Turning on tables

The downgrade of billions of dollars of asset-backed securities referenced to US subprime mortgage loans has raised questions about the accuracy of credit ratings. For the new Basel Capital Accord, much of which is based on rating inputs, it has raised doubts over certain aspects of the framework. Duncan Wood reports

The biblical parable of the two builders recounts the contrasting fortunes of two men - a wise man who builds his house on rock and a foolish man who builds on sand. At first, all seems fine, but then the heavens open, the rain falls heavily and the house on the sand collapses. After building parts of their new regulatory framework upon apparently sound rating inputs, the Basel Committee will be able to empathise: mass downgrades of structured credit deals since July has revealed ratings to be a wobbly foundation for calculating bank regulatory capital.

This is a particular problem for the treatment of securitisation under Basel II. Even though the framework's advanced internal ratings-based approach allows banks to calculate their own capital requirements, external ratings are the main starting point for all securitisation transactions not held for trading. The Basel Committee is now reviewing the issue. While it does so, banks may be forced to raise their capital under Pillar II of the new framework - a section of the Basel Capital Accord that is provoking dark mutterings from practitioners (see box).

No proposals for amendments to the securitisation chapter are on the table and no timetable has been set, but practitioners and regulators are already anticipating three possible changes. The first would leave the mechanics of the existing framework untouched, but would raise the risk weights attached to certain rating grades, producing higher capital. The second would scrap the single existing risk-weight table - which applies to all types of securitisation - and replace it with a number of tables tailored to specific sectors, allowing the capital regime to differentiate between benign credit card deals, for example, and Wall Street's current nemesis, collateralised debt obligations of asset-backed securities. Finally, and more radically, ratings could be ditched entirely - a suggestion some experts immediately dismiss as unworkable.

"It's like former UK prime minister Margaret Thatcher used to say: there is no alternative," says William Perraudin, chair of finance at Imperial College in London. "Very few people in the market have enough data to do a bottom-up assessment of risk, so it has to be pinned to the risk indicators people will habitually use - and that, for better or worse, is the rating."

Perraudin has been through this before. While Basel II was being drafted, he worked as a special adviser to the Bank of England, and his technical work contributed to what's known as the ratings-based approach (RBA) to securitisation. It's the most important of three approaches spelled out in the Basel II framework because it applies whenever a securitised asset has been given a public rating - which the vast majority of the market has. He accepts the agencies were "catastrophically slow" to respond to the subprime crisis, but insists their ratings have to play a part in the capital framework, if only because they are a public measure of risk and a language the whole market understands.

One senior US bank regulator agrees: "If we don't use ratings for these structures, one obvious alternative is to use banks' own internal credit risk models. But Basel II decided - and I think it still holds true - that a full credit risk model for some of these deals is premature at this stage. It's important to have some external, independent review to get a better relative rating of these exposures."

There is another alternative. The existing securitisation chapter already contains a ratings-free methodology, known as the supervisory formula approach (SFA). Developed by a staffer at the Federal Reserve Bank of New York, the SFA was used within Basel II as a way of calculating capital for unrated exposures - typically, the risky slices of a transaction. But it was originally touted as a direct competitor to the RBA, says Perraudin: "The Fed was anti-ratings, but the rest of the regulatory community was quite clear you simply couldn't have capital charges that didn't run off ratings. Apart from originators, people didn't have much information about the underlying pools."

That remains the principal objection to broader use of the SFA, although Perraudin concedes it might be worth exploring if wider access to underlying information was obtainable. A New York-based head of one US bank's Basel programme has a more fundamental objection: "To be honest, it's hard to work with. In dissecting a transaction, you end up with as many questions as answers sometimes. I think we need practice working with it before we start amending it."

Unless regulators decide to start from scratch, it means the most likely option will be a change to the RBA. As it stands, the approach is quite simple - each rating grade corresponds to a given risk-weight, which is further determined by the thickness and granularity of the tranche in question (see table A). During the drafting of Basel II, the table led to some fiery exchanges with bankers, who were nonplussed by the regulators' decision to assign different risk-weights to securitisation tranches and bonds with comparable ratings, says Perraudin: "One of the first things said to us by industry - and they said it repeatedly - was, 'If you're charging us this amount for a BBB bond, shouldn't you charge us the same for the BBB tranche of a securitisation?'"

Not according to Perraudin. He argues that credit ratings measure the likely fractional loss that will be suffered by a securitisation's underlying pool of assets, but don't reveal much about the volatility of that figure. So, for example, a pool of credit card loans might lose an average of 6% a year with an annual range of 5%-7%, while corporate loans might have a lower 2% average default rate, but vary from 1%-5%. Even though both might have the same rating, they should attract different amounts of capital, he says.

Ashish Dev, head of the risk practice at consulting firm Promontory Financial in New York, agrees. In his previous job as head of enterprise risk at US bank, KeyCorp, Dev wrote a paper that identified tranche thickness and granularity as key risk factors in securitisation transactions, and also found that mezzanine tranches of securitisations are far riskier than comparably rated loans. Perraudin's own modelling work at the Bank of England tallied with those findings and the two models were condensed into the final risk-weight table.

