Capital sums set to drive FRTB desk decisions

Using Volcker desk structure may hurt model approval chances, banks say

Search for optimal FRTB desk structure will be tricky

Volcker rule compliance offers banks a place to start when deciding on a desk structure for incoming trading book capital rules – but it will only be a start, practitioners warn. Capital efficiency and other incentives could see the industry opt for larger, more diverse desks in some cases, and smaller, more focused ones in others.

Under the Fundamental review of the trading book, finalised in January, banks that want to calculate capital requirements using the internal models approach (IMA), must apply for supervisory approval on each of their desks separately. Desks that fail either of two key tests must revert to the standardised approach, which would generally result in a higher capital charge.

Banks that have had to comply with the US Volcker rule since it took effect in July 2015 are considering using it as a template for their FRTB preparations. The rule requires reports to be filed by individual trading desks, to show they have not breached the Volcker prohibition on proprietary trading – but these groupings may include products that would cause the desk as a whole to fail one of the IMA tests.

"I think [Volcker] is definitely the starting point. There is currently an evaluation process going on in the institution to figure out what the right desk structure is. If we have portions of books where we do not think we are going to pass the test, we will think about separating them," says a regulatory policy adviser at a major US bank.

Anke Raufuss, a partner at McKinsey, says all of the banks she has encountered are assuming they will use the Volcker desk structure for FRTB. But she adds that banks may then ring-fence certain products from each desk's portfolio if the data available for those items is insufficient to pass the test to qualify for IMA.

"[Banks] would rather remove [products] from one desk in order to increase the likelihood of getting the remaining products on the desk under the internal model approach," says Raufuss.

Desk tests

The first test a desk must pass is a profit and loss (P&L) attribution test. This requires banks to compare their hypothetical P&L, which comes from the front-office for pricing systems, with a theoretical P&L, based on internal model-generated risk measures. If the values differ significantly then it is seen as a weakness in the measurement models used by the desk.

The results are measured using two metrics. The first is the mean of the difference between the theoretical and hypothetical P&L – also known as unexplained P&L – divided by the standard deviation of the hypothetical P&L. The difference between theoretical and hypothetical has to be within 10%.

The second metric is the variance of the unexplained P&L divided by the variance of the hypothetical P&L. The threshold applied to this calculation is 20%. These tests are conducted and reported monthly to the regulator. If four breaches occur on either test within a period of 12 months then the bank loses IMA approval for that trading desk and will have to fall back to the standardised approach.

We need to find a structure that is a best fit for them and can give them the most stable capital charge
Britta Achmann, RBS

In an interim quantitative impact study in November 2015, the Basel Committee estimated that capital charges under the standardised approach would be two to three times higher than the IMA. However, industry estimates suggest the capital requirement could be as much as 4.6 times higher for some asset classes.

Britta Achmann, head of market and counterparty credit risk capital at Royal Bank of Scotland (RBS), described the process of starting with Volcker desks and then making them fit for purpose under FRTB in a Risk.net webinar in March. It means running the P&L attribution tests on each Volcker desk. If a desk proves too volatile under the tests, with the risk of switching between the IMA and standardised approaches, then a bank might change its mix of products and risks.

"It's really a business optimisation problem, because as a business we need to find a structure that is a best fit for them and can give them the most stable capital charge," said Achmann, who stressed she was offering personal opinions in the webinar, not discussing RBS policy.

The final structure some banks employ may be quite different from existing Volcker definitions.

"I don't think alignment is necessarily the right way because the purposes defined are different. Obviously, operationally, the fewer differences the better, but we need to recognise these [desk definitions] are for different purposes," says a head of risk analytics at a major UK bank.

Backtesting and scope

One risk manager at a European bank says Volcker desks were set at a relatively high level. For FRTB compliance, there could be advantages to adopting more granular trading desk definitions.

"[A larger desk] gets more of a diversification effect and it might be easier to run the P&L test-submission process. But if you don't pass, you face a bigger capital hit, so you might prefer more granular desks," says the risk manager.

However, banks must also carry out backtesting to pass the IMA approval process. If the P&L values are higher than the value-at-risk figure generated by the internal model, then an exception occurs. If, within 12 months, 12 exceptions occur at the 99% confidence level VAR, or 30 exceptions at the 97.5% confidence level, the desk could be stripped of IMA approval. More granular trading desk definitions could result in outlier data points having a greater impact on backtesting results, because there will be fewer total trades per desk.

"If you narrow down the scope of your desk too much, the chance that you are going to fail on backtesting increases. So if I defined a desk with one product, like credit default swaps, the chance that the desk failed on backtesting is much higher than a diversified credit desk because there is no offsetting benefit in the first desk structure," says a London-based risk and regulation consultant.

Moreover, VAR models are based on historical data. A smaller dataset may include fewer examples of historic volatility, resulting in more exceptions when compared to the P&L values. Idiosyncratic risks would also have a greater impact on model failure under a more granular structure.

"Models, by definition, work on homogenous portfolios. If you create too granular a desk, the idiosyncratic risk will go so high that the models will just fail all the time," says the UK bank head of risk analytics.

An alternative would be to pool the risk factors across a particular asset class into one desk. By minimising the number of desks per asset class, this could reduce the impact of individual risk factor events to limit the danger of model exceptions.

There is a general consensus on the need for more analysis to fully understand and optimise FRTB desk structures before the implementation of the new rules in 2019.

FRTB – a timeline

• Risk management

Banks make new push on FRTB’s P&L test

• 08 Mar 2018

• Regulation

Basel delay does not ensure global FRTB consistency

• 18 Dec 2017

• Regulation

Basel rules risk fragmentation after key compromise

• 07 Dec 2017

• Risk management

ECB tells IMA banks to apply before rules are complete

• 19 Oct 2017

• Risk management

FRTB: industry pushes to use own quotes in risk factor modelling

• 08 Sep 2017

• Basel Committee

EU lawmakers consider extending FRTB deadline

• 24 May 2017