Risk managers defend IRB against Tarullo criticism

The internal ratings-based (IRB) approach has been in intensive care for the past year. Now, one of the world’s most senior supervisors wants to pull the plug, but bank risk managers argue that – with some therapy – it can still lead a full and useful life. Fiona Maxwell reports

william-coen-0043

The models banks use to calculate capital for their loan books are complex and opaque; they can be gamed, offer little insight into bank balance sheets, do little for market discipline, and are both pro-cyclical and blind to tail risk – or so says Daniel Tarullo, one of seven Federal Reserve Board governors.

In a speech on May 6, Tarullo argued the 13-year old internal ratings-based (IRB) approach fails on pretty much every level, and said the industry would be better served by a combination of three newer prudential tools – Basel III's leverage ratio, a capital floor based on standardised measures, and regulatory stress tests.

Not so fast, say bank risk managers. They accept the IRB needs some work, and agree stress tests have a role to play in post-crisis supervision, but insist the arguments that led regulators to adopt it still hold.

"The logic behind internal models is that they are closer to the real risks. Using them on a daily basis helps banks to better manage those risks," says Jacques Beyssade, chief risk officer at Natixis in Paris.

Louise Lindgren, Stockholm-based chief risk officer at Sweden's fifth-biggest lender, Länsförsäkringar Bank, echoes that. "I'm very fond of stress tests, but it's difficult to immediately stress every single counterparty. It works more on a consolidated, portfolio level than the IRB approach – you can take segments, you can take geographies, you can take certain parts of the portfolio."

IRB is an important part of the framework and will continue to be unless the committee determines there is a better alternative

One London-based senior risk manager also defends the approach: "Why target IRB when there are so many models out there? I thought the comment seemed somewhat strange – frankly, I would defend the IRB, and I would also defend all internal models," he says.

The IRB was a revolution when proposed in 2001, offering banks the freedom to model the risks arising from huge portfolios of residential, small business and large corporate loans, and to hold capital accordingly rather than being tied to standardised risk-weights. It has three components. Under the advanced IRB, banks are allowed to calculate the chance of a loan going bad, the size of the exposure at that point, and the amount of the resulting loss. The foundation IRB only gives banks freedom to work out the first of these, but both approaches are subject to certain conditions – minimum amounts of historical data when working out probabilities of default (PD) for example.

It was explicitly presented as a way of reducing capital requirements – an incentive designed to attract small and medium-sized banks, as well as the biggest lenders. While total credit risk capital would shrink, regulators hoped that tying these reserves to accurate measures of risk would force banks to police themselves – taking a lot of risk would tie up prohibitive amounts of capital.

The Fed's Tarullo called this "badly misguided". The crisis showed up weaknesses in market and credit risk models, and subsequent years have revealed huge differences in capital numbers from one bank to the next (Risk April 2012). A study of bank capital models published by the Basel Committee on Banking Supervision in July 2013 – part of its regulatory capital assessment programme – used a benchmark portfolio to show the capital requirements calculated by individual banks could vary by as much as two percentage points from a 10% capital ratio benchmark.

Bank risk managers know all of this, but insist the IRB should be rebuilt, rather than replaced – and many regulators agree.

William Coen, the Basel Committee's secretary general, says there are no plans to scrap the approach: "IRB is an important part of the framework and will continue to be unless the committee determines there is a better alternative. We are doing a lot to address some of the deficiencies and some of the factors that lead to the wide dispersion in results."

And the industry is pitching in. Barbara Frohn, senior adviser at the Institute of International Finance (IIF) – seconded from her position as director of public policy at Santander – says banks are working on transparency and harmonisation, but she also takes a swipe at one of the common criticisms of capital modelling.

"Most banks by now are realistic, and accept that some things may need to change to restore the credibility of a capital framework based on internal models. One of the angles is transparency, and we are also exploring with our members whether in some areas, their modelling practices can be harmonised. But we have to remedy the real, rather than the alleged, shortcomings," she says.

That is a reference to Tarullo's claim that models can be gamed by deliberately selecting data or parameters to suppress risk and capital. One obvious way to do this is to feed the model a dataset that starts just after a major meltdown – a Swedish bank, for example, might base its residential mortgage model on data running from the mid-1990s, after the country's devastating credit crisis. Risk managers bristle at this notion.

"It's not so easy to choose the length of time. You have to use a certain number of years, depending on the parameter, and on top of that you have a margin of conservativeness on the actual values. It's not so easy to game the system," says Länsförsäkringar's Lindgren.

The IIF's Frohn says much the same: "It's not true that banks can just game the system, nor would that benefit them. Speaking from my time as a global head of internal model validation, all models undergo extensive scrutiny – not just at inception but on an annual basis – by home and host supervisors. Such model approval processes sometimes take up to two years. In our case, every single model choice and all data inputs were monitored, checked and discussed extensively."

An innocent way of explaining the dispersion is to say banks are making legitimate modelling choices, but which produce very different outcomes.

"The interpretations and the range of data inputs provide banks with an extraordinary amount of discretion," says Simon Samuels, a London-based former banking analyst at Barclays. "When you really get into the weeds of it, you realise there's an awful lot of opportunity for banks to interpret the number differently. And I think, at its core, this is why confidence has been sapped."

It could also be the way to rebuild confidence. In its annual report, published at the end of June, the Bank for International Settlements – home to the Basel Committee – makes a number of suggestions on how to improve the comparability of models. These include additional guidance from regulators, harmonisation of national requirements, the use of capital floors and constraints on IRB parameter estimates. All would have the effect of limiting bank discretion, and some national regulators have jumped the gun – in mid-2012, the UK prudential regulator floored the loss-given-default component of the IRB at 45% for sovereign exposures, while the Swedish regulator has imposed risk-weight floors for mortgage lending (Risk November 2012 and Risk July 2012). Similar action has also been taken by regulators in Belgium, Denmark and Switzerland.

In other words, the cost of saving the IRB is likely to be greater standardisation, but Eduardo Epperlein, global head of risk methodology at Nomura in London warns against going too far: "There's a misconception that if you replace internal models with so-called standard models it will be better; perhaps due to a perception that the word ‘standard' means it will be more stable and conservative. That's not necessarily the case."


BOX: "It's going to require a dramatic dumbing-down ..."
Risk managers may think Daniel Tarullo went too far in suggesting the internal ratings-based (IRB) approach should be canned, but the Federal Reserve Board governor is not alone in his concerns.

"I am grateful that somebody powerful and in a very senior position, like governor Tarullo, has begun to speak candidly on a topic that frankly hundreds of my peers have been talking about quietly for years, and saying ‘Why is so much going into this exercise when it has no statistical validity?' says Donald van Deventer, chief executive officer and founder of risk management solutions software vendor Kamakura Corporation in Honolulu, Hawaii.

Simon Samuels, London-based former banking analyst at Barclays, agrees a change is needed. Abandoning the IRB approach may be a bit extreme, he says, but a revised version may be unrecognisable from the one used today.

"It's going to need a dramatic dumbing-down of the process. It will require a ton of the discretion that's currently allowed by Basel to be removed so we end up with something much closer to the current standardised approach," says Samuels.

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 info@risk.net or view our subscription options here: http://subscriptions.risk.net/subscribe

You are currently unable to copy this content. Please contact info@risk.net to find out more.

Credit risk & modelling – Special report 2021

This Risk special report provides an insight on the challenges facing banks in measuring and mitigating credit risk in the current environment, and the strategies they are deploying to adapt to a more stringent regulatory approach.

The wild world of credit models

The Covid-19 pandemic has induced a kind of schizophrenia in loan-loss models. When the pandemic hit, banks overprovisioned for credit losses on the assumption that the economy would head south. But when government stimulus packages put wads of cash in…

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