PRIIPs stress scenario 2 calculation period of calculation

Question

I have recreated the calculations based on the examples in the JC 2017 49 flow diagram. A couple of questions that I am hoping someone can help with.

I can get back to the exact numbers for everything apart from the stressed scenario. For 5 yrs I get the same result. For 3 yrs I get the correct result if I use 3 years as the RHP and for 1 yr, I think they have taken the SD of 22 days instead of 21 days, but can get back to a result (though the number of days they have used seems strange).

Has anyone else had this problem?

Looking at the final annexes, in Annex IV page 20 it states: Calculation of scenario values for the recommended holding period 4. The scenario values under different performance scenarios shall be calculated in a similar manner as the market risk measure. The scenarios values shall be calculated for the recommended holding period.

This suggests that the RHP period should be used for the calculation of all performance scenarios, however, having read a number of different responses to questions on this site the general consensus is that the full 5 years data should be used. This seems inconsistent with the annexes and also the fact the other stress scenarios are based over the RHP. I am confused as to why 5 years be used when the guidance seems to confirm that the RHP should be used?

Thanks in advance for all your answers.

Answer 1

Well the thing is that you are using the RHP when you are simulating into the future. But when you want to rescale data with stressed volatility you should use all the data for calculation of the stressed volatility. Few weeks ago there were workshop in Frankfurt (link to the presentations was posted by Matúš Košík in this question link) where were presented new presentation with examples of the calculations. In this presentation the stressed volatilities are no longer different for the >1Y RHP (there is still problem to get the values they are presenting i was able to get only the volatility only for 1Y). Also in this question you can find some of the problems which would rise if you won't use all the historic data for the calculation of the stressed volatility. For example when you want to calculate stressed scenario for the Category 3 product with some RHP you should get the 5Y data, then calculate rolling volatilities (rolling window is based on the RHP) on all of the 5Y data, then choose the stressed volatility based on the percentile (this is also based on the RHP), after that you should rescale your data and after all this you will make the simulation for the RHP using bootstrap.