Apuntes del CENES
Estimation of risk in a portfolio of assets
Record only. The repository marks this work as “Acceso abierto”: the PDF is not public, so there is no text to read here. View the record in the repository.
Abstract
This paper introduces the use of extreme value theory (EVT) and copula for the estimation of value at risk (VaR) for a three asset portfolio representative of the Colombian market. Returns on risk factors are adjusted by ARMA GARCH models and innovations for each of them are modeled by Pareto’s generalized distribution in order to estimate one-day volatility. Copulas are built on the assumption that innovations follow an empirical marginal distribution so as to represent the dependence structure among risk factors. Performance tests for a series of three month VaR estimations show that modeling volatility and dependence through the use of these theories result more appropriate than those based on normality assumptions.