Measurement of Extreme Market Risks: An Empirical Analysis Across Asset Classes and Financial Markets
Posted: 24 Oct 2019
Date Written: September 28, 2019
Financial Risk Management (FRM) crucially relies on the distributional assumptions of asset returns. In this paper, we apply methods from extreme value theory (EVT) to examine the limiting distribution of large returns across asset classes viz equities, bonds, foreign exchange, and commodities selected from mature, emerging and frontier markets across geographies namely Americas, Europe, Asia, and Middle East and Africa (MEA). Our initial findings suggest that the much heralded generalized extreme value (GEV) and generalized pareto (GP) distribution may not most precisely represent the distribution of extreme gains and losses; rather the generalized logistic (GL) distribution may be a superior candidate distribution owing to its improved fit with tail returns. We try to add to the robustness checks of our empirical findings by highlighting the time-varying characteristics of our distributional parameters. Furthermore, we argue that the empirical results may have substantive inferences for risk modeling, especially that for tail risk measures viz Value-at-Risk (VaR) and Expected Shortfall (ES). This is important since the market risk models that cannot accurately estimate the fat tails of asset returns’ distribution can severely underestimate (downside) tail risk to possible catastrophic detriment of investors.
Keywords: Extreme Value Theory, Generalized logistic distribution, Generalized Extreme value, Generalized Pareto distribution, VaR, ES, L-Moments
JEL Classification: C58, G24, G31, G15
Suggested Citation: Suggested Citation