economies. This study tests the relative performance of a range of volatility models (RiskMetrics (EWMA), GARCH, EGARCH) in forecasting Value-at Risk (VaR) and Expected Shortfall (ES). Bi-weekly returns of Fan Milk Limited (FML) and GSE Composite Index (GSECI) spanning over a period of fourteen years (Jan 2000-Dec 2013) were used. Residuals of the GARCH and EGARCH models were assumed to follow a normal and student-t distribution. The first 215 observations were used to estimate model parameters while the last 150 observations used to verify and back-test the VaR forecasts. From the empirical results the ARMA-GARCH with normally distributed error terms was seen as the best volatility model for VaR and ES estimation for both FML stock and GSECI. Even though other models passed the Kupiec Unconditional coverage test, the number of VaR violations outnumbered that of the ARMA-GARCH model.