Evaluating Value at Risk 部分内容如下： Evaluating Value at Risk Methodologies: Accuracy versus Computational Time
First Version: September 1995
Current Version: November 1996
Abstract: Recent research has shown that different methods ofcomputing Value at Risk
(VAR) generate widely varying results, suggesting the choice of VAR method is very
important. This paper examines six VAR methods, and compares their computational time
requirements and their accuracy when the sole source ofinaccuracy is errors in
approximating nonlinearity. Simulations using portfolios of foreign exchange options
showed fairly wide variation in accuracy and unsurprisingly wide variation in computational
time. When the computational time and accuracy ofthe methods were examined together,
four methods were superior to the others. The paper also presents a new method for using
order statistics to create confidence intervals for the errors and errors as a percent of true
value at risk for each VAR method. This makes it possible to easily interpret the
implications of VAR errors for the size of shortfalls or surpluses in a firm's risk based capital.
Matthew Pritsker is at the Board of Governors ofthe Federal Reserve System. The views expressed in this paper
reflect those ofthe author and not those ofthe Board ofGovernors ofthe Federal Reserve System or other
members ofits staff. T he author thanks Ruth Wu for extraordinary research assistance, and thanks Jim O'Brien and
Phillipe Jorion for extensive comments on an earlier version, and also thanks Vijay Bhasin, Paul Kupiec, Pat White,
and Chunsheng Zhou for useful comments and conversations. All errors are the responsibility ofthe author.
Address correspondence to Matt Pritsker, The Federal Reserve Board, Mail Stop 91, Washington, DC 20551, or
send email to mpritsker@Frb.gov. The author's phone number is (202) 452-3534.
This paper was presented at the Wharton Financiallnstitutions Center's conference on Risk Management in
Banking, October 13-15, 1996. |