Journal of Operational Risk

Evaluation of backtesting on risk models based on data envelopment analysis

Grigorios Kontaxis and Ioannis E. Tsolas

  • Value-at-risk models' efficiency can be heavily impacted when horizons or confidence intervals change.
  • Backtesting results can vary among value-at-risk models with different horizons.
  • Data envelopment analysis can be a suitable alternative for the selection of accurate risk methodologies.

In this study, different value-at-risk (VaR) models, which are used to measure market risk, are analyzed under different estimation approaches and backtested with an alternative strategy. The methodologies examined include filtered historical simulation, extreme value theory, Monte Carlo simulation and historical simulation. Autoregressive-moving-average and generalized-autoregressive-conditional-heteroscedasticity models are used to estimate VaR. Selected VaR functions, marginal distributions and different horizons are combined over a set of extreme probability levels using the time series of the Financial Times Stock Exchange 100 index. Data envelopment analysis, which investigates the efficiency of VaR models using a number of different parameters, is carried out in lieu of standard backtesting techniques. This study shows that, for short horizons, some approaches underestimate VaR. However, a sufficient number of models present violation estimates that almost converge to the desired ones. Surprisingly, aside from historical simulation and some extreme value theory models, overlapping returns tend to yield conservative ten-day VaR estimations for most models; in cases of nonoverlapping returns, the results are satisfactory.

Sorry, our subscription options are not loading right now

Please try again later. Get in touch with our customer services team if this issue persists.

New to View our subscription options

You need to sign in to use this feature. If you don’t have a account, please register for a trial.

Sign in
You are currently on corporate access.

To use this feature you will need an individual account. If you have one already please sign in.

Sign in.

Alternatively you can request an individual account here