Abstract
This dissertation evaluates the forecasting performance of asymmetric generalized autoregressive conditional heteroskedasticity (GARCH) models for estimating Value-at-Risk (VaR) in emerging markets. By challenging the traditional assumption of symmetric risk, the study introduces models that better capture the dynamics of financial volatility. Using data from five rapidly growing emerging markets, namely, Brazil, India, China, South Korea, and Taiwan. Data is sourced from the MSCI EM database via Bloomberg, the analysis spans May 11, 2000, to March 19, 2025.
Key findings reveal strong evidence of time-varying volatility, asymmetry, and volatility clustering. Among the models tested, the GJR-GARCH model demonstrates superior performance in forecasting asymmetric volatility and provides more accurate VaR estimates, particularly during market downturns. These results underscore the importance of accounting for the leverage effect and fat tails in return distributions. These findings suggest that regulators and financial institutions in emerging markets should integrate asymmetric volatility models, such as GJR-FARCH into VaR frameworks to better capture downside risk. Consequently, the evidence can help strengthen risk management and improve resilience during market turbulence.