Value at Risk models of autoregressive quantiles for improved performance on financial criteria
41 Pages Posted: 24 Aug 2015 Last revised: 25 Oct 2017
Date Written: October 25, 2017
Abstract
We study alternative specifications of conditional quantile models that are used to estimate Value at Risk (VaR). Our proposed specifications include the incorporation of a slow moving component in the quantile process, along with recent aggregate returns as regressors. We consider a range of criteria with the aim of identifying models with improved performance in both statistical and financial terms. These criteria include the potential to lower transaction costs and realized losses in excess of the VaR on exceedance days. We find that for many assets, the proposed specifications lead to improved performance.
Keywords: Value at Risk; CAViaR model; Component model; Conditional loss; Conditional excess loss
JEL Classification: G11, C58
Suggested Citation: Suggested Citation