CMStatistics 2016: Start Registration
View Submission - CFE
Title: Assessment of Value-at-Risk estimation of long and short-memory GARCH-class using filtered historical simulation methods Authors:  Pilar Grau - Universidad Rey Juan Carlos (Spain) [presenting]
Abstract: The use of both filtered historical simulation and bootstrap filtered historical simulation of Value-at-Risk (VaR) is proposed under different realistic generating processes that include short and long-memory. Various non-linear short and long-memory GARCH-class under three different density function, Gaussian, Student and skewed Student are used to evaluate the predictive performance of the method. Additionally, daily data on three well known active stock indices are used to empirically evaluate the VaR estimates. The predictive performance is evaluated in terms of different criteria, such as the tests of unconditional and conditional coverage and the independence test. Results from different models specifications and methods are ranked showing that VaR using bootstrap filtered historical simulation under long-memory GARCH-class models outperforms over short memory ones.