Using the Full Five-Year History for PRIIPs Stress Volatility
Summary
The discussion concerns which historical returns to use when calculating the PRIIPs stress scenario. The question is whether the return series should cover the same maximum five-year period used for other performance scenarios, or whether the product’s recommended holding period should determine the history used for stress calculations.
The responses favor calculating rolling standard deviations across the available five-year history, using windows of 21 or 63 days according to the scenario horizon. One response cautions that restricting the sample to the recommended holding period could leave too little history for short-horizon products. Another notes that a short sample could produce a stress estimate below the unfavourable scenario, undermining the intended relationship between scenarios. These are interpretations of regulatory text in a forum discussion, not a formal regulatory ruling; the source does not provide independent validation or detailed implementation guidance.
Key ideas
- The proposed interpretation applies rolling volatility windows across the available five-year history.
- The window length discussed is 21 or 63 days, depending on the stress horizon.
- Using only the recommended holding period could leave too little data for short-horizon products.
- A short estimation history could produce a stress scenario less severe than the unfavourable scenario.
- The answers express practitioner interpretations rather than a formal regulatory determination.
Tags
Full text
# PRIIPs Stress Scenario # PRIIPs Stress Scenario Let me ask you about the following rationale regarding stress scenario (IV Annex 10 (c)): " Identify for each sub interval of length w the historical lognormal returns rt, where t=t0, t1, t2, …, tN. " I find myself wondering what the t0,...,tN are: in fact, given that RTS only mentions for historical data of up to 5 years, I interprete the above as saying that t0,...,tN should be the 5 years data used in the rest performance scenarios as well. Otherwise the RTS should be more specific of how and where exactly these "new data" come from. Would you agree to that interpretation or are you of the opinion that for stress purposes one needs to get more historical data? Thank you very much in advance for your answer! ## Answer by Rick N Backer (score 1, accepted) https://quant.stackexchange.com/a/36434 I interpreted it as the rolling standard deviation in 21 or 63 day Windows for the whole five years (less 21 or 63 days respectively) depending on whether you are doing one year or longer stress scenarios. I believe some think that the one year stress scenario means you only go back one year with the 21 day rolling window but this would surely be inconsistent with the other scenarios where regardless of the RHP you still use the maximum amount of historic returns. ## Answer by Andrej Iring (score 0) https://quant.stackexchange.com/a/37057 The Rick n Backer approach to the rolling window for stress scenario is the only logical and right approach you can choose. These rolling windows should be calculated for all the 5Y(if avaiable) data history. If we would only use data based on RHP we would get in quite big troubles with products with short RHP e.g when the produdct has shorter RHP than 21 days we wouldn't be able to calculate the risk scenario but all the scenarious are required to be shown in KID. Or if RHP would be more than 21 days but still somethig short we could get that stressed volatitily would be lower than the historic volatility which could result in that the stress scenario would be more optimistic than the unfavourable scenario which would be contradicting the idea behind the stress scenario.
Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.