Modelling and Forecasting Volatility of Returns on the Nairobi Stock Market using Arch Models

This study empirically models and forecasts volatility (conditional variance) on the Nairobi Stock Market using the ARCH models namely; GARCH-M (1,1), EGARCH-M (1,1) and TGARCH-M (1,1). The daily NSE 20-share index data over a period of 10-years was used in the analysis. The competing volatility models were estimated and their specification and forecast performance compared using RMSE, MAE, MAPE, TIC and R2. The NSE stock returns exhibits volatility clustering, asymmetric effects, leptokurtosis which are common characteristics for most financial time series data. Overally, the EGARCH-M (1,1) emerged the best model with the t-distribution over the GARCH-M (1,1) and TGARCH-M (1,1) due to it’s lower values of the RMSE, MAE and MAPE. Comparison using the R2 also gave the same results in that the EGARCH-M (1,1) emerged the best due to its highest value of R2 (0.187010) unlike the TGARCH-M and GARCH-M