RiskMetrics assumes zero mean for the calculation of value at risk (https://www.msci.com/documents/10199/5915b101-4206-4ba0-aee2-3449d5c7e95a)

In our data, the mean return is quite negative. Is there an alternative approach we can use to calculate VaR? The negative returns are due to the crisis period, while there is a reverse in the trend lately with returns being positive. Is there a way to include this new trend in our calculations in order to get a better result from the one that classic VaR would get?

  • 1
    $\begingroup$ The var tag is for Vector AutoRegression, not Value at Risk. Since VaR is just a quantile of a distribution, I have added the quantiles tag. $\endgroup$ Jan 16 '19 at 11:40

As far as I know, the mean is not necessarily zero in Riskmetrics. It depends on your definition of the mean equation. You should have a look in [Bawens (2006)][1] at page 83.

  • $\begingroup$ onlinelibrary.wiley.com/doi/full/10.1002/jae.842 $\endgroup$ May 10 '20 at 12:31
  • $\begingroup$ Can you summarise the contents of the link so the answer stays useful if the link rots? $\endgroup$
    – mdewey
    May 10 '20 at 12:50
  • $\begingroup$ $H_t=\lambda H_{t-1}+(1-\lambda)\epsilon_{t-1}'\epsilon_{t-1}$ with $\epsilon_{t-1}$ denoting the vectors of at least two returns' errors. $\endgroup$ May 11 '20 at 13:12

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.