We provide an economic valuation of the riskiness of risk models by directly measuring the impact of model risks (specification and estimation risks) on VaR estimates. We find that integrating the model risk into the VaR computations implies a substantial minimum correction of the order of 10-40% of VaR levels. We also present results of a practical method - based on a backtesting framework - for incorporating the model risk into the VaR estimates.
This document is published in Journal of Banking and Finance (March 2014) http://dx.doi.org/10.1016/j.jbankfin.2014.03.019