A comparative cross-regime analysis on the performance of GARCH-based value-at-risk models: Evidence from the Johannesburg stock exchange

A topic of recent interest in financial risk management is the predictive accuracy of Value-at-risk (VaR) models for adequate capitalization under different market conditions (or regimes). This article assesses the forecasting performance of popular GARCH-based volatility models in the context of VaR estimation. In particular, we conduct a cross-regime analysis between time periods whereby market conditions experiences a shift. Stock returns data from the FTSE/JSE Africa All Share index were selected for the evaluation of both long and short positions of trade. Despite prior findings of the long memory models dominating in the South African financial market, we conclude that such dominance does not necessary hold when assessed under different regimes of the market. Moreover, our findings indicated a need for implementations of model switching policies, which may provide significant improvements in forecasting and minimize chances of VaR estimates falling short of actual trading losses.