recently new document has been provided by EU supervisors regarding the new PRIIPs KID methodology. (PRIIPs KID is 3page document about the fund/product that is being sold to clients. It has to inform about scenarios and risk of the investment)

This new document include examples of calculation and my question is regarding the stress performance scenario for category 2 PRIIPs, which is on page 23 and 24.

I tried to reconstruct this example and was successful with all other scenarios except the stressed. That means that I should have same dataset as they have. However my results for stressed scenario are slightly shifted. Moreover the stressed volatilities (red bold on page 24) are different for 3 and 5 holding period, which does not correspond to the methodology on page 23. The formula for stressed volatility is based on historical data only and should NOT depend on the holding period, aside from 1 year period which is special.

So my question is:

  • Why are the stressed volatilities for 3 and 5 holding period different?
  • Is anyone able to reconstruct their numbers precisely?

If code or excel is needed I can share, however I do not know how to share datasets here.

Thank you.


Requested info about the dataset.

  • Source: Bloomberg
  • Start date: 01.05.2012
  • End date: 28.04.2017
  • Remove two NAs

After there should be 1280 observed returns (1281 prices). For favourable, unfavourable and neutral scenario I get 1e-10 (rounding) error.

EDIT 3Y and 5Y stress volatilities should be equall, as I found out in slide 26 of presentation from the latest ESA's workshop 27.11.2017 in frankfurt. Which means full dataset should be used for rolling volatilities.

  • $\begingroup$ Matuš, I have also tried to reconstruct the example given in the document, but without success. Could you please share the sample you have used or at least the basic characteristics (start date, end date, # of observations)? $\endgroup$
    – Ryko
    Commented Aug 30, 2017 at 8:31
  • $\begingroup$ Sure. I will edit it into the question. $\endgroup$ Commented Sep 4, 2017 at 9:39
  • $\begingroup$ Matus, just saw your edit today 1st December (quote EDIT 3Y and 5Y stress volatilities should be equall, as I found out in presentation from the latest ESA's workshop 27.11.2017 in frankfurt. Which means full dataset should be used for rolling volatilities.unquote) and I am not comfortable with it as there is no official document denying the JC 2017 49 flow diagram result, which indicates on page 24 different stressed volatilities for 3Y and 5Y. Making no difference between 3 and 5Y stressed volatilities would mean that the stressed volatility is not determined on a 3 or 5Y observation $\endgroup$ Commented Dec 1, 2017 at 11:13
  • $\begingroup$ (continued) but only on a 5Y observation. The same would apply for 1Y, which would mean that you would look at 5Y data on 21D volatility, and I do not understand how it could be called 1Y stressed volatility. Could you please explain or provide supporting official document source ? $\endgroup$
    – Bob Jansen
    Commented Dec 1, 2017 at 15:37
  • $\begingroup$ I have uploaded the presentation myself for you since iam unable to find it online anywhere... it is edited into the question $\endgroup$ Commented Dec 4, 2017 at 8:38

1 Answer 1


Well probably it is some sort of mistake on EU supervisor side because if you look at stress scenario for Cat 3 in that document you will find that it's just nonsense. For example they are using only 2Y (which is the RHP in their example) of data but they should use full 5Y. Next if you look at the returns (page 31) the first nine returns are corresponding to the dates on the left but the other returns and even rolling vol. are not (the last return is for 28.3.2017 not 27.1.2017 as they are presenting). Furthermore the volatilities they calculated are wrong. The last five volatilities are rolling volatilities for 22.3.2017 to the 28.3.2017 and the rolling window is not 63 (as it should be due to RHP is 2Y) but 21. And the first volatilities well I don't know how they calculate them I have try to reproduce them from data but I was not successful. So that is that, they probably publish corrected document but until then you should rely only on RTS when it comes to the stress scenario.

  • $\begingroup$ Thanks. I suspected it is error from their side as well. Also, I have noticed some mistaken dates for Category 2 as well. For example their say they use Stoxx50 untl 25.5.2017 on page 8, but I found out they use data only until 28.4.2017. So I hope they will put more effort into it and publish correct calculations soon. $\endgroup$ Commented Sep 4, 2017 at 9:36

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