Why PRIIPS KID will not satisfy MIFID IIMay 16, 2017 - 14:45
Manufacturers and distributors of packaged retail investment and insurance products (PRIIPS) will have two main requirements:
- Manufacturers will be required to generate a KID using the prescribed methodology defined by the PRIIPS Regulatory Technical Standard (RTS)
- They have a separate requirement to comply with MIFID II, and show that products meet the needs of the target clients.
Manufacturers and distributors may be tempted to think that the PRIIPS KID will be sufficient. Legal and regulatory experts (with no quantitative skills or expertise) may conclude that if one part of the EU has determined how a product should be presented to investors, that this will be sufficient for another part of the regulatory framework. Unfortunately, for manufacturers of category 3 PRIIPS (structured products) the detail on the RTS mean that in our opinion the resultant KID is not fit for purpose, and will not enable manufacturers to comply with their MIFID II requirements.
In this note we highlight three main areas where the PRIIPS KID is insufficient for manufacturers of structured products to comply with MIFID II.
PRIIPS IS NOT A GOOD WAY TO TEST STRUCTURED PRODUCTS
MIFID II requires Manufacturers and Distributors to show that the product meets identified target market needs, characteristics and objectives; these would include generating and reviewing metrics for:
Market risk equivalent volatility (to address “attitude to risk”)
- Probability and extent of losses, especially large ones (to address capacity for loss).
- Benchmark-relative metrics, such as probability of and expected shortfall (to address relative risk and capacity for loss).
- Probability of various product events, for example the chance of, and expected losses from, the possible triggering of a down-and-in put option typical of autocalls (giving the investor more information on how they might suffer losses).
Unfortunately, the PRIIPs KID provides no information with which to do this. We highlight three main reasons why the PRIIPS KID is not fit for purpose:
- It is unstable
o The SRI measures risk using historic data from a rolling 5-year window, this is too short, and will result in risk scores that are variable depending on when the product is tested
o Use of 97.5% VaR means that products with a chance of loss of about 2.5% can bounce back and forward from being rated as very risky to having little risk.
- It does not reflect the risk
o It’s not clear how the adjustment to the bootstrap results should reflect dividends or yield
o Risk bands are too wide and inconsistent with current UCITS KIID scale
o Using a risk neutral distribution is not an accurate reflection of the risk of structured products
- Results are unusable
o The SRI is a single meaningless dimensionless scale that combines market and credit risk using arbitrary weights
o The scenario analysis is unrealistic, probably misleading and hard to interpret.
THE RESULTS ARE UNSTABLE
- The RTS proposes that volatility is calculated using VaR rather than conditional-VaR; VaR is generally unstable as it just measures one point on the risk landscape, whereas conditional-VaR measures the extent of a region of risk; for example, some products have a 97.5% VaR that, in response to very small moves in the underlying, oscillates between zero and a significant loss of capital. Conditional-VaR would provide a more realistic continuous risk response; note also the use of conditional-VaR in FRTB bank risk regulations; in addition, the overly extreme 97.5% confidence level is unstable since extremes are rarely encountered - a 90% confidence level would give a more stable risk measure.
- The SRI uses 5y of historical data; a much longer history is needed to capture as many relevant business and banking cycles as possible. It means that results will be very variable. The same product would have a very different risk score in 2010 than if it was tested today. Where data is unavailable a beta-to-a-reasonable-proxy approach can be used, and is in fact allowed under PRIIPs in certain circumstances.
THE SRI IS NOT A GOOD REFLECTION OF THE RISK
The Market Risk Measure (MRM) is based on simulations that:
- Ignore asset yields such as dividends: Underlying’s are simulated as though their dividend payouts do not exist; for example, a synthetic underlying can be constructed that has high dividends (and so posing a high risk, for example, of a downside put being exercised against the investor) but the MRM simulation will assume the index drifts at the risk-free rate.
- Use volatility bands that are much higher than the CESR bands for UCITs SRRI (to convert VaR-equivalent-volatility to a MRM score).
THE RESULTS ARE UN-USEABLE AND POSSIBLY MISLEADING
SRI uses just one number, and a meaningless dimensionless scale between 1 and 7:
- This scale does not map to any risk scale in common use by investors; for example, equity investors typically use volatility of returns, credit investors typically use CDS spreads or distance-to-default probabilities.
- The SRI combines market (MRM) and credit (CRM) risk measures using arbitrary relative weights. Condensing different risks into a single number loses information on these different product risk dimensions; providing investors with fewer numbers, in the interests of clarity, goes too far here and detracts from, rather than improves, product understanding
The “Scenarios” aim to answer the question “what could I get in return” using 4 scenarios: stress/unfavorable/moderate/favorable. The calculated numbers seem intended to reflect the likely fair values in those scenarios; however:
- The Recommended Holding Period (RHP) Stress scenario corrects underlying drift rates to a risk-free rate (the other 3 scenarios use real-world drift rates); not only is this inconsistent, but could lead to a less stressful scenario than, say, the unfavorable if an underlying has a high dividend yield.
- The simulations at intermediate holding periods appear to use risk-free drift rates for the underlying: para24 “shall pick three underlying simulations as referred to in points 16 to 24 of Annex II used for the calculation of the MRM and one underlying simulation as referred to in point 13 of this Annex”. This is inconsistent with the 3 RHP scenarios which appear to use real-world drift rates.
- As noted above, correcting drift rates to risk-free provides neither consistency between products with different underlying, and could make the stress scenario less conservative than the unfavorable.
As a result, the values in the scenario table are almost certainly inaccurate, they are inconsistent and they have the potential to mislead investors.
Stress scenario is difficult to assess. The stress scenario is based on scaling-up asset returns to high percentiles of their distribution of historical volatility, generated with sliding windows whose length varies with the simulation horizon, as does the percentile to use. These inflated returns are then corrected to a risk-free drift rate. It is not clear what probability an investor could attach to the resulting PRIIPs values. Nor is it clear, as mentioned above, that a risk-free drift rate is conservative.
Finally we think that it is difficult to see how the results from the scenarios could be used by advisers and investors.
Manufacturers of structured products will have to generate a PRIIPS KID, but it is hard to see this document being used by investors. Manufacturers and distributors will need better information about the returns that investors may receive and the risks that they face to comply with the requirements of MIFID II.
David has been involved in equity derivatives, equity structuring and the structured product market for over 25 years. Before setting up CUBE in 2013 David worked at J.P. Morgan, Barclays and RBS. David has worked with and for retail product providers, discretionary managers and institutional investors.