Measuring the Effectiveness of Contemporary Models to Forecast Market Risk of Commercial Banks in Bangladesh

Abstract

This paper investigates the validity of Hybrid model in forecasting market risk considering VaR (Value at Risk) recommended by (Boudoukh, Richardson and Whitelaw, 2007). It then considers the coherent version of the risk measure VaR, ES (Expected shortfall) in forecasting daily investment risk for six commercial banks operating in Bangladesh for the first time applying the Hybrid Model. We consider the daily log return covering from 2010 to 2020, a particularly idiosyncratic performance period for Dhaka Stock Exchange (DSE). We have adopted RiskMetricsTM(RM) and Historical Simulation (HS) models as less complicated alternate to Hybrid model, which is a combination of HS and RM model. We then consider GARCH (generalized autoregressive conditional heteroscedasticity) and GARCH-t models being more complicated alternate to the Hybrid model. Our investigation reveals that Hybrid model, being a parsimonious model, performs poorly compared to GARCH and GARCH-t models in case of VaR forecast and it also performs poorly compared to RM, HS, GARCH and GARCH-t models in case of ES forecasts as evidenced by the outcomes of several backtests as per BASEL-III accord.

Keywords:
Market risk, BASEL-III, Hybrid model, Value at Risk, Expected Shortfall