The RBA table is now set for another spell in the limelight. "Ensuring that capital is appropriate is clearly the most important issue here," says the US bank regulator. "The overall framework still works, but there's an acceptance that we need to look at the risk-weight tables in the RBA to see if they still make sense."

If they are deemed not to make sense, regulators are likely either to bump up the risk weights or develop new product-specific tables. Both approaches have drawbacks, however. The New York-based head of the US bank's Basel programme says raising risk weights in response to a period of unprecedented stress in the structured credit market may be counter-productive in the long run. "You don't necessarily capitalise to the worst thing you've ever seen because you'll be embedding inefficiencies into the allocation of capital for a long time down the road. Practices change, financial systems adjust; from a supervisory perspective, you need to factor that into the picture."

He admits regulators tend to be sceptical of the self-policing promises made by banks in the aftermath of each new financial crisis, but points to the commercial real estate market in the US as evidence that a leopard can change its spots: "We had big problems in that market 15 to 20 years ago, but we changed our practices and we've had good times since then, with very low default rates."

Devising separate risk-weight tables for each product type is the change that initially appeals most, he says - but that, too, comes with problems. "Who's to say the tables we develop tomorrow to solve yesterday's problem will still feel right five years from now?" he asks.

The suggestion that the existing RBA table could now be changed does not go down well with Dev. "One shouldn't fall into the trap of thinking that because the rating is wrong, everything is wrong. If the rating is wrong, it's because the rating alone can't reflect all a deal's risk characteristics, particularly tail risk. It would be a shame to throw the baby out with the bathwater." He argues that, in using ratings as just one input and in making capital requirements more conservative for securitisation tranches versus other comparably rated assets, Basel II got far more right than it got wrong: "Quite a few risks were captured in advance, well before the market as a whole had faced up to them - and recent events seem to vindicate us, if anything."

Of course, structured credit losses have yet to fully play out, which complicates the Basel Committee's task considerably. At this point, many super-senior tranches have not yet been hit by subprime losses, but banks are already taking mark-to-market losses on their holdings. Dev concedes the research underpinning the RBA only focused on credit - or principal losses - and did not cover the type of mark-to-market loss suffered by many institutions recently, but he argues these recent losses result from liquidity risk and mark-to-market accounting, which was not part of the securitisation chapter's remit. "The underpinnings of the RBA provide considerable insight for anyone interested in understanding credit risk in securitisations after the events of 2007. There is always scope for improvement, of course - such as incorporating liquidity risk into the approach," he says.

Perraudin, too, argues the RBA gets a lot right. "I know you could say there's turned out to be a stunning amount of risk in the securitisation market that people didn't realise was there a few years ago, but our models showed at the time that capital should be much more conservative than the banks wanted, and that has been very much justified by what we've subsequently seen."

To prove the point, some practitioners are still aggrieved by the high risk weights attached to mezzanine tranches under the RBA. A London-based credit risk manager at one investment bank complains that if a tranche is downgraded from AAA to B, its risk weight can jump from 7% to 1,250%: "It basically means you have to hold capital equal to your exposure, which is pretty ludicrous."

He hopes the coming review of the securitisation chapter results in capital requirements actually being lowered, but concedes that's unlikely to happen in the current environment.

This article originally appeared in Risk, March 2008 (www.risk.net)BANKERS WARY OF PILLAR II

In the eyes of the banking industry, the success of Basel II now rests upon the least substantial of its three pillars. The first and third pillars are solid, concrete structures that tell banks how to work out their own capital and what information to disclose, respectively. The second pillar is completely different: it gives supervisors discretion over whether to force individual banks to add more capital to their Pillar I numbers and, if so, how much.

A cynic might describe Pillar II as a regulatory get-out clause, enabling supervisors to ignore a bank's own capital assessment if they don't trust the number - something many bankers feel is increasingly likely given the recent turmoil in the financial markets.

One senior US bank regulator phrases it rather differently: "Pillar II was designed to enable us to be nimble, to react to changing markets. We needed to have the ability to say 'I understand what Pillar I says, but the capital number it generates in this instance is not reflecting the risk you're taking and we need to change that'."

What banks fear is that regulators will use this discretion to ensure individual bank capital numbers remain at pre-Basel II levels. The signs are not promising, says the New York-based head of Basel at one large US bank. In their draft rules, the US regulators included a clause that would trigger a wholesale revision of the new framework if it produced a drop in industry capital of 10% or more - a sign they are not committed to Basel's stated goal of making capital more risk-sensitive, he says. "We resisted that very strongly. If capital is going to reflect risk, then the regulators have to accept the possibility that capital could go down, and banks have to accept the possibility that capital could go up. We're willing to do that, but if they're not, then this whole game is not worth playing."

The clause was dropped for the final US version of the rules, but bankers remain pessimistic about how Pillar II will be used in practice, especially while credit markets remain troubled. "Given the stress in the economy, will regulators have the stomach to allow capital to go down? Probably not. But regulators have been very unidirectional in this whole process anyway. Everything they've tried to do, everything they think about is conservatism, covering themselves and therefore raising capital," says a London-based credit risk manager at one investment bank.

The US regulator insists individual banks will be able to achieve capital relief