EBA relaxes modellability hurdles for market risk capital

Flexibility granted for assessing NMRFs on options, but constraints remain on committed quotes

Europlaza tower Paris
Europlaza, home to the EBA
© Daniel Rodet

The European Banking Authority has dialled back hurdles in a test for determining whether risks are hard-to-model under forthcoming trading book capital rules, but a restriction remains on the way quotes from the market are used in the test.

Derivatives traders are particularly relieved by a change to the way complex risks used to price volatility in options and swaptions will be capitalised, as two previously proposed approaches were found to be impractical.

“It seems the EBA understood our concerns and they acknowledge the two approaches did not make much sense,” says a market risk specialist at a European investment bank. “They now say the best way is to keep a principles-based approach and leave flexibility to banks to actually decide how to assess the eligibility of each model parameter.”

On March 27, the EBA published a series of technical standards fleshing out rules surrounding the use of internal models for calculating capital requirements under new market risk rules, known as the Fundamental Review of the Trading Book (FRTB).

One of the three technical standards lays out the rules of a quarterly test banks must undertake to determine whether they have enough data to accurately model their risk factors. Risk factors that fail the test are deemed non-modellable (NMRFs) and must be capitalised separately using a stressed capital surcharge.

The final FRTB – published by the Basel Committee on Banking Supervision in January 2019 – requires the test to be applied to the underlying inputs of parametric models that are used to represent the risk sensitivities of complex options functions such as curves, surfaces or cubes. But the text is silent on how to determine modellability of the whole function and capitalise different points on the curve if some are modellable and some non-modellable.

Bankers feared it would lead to an all-or-nothing approach, where whole curves would be treated as non-modellable if a single point on the curve failed the test.

The two alternatives the EBA put forward for consultation in June 2019 had aimed to ease the capital impact from this situation by allowing modellable points of a curve to be subject to lighter capital requirements. However, risk managers feared they were too complicated to be practical, and could be subject to conflicting interpretations.

The EBA has instead opted for a different approach that will allow banks to determine themselves how best to assess modellability of the function as a whole. However, this will be subject to dividing the function into maturity buckets and assessing the modellability of each bucket. The threshold per bucket will be at least 100 verifiable price observations over the past year; or at least 24 verifiable price observations over the past year, and no 90-day period that has fewer than four prices.

It seems the EBA understood our concerns and they acknowledge the two approaches did not make much sense
Market risk specialist at a European investment bank

As long as more than one bucket is modellable, the curve can be capitalised using the expected shortfall method set out in FRTB internal model rules. Those maturities that are classified as non-modellable will be subject to a capital add-on based on the worst 12 months of expected shortfall numbers for that risk factor.

“There is some additional flexibility now, and how we decide to model that [risk factor] will need to be approved by supervisors. As firms document and put in their model applications, I’m sure supervisory scrutiny will be there to ensure they are doing it correctly,” says a capital manager at a second European investment bank. “This is a positive change.”

The new approach will allow banks to consider simpler methods for assessing the modellability of the whole curve. The market risk specialist at the first European investment bank says an alternative method they envisage implementing would focus on the parameter that is most important for determining the output of their models.

Stephane Boivin - web.jpg
Stephane Boivin

“In a model, there is always one parameter that drives the overall curve or the surface, so that is a good proxy for assessing modellability,” says the market risk specialist. “If that is all modellable, then it is a good step for treating the curve as modellable.”

They point to the stochastic alpha beta rho model – typically used to price interest rate options and swaptions to represent the sensitivity of the position to volatility – as one model their approach would work for. They say implied volatility is mainly driven by a single parameter: at-the-money volatility. By using this single parameter as a proxy for the whole curve, the bank can simplify both the test and capitalisation of the different points along the single parameter’s output.

Stephane Boivin, a senior policy expert at the EBA, confirms the new rules will allow banks to recognise hierarchies in their models, and more generally give them extra freedom to define their own risk factors. Their risk factor definitions, however, need to be justified as eligible to regulators, and to pass two other tests applied to banks’ trading desks assessing whether their internal models accurately reflect actual risks and losses.

“This approach doesn’t pre-empt the choices banks make in terms of the specification of their risk factors,” says Boivin. “The options we had for consultation were a bit restrictive in this respect. Now it is clearer that banks can specify their risk factors as they want, as long as this is consistent in the backtesting, the P&L attribution test and the risk factor eligibility test.”

Two single-sided quotes

Banks say the technical standards on NMRFs are “broadly positive”, but highlight one rule that they say will reduce the sources of market pricing used to prove modellability. The idea is that a transaction does not have to have taken place to provide a verifiable price for internal models – a committed quote at which a market participant is obliged to transact could be enough.

The final rules require quotes to have both a bid and ask price from the same day, which can come from different dealers.

“The EBA still wants both bid and ask, but the quote can come from different sources,” says the market risk specialist at the first European investment bank. “So you could have a market-maker that would quote a bid price and another market-maker on another quote that gives you an offer price, and then it would be modellable provided the two quotes are for the same trading day.”

It’s still a constraint, but it is better than [what was first proposed] – getting rid of all the one-sided quotes that are a huge amount of the available quotes
Market risk specialist

The rule is softer than the approach the EBA had originally proposed, which would have required the bid and offer prices to come from the same dealer – so-called two-sided quotes. Banks had complained in responses to the June 2019 consultation that this stopped them from using a deep pool of single-sided quotes.

Although the final rule will mean single-sided quotes can be used, two sources warn it will still limit the number of available quotes for the test.

“It’s still a constraint, but it is better than [what was first proposed] – getting rid of all the one-sided quotes that are a huge amount of the available quotes,” says the market risk specialist.

Single-sided quotes are commonly found in asset classes that trade predominantly over-the-counter, which includes corporate bonds and derivatives.

Jacob Rank-Broadley, a director at technology and data vendor Refinitiv, says the ruling will most likely affect risk factors where committed quotes would have been vital to passing the test.

“If a risk factor is super illiquid then it shouldn’t be passing the eligibility test, and if it is trading every day then you don’t need quote data,” says Rank-Broadley. “But the whole point of committed quotes was the fact that you have got this bit in between where the quotes do provide value, [but] you only end up with a bid or offer because there might have only been one or two firms looking for a price on a given day. So our concern is that they have just eliminated that middle ground from being modellable.”

The Basel way?

Rank-Broadley says the provision will mean the EU rules are stricter than the Basel Committee’s version of the FRTB. A Basel FAQ document published in February 2019 defines a verifiable price simply as a committed quote to buy or sell an instrument, rather than referring to two-way quotes.

“Buy or sell doesn’t seem quite consistent with what the EU has proposed,” says Rank-Broadley. “If we have a scenario where the non-EU based regulators take a more literal interpretation, then you go down a scenario where non-EU banks have a higher likelihood of passing the modellability test and could have lower capital than an EU bank.”

Boivin of the EBA says asking for both bid and offer prices will help ensure committed quotes are reflective of actual transactions in the market. By having a requirement to evidence prices for both a buy and sell order, it provides an additional safeguard against dud quotes being used to prove modellability.

If a risk factor is super illiquid then it shouldn’t be passing the eligibility test, and if it is trading every day then you don’t need quote data
Jacob Rank-Broadley, Refinitiv

“We don’t think it is in the spirit of Basel rules to allow all one-sided quotes to be counted as verifiable prices, so we require bid-offer to ensure committed quotes remain not too far from potential transactions and can, therefore, be counted as verifiable prices,” says Boivin.

Having both bids and offers also makes it easier for supervisors to check the quotes aren’t based on negligible volumes compared with other orders in the market – a requirement set out in the FRTB. The EBA plans for supervisors to use the bid and offer spreads as a gauge to check committed quotes accurately reflect prevailing market conditions by checking against other bid and offer spreads in the market.

“The onus will always be on the bank to provide a justification for the committed quotes they count as verifiable prices for the purpose of the risk factor modellability assessment,” says Boivin. “Banks are required to document their approach, in particular what they consider a reasonable bid–offer spread. Competent authorities will then have to check and challenge what banks are doing based on the information provided by banks and any other information they may have.”

  • LinkedIn  
  • Save this article
  • Print this page  

Only users who have a paid subscription or are part of a corporate subscription are able to print or copy content.

To access these options, along with all other subscription benefits, please contact [email protected] or view our subscription options here: http://subscriptions.risk.net/subscribe

You are currently unable to copy this content. Please contact [email protected] to find out more.

You need to sign in to use this feature. If you don’t have a Risk.net account, please register for a trial.

Sign in
You are currently on corporate access.

To use this feature you will need an individual account. If you have one already please sign in.

Sign in.

Alternatively you can request an individual account here: