Cold comfort for dealers at crunch FRTB meeting

Banks and regulators make little headway on P&L variance test

roadsign frozen
On ice: any agreement between dealers and supervisors on FRTB tests is on hold

A meeting this week between dealers and international regulators has underscored how difficult it will be to address one of the industry's key complaints: the toughness of one of the tests used to decide whether banks can use internal models to calculate market risk capital requirements.

Dealers claim the test would be relatively easy to pass for directional, unhedged portfolios, but practically impossible to pass when positions are hedged.

"Clearly there is an appetite for a solution that would save face and would actually work at this advanced stage of the guidelines, but it is not easy. In some cases the right solution is to remove the test, and that may be entirely unpalatable [to regulators]. It was a receptive audience listening, but they will struggle to address the issues raised," said Michael Sternberg, a managing director in the risk analytics division at Morgan Stanley, speaking at the Traded Risk Europe conference on April 14.

The tests are part of the Fundamental review of the trading book (FRTB), which was published by the Basel Committee on Banking Supervision in January. Members of the committee's trading book group spent six hours talking with industry representatives on April 12 in London, but attendees at the conference said the meeting broke up with no sign of a solution.

Industry concerns focus on the second of two profit-and-loss (P&L) attribution tests a trading desk must pass in order to model its own capital requirements.

The first test, called unexplained P&L, is a comparison of the theoretical P&L produced by risk models and the hypothetical P&L derived from front-office pricing models. The second test divides the variance of unexplained P&L – the difference between the theoretical and hypothetical numbers – by the variance of hypothetical P&L. As calibrated in the final FRTB paper, the threshold for this variance test is 20%. If a trading desk fails either test four times in 12 months, it must instead apply the regulator-set standardised model, the sensitivities-based approach (SBA).

At the meeting on April 12, regulators apparently asked banks how far this variance test ratio might need to be raised to make it practical for banks to comply. Speaking at the Traded Risk conference, Deutsche Bank's head of risk methodology Lars Popken said such a concession would not be enough.

"Unfortunately, the answer is the test is fundamentally flawed; you cannot set a threshold high enough for it to make sense. The test might work well for outright positions, but as soon as you have hedge positions in the portfolio, very small deviations in the P&L between the outright and hedge positions can lead the variance ratio to fail. We demonstrated that in the meeting with very simplified portfolios. There is now a big concern that in order for banks to pass the test, it is better for them not to hedge their risk any more," said Popken.

One presentation at the conference offered support for these claims. It showed the variance test results for a portfolio made up of bought and sold call options over a four-month period. In each month, the long and short positions – tested individually – had variance ratios of 2.7% on average, with a high of 6.1%, well below the 20% threshold. But when the two positions were tested as a single portfolio, the ratio exploded, ranging between 80.6% and 1025.4%. The combined, hedged portfolio failed the test in all four months, which would subject it to the SBA.

The industry this week produced its own quantitative impact study which suggests the capital increase when using the SBA will be several times higher than the Basel Committee's own estimates, especially for certain asset classes. For foreign exchange and equities, the capital requirement under SBA is 6.2 times and 4.1 times the requirement under IMA, respectively, according to the industry study.

Speaking at the conference, Deutsche Bank's Popken also warned a move to the SBA could adversely affect the quality of risk management.

"A lot of the component parts of the FRTB internal model world are actually very valuable – it is about are you conscious of the risk drivers that you don't model very well, how accurate are your valuation models that you use in risk, how good is your valuation data? A number of components, if the worst comes to the worst, would disappear, which means the banks would have lesser incentives to invest in improving their risk models," he said.

In this context, a compromise suggested by the trading book group looks likely to divide opinion in the industry. This is to start with a weaker set of P&L attribution tests, offset by the use of a high standardised output floor to all internal models. In a consultation published on March 24, the Basel Committee indicated it planned to introduce output floors for all internal models at between 60% and 90% of the standardised approach. The trading book group apparently suggested adopting the 90% upper limit as a stopgap while banks and regulators consider how to implement the P&L attribution tests.

"That floor, say 90%, is a deal-breaker, because no-one would bother doing internal models," said Etienne Varloot, head of global markets regulatory strategy at Natixis, speaking at the conference. "One approach is to say the P&L test is stopping you from getting validated, so let's get a P&L test that is easier, then come back to the floor. If we can get a bit of leeway there, it is a better trade-off."

But he warned that official participants at the trading book group meeting were not the Basel Committee's decision-makers. At the top level, he felt regulators had a strong view that "the less modelling, the better", which would make any further FRTB compromises difficult.

  • LinkedIn  
  • Save this article
  • Print this page  

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: