Estimating PRIIPs Category 2 Stress Volatility from Rolling Returns
Summary
The document clarifies a calculation for estimating volatility in a PRIIPs Category 2 stress scenario. The questioner first forms overlapping 63-day log returns, then calculates the standard deviation of those returns. The response says to calculate daily log returns instead and estimate standard deviation over rolling windows of 63 daily observations. The resulting series of rolling standard deviations can then be used to determine percentiles for the stress calculation.
The exchange illustrates how confusing a horizon return with a rolling estimate of daily-return volatility can produce a very different value. It gives no worked recalculation or percentile results, and does not discuss details such as annualization or alternative window choices. The recommendation is presented as a practical interpretation rather than a full derivation of the PRIIPs methodology.
Key ideas
- Calculate daily returns before estimating volatility over rolling windows.
- Use rolling windows of 63 daily observations to create a series of volatility estimates.
- Determine stress percentiles from the distribution of rolling standard deviations.
- A standard deviation of overlapping 63-day returns is not the same calculation as rolling daily-return volatility.
Tags
Full text
# PRIIPs category 2 stress scenario calculation steps # PRIIPs category 2 stress scenario calculation steps I have not been able to get to the results of the stress scenarios. I am using the series suggested between 1.05.2012 and 1.05.2017 where I have 1283 daily values including both dates. My steps in excel are the following: - Since there are 5 years of daily prices , I calculate, for each possible day, the log normal performance on 63 days. For example in cell C65 =LN(B65/B2), then in C66 =LN(B66/B3), then in C67 =LN(B67/B4), until C1284 =LN(B1284/B1221). This makes 1282 - 63 = 1220 observations of returns. The average return of the observations is +2.26%. - In column D I calculate the difference between the return and the average return, so in D65 = -0.24% - 2.26% = -2.49%, in D66 = +2.16% - 2.26% = -0.1% and so on until in D1284 = 8.71% - 2.26% = 6.45%. - In column E I raise the results from each cell of column D to square. The column total is 6.01199142. - I divide this by 1220, to obtain 0.004927862. - I get the square root of this to get to the standard deviation and I obtain 0.07019873. This is very different than 0.017152366 indicated on page 24 of the flow diagram document. So I am obviously missing something. Can someone please lead me in the right direction ? Thank you. ## Answer by RDA (score 1, accepted) https://quant.stackexchange.com/a/36319 This is how I understand it. You first check the recommended holding Period for your fund. If it is bigger than 1 Year and you have daily prices you will have sub intervals of 63 observations as you say. However you compute the returns daily as usual (e.g. Ln(B2/B3)) and you take a rolling window for the standard deviation Estimation of 63 Variables (e.g. St.dev(C2:C64,C3:C65...etc). From this series of Standard deviations you take the different percentiles. This approach will lead you to reasonable outputs. Hope this helps
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.