The recent regulation (page 32) on PRIIPs requires to compute a VaR-equivalent volatility defined as

$$\mbox{VEV}=\frac{\sqrt{3.842-2\ln \mbox{VaR}}-1.96}{\sqrt{T}}$$

Does anyone have an idea how they came up with that formula?

  • $\begingroup$ The formula should be corrected. It should be ln (VaR/ initial value of the PRIIPS) and not ln(VaR) This is a typo. $\endgroup$ – user22115 Jun 12 '16 at 12:01
  • $\begingroup$ Do you have a source for that? $\endgroup$ – Bob Jansen Jun 12 '16 at 14:53
  • $\begingroup$ @despieg I sort of guessed that, but still the formula is unclear. The number 3.842 is the 95% quantile of the chi-square distribution, but I don't see why it should play a role here. $\endgroup$ – splinter123 Jun 12 '16 at 16:26

Let's assume T=1 and let S be a geometric gaussian process with zero drift, i.e. $\ln(S_1/S_0)$ is normally distributed with mean $-1/2\times\mathrm{VEV}^2$ and volatility VEV.


$$\ln(\mathrm{VaR}/S_0) = -1/2\mathrm{VEV}^2 - \mathrm{VEV} \times 1.96$$ with the VAR at $0.975$ quantile.

This is a quadratic equation in VEV, with solutions

$$\mathrm{VEV} = -1.96 \pm \sqrt{1.96^2 - 2\ln(\textrm{VaR}/S_0)}.$$

We take the positive solution and are done.


To answer this question it might make sense to mention the VaR part and VEV part separately.

  1. VaR example using a parametric approach to VaR: assuming an investment of $V_0 = 10,000 $ into the financial asset and its daily log return following a normal distribution such that $r_t \backsim N(\mu, \sigma^2)$ with mean $\mu=0.02$ and standard deviation $\sigma = 0.7$, compute the VaR of the investment at $p = 2.5\%$ for 1 day. Since the VaR may be defined in several ways, e.g. as value $\bigtriangleup V_1 $ on which the investment can depreciate or as value $V_1$ ($V_1 < V_0$), which shows the new level of the investment, the later definition is used in the example. The solution would be:
    \begin{equation} VaR = V_t(exp(\Phi^{-1}(0.025)\sigma + \mu)). \end{equation} 2.5-quantile of standard Normal distribution is $\Phi^{-1}(0.025)$ is -1.96. \begin{equation} 10,000(exp(-1.96 * 0.7 + 0.02)) = 2587.22, \end{equation} meaning that with probability $2.5\%$ a one-day investment of 10,000 into the asset will be 2587.22 or less.

  2. As one may guess, given a Normal distributional assumption on daily log-return and already calculated VaR, for example by means of a Monte Carlo simulation, one may infer $\sigma$. It seems that a question being answered by VEV is: "what would be a scale parameter labelled $\sigma$ of a Normally distributed random variable, if one assumes "demeaned" distribution $N(-\dfrac{1}{2} \sigma T, \sigma^2T)$ and value of 2.5-quantile at the VaR level. $T$ denotes the period in years.
    [ see this link for more info on T and distributional assumption Is This A Viable Alternative Options Pricing Method? ]. \begin{equation} VaR = V_t(exp(\Phi^{-1}(0.025)\sigma \sqrt{T} +(-\dfrac{1}{2} \sigma T) )), \end{equation} which is rewritten in as a quadratic equation: \begin{equation} \dfrac{1}{2} T\sigma^2 + 1.96\sqrt{T}\sigma + \ln{\dfrac{VaR_0}{V_0}}=0, \end{equation} which can be solved for sigma via discriminant as $ax^2 + bx + c = 0$, when $\sigma$ replaces the $x$. Finally, a negative solution being disregarded one gets VEV formula: \begin{equation} \sigma = \dfrac{\sqrt{3.8416-2\ln{\dfrac{VaR_0}{V_0}}}-1.96}{\sqrt{T}}. \end{equation}


protected by Community Sep 28 '16 at 10:31

Thank you for your interest in this question. Because it has attracted low-quality or spam answers that had to be removed, posting an answer now requires 10 reputation on this site (the association bonus does not count).

Would you like to answer one of these unanswered questions instead?

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