A Coupling of Extreme-Value Theory and Volatility Updating with Value-at-Risk Estimation in Emerging Markets: A South African Test

21 Pages Posted: 7 Jul 2015

See all articles by Anthony Seymour

Anthony Seymour

University of Cape Town (UCT)

Daniel A. Polakow

University of Cape Town (UCT)

Date Written: July 7, 2015

Abstract

This research is aimed at a formal appraisal of recent advancements in stochastic volatility modeling and extreme-value theory to application of value-at-risk computation in particularly volatile markets. Established methods such as historical simulation are prone to underestimating value-at-risk in such developing markets. Two contemporary methods of value-at-risk calculation are tested on a representative portfolio of South African stocks. The first method incorporates extreme value theory. The second model includes both extreme value theory and volatility updating (via GARCH-type modeling). The combined GARCH-type time-series approach and extreme value theory model is found to provide significantly better results than both straightforward historical simulation as well as the extreme value model. In no instance, however, were results on these VaR methods as good as those obtained when the same methods were tested in developed markets.

Keywords: backtesting; extreme value theory; GARCH, historical simulation; RiskMetrics; value-at-risk

JEL Classification: D81, G10

Suggested Citation

Seymour, Anthony and Polakow, Daniel A., A Coupling of Extreme-Value Theory and Volatility Updating with Value-at-Risk Estimation in Emerging Markets: A South African Test (July 7, 2015). Multinational Finance Journal, Vol. 7, No. 1/2, p. 3-23, 2003. Available at SSRN: https://ssrn.com/abstract=2627556

Anthony Seymour (Contact Author)

University of Cape Town (UCT) ( email )

Private Bag X3
Rondebosch, Western Cape 7701
South Africa

Daniel A. Polakow

University of Cape Town (UCT) ( email )

Private Bag X3
Rondebosch, Western Cape 7701
South Africa

Register to save articles to
your library

Register

Paper statistics

Downloads
31
Abstract Views
446
PlumX Metrics