Bafin's Hufeld: op risk modelling 'almost impossible'

AMA can go, but other models will stay, Felix Hufeld tells Risk.net

ofc-cr105-cover

In this issue, Germany's top supervisor says the modelling of operational risk capital has proved "almost impossible" at the standards of accuracy required by Basel II's advanced measurement approach (AMA) – but he also reassures the industry that capital models have a part to play for other risk types.

Regulators met in New York on December 1-2 to discuss the issue and are expected to propose that the AMA be killed off and replaced with a new, standardised approach.

In an interview that took place just prior to that meeting, Felix Hufeld, president of German supervisor Bafin, says: "Just looking at operational risk, frankly we probably wouldn't mind being more restrictive and possibly abandoning the AMA approach. But I wouldn't say the same thing for credit risk. Over the whole range of risk categories we very much support using internal models, just in a constrained fashion."

In October, Risk.net revealed the Basel Committee on Banking Supervision was leaning towards scrapping the AMA. The committee consulted on simpler methods of calculating operational risk in October 2014. Aspects of its revised standardised approach are expected to form part of a new standardised measurement approach, which all banks would have to follow.

The AMA requires banks to calculate their operational risk with a 99.9% confidence interval over a one-year period, but the industry's models proved unable to cope  with the wave of costly post-crisis conduct settlements. As an illustration of the increase in losses, when the Basel Committee conducted a 119-bank loss data collection exercise in 2008, it found an aggregate annualised loss of €5.1 billion ($5.5 billion) in the category that would contain most legal and regulatory settlements – that is, roughly half the amount Bank of America Merrill Lynch agreed to pay in a single mortgage-related case in January 2013.

US regulators had been particularly cautious about allowing their banks to use the AMA, insisting on a years-long 'parallel run' of the models alongside a simpler approach. Approval was granted for eight banks in early 2014. Banks elsewhere have been using the AMA for longer, but ultimately appear not to have satisfied regulators that their modelled numbers can be trusted.

"It was almost impossible to come up with a sufficiently robust and broadly supported approach at an AMA level to capture operational risk," Hufeld says.

Bankers fear any standardised approach will not be risk-sensitive enough. If a reduction in risk is not reflected in the capital numbers – the argument goes – banks will be discouraged from investing in operational risk management.

For market and credit risk models, regulators are being more selective, continuing a trend that started in 2009, with Basel 2.5's quick-fire overhaul of trading book capital rules; banks were told they could no longer model capital requirements for securitisations, while for correlation trading portfolios models were allowed, but capital levels could not go below 8% of the number calculated by a cruder, standardised approach.

The same solution is now being applied more widely. Regulators are consulting on revised standardised approaches for market, credit and counterparty risk, as well as a framework of floors that will ensure capital savings are limited.

Click on the links below to read the articles from this issue.

 

  • LinkedIn  
  • Save this article
  • Print this page