Measuring Extreme Market Risk: The Sri Lankan Context

Abstract

Many empirical studies have been carried out in developed and emerging markets using methods to account for extreme market risk. A frontrunner in the methods used is Extreme Value Theory (EVT). In this study, these methods are applied to the All Share Price Index (ASPI) of the Colombo Stock Exchange (CSE) to obtain risk forecasts for VaR (Value at Risk) and ES (Expected Shortfall) at 99% confidence level. Recognizing the merits of both conditional and unconditional models for risk measures the most suitable unconditional model and conditional model are found for the ASPI. Contrary to other studies the Historical Simulation method was found to be more appropriate for obtaining static estimates than the static EVT model. The Two-Step Approach of McNeil and Frey which combines EVT and Generalized Auto-Regressive Conditional Heteroskedasticity (GARCH) methodologies was found to be the most appropriate conditional model. Backtesting of models is done using a binomial test, Christoferssen's conditional coverage test and McNeil and Frey's ES test.

Measurement of extreme market risk: Insights from a comprehensive literature review

“…Select empirical studies that belong to the first sub-strand include Bali andNeftci (2003), Brooks et al (2004), and Chan and Gray (2006), Ozun et al (2010), Ren and Giles (2010), Karmakar (2013), Chou and Wang (2014), Wijeyakulasuriya and Wickremasinghe (2015), Dahlen et al (2015), and Youssef et al (2015), Kellner and Rösch (2016), Muela et al (2017), Gkillas and Katsiampa (2018), and Liu et al (2018). Unlike majority of these studies, Ren and Giles (2010) and Dahlen et al (2015) did not employ the usual Hill scatterplots and the mean excess function to identify the threshold for extreme returns.…”