Modeling Extreme Risk in Major U.S. Tech Stocks: Integrating GARCH, EVD, and CVaR

This study aimed to identify the most accurate method to assess the extreme risk associated with major technology stocks in the United States, focusing on Apple, Microsoft, Alphabet, Amazon, and NVIDIA stocks. In the intricate and ever-evolving financial landscape, conventional methods such as Value-at-Risk (VaR) frequently prove to be inadequate in capturing events that are rare but have the potential to have a substantial impact. To address this shortcoming, this study proposes a novel integration of the Generalized Autoregressive Conditional Heteroskedasticity (GARCH) model with the Extreme Value Distribution (EVD) approach, complemented by Conditional Value at Risk (CVaR), to formulate a more robust and conservative risk measurement. The findings indicate that the GARCH-EVD-CVaR approach generates risk estimates with greater precision, particularly in volatile market contexts. Specifically, the GARCH-EVD model reduced the Akaike Information Criterion (AIC) values by an average of 25.3 points and decreased VaR estimates by up to 14% at the confidence level of 99%. In practice, this model can help institutional investors conduct more accurate portfolio stress tests and offer regulators a framework for assessing systemic tail risk under volatile market conditions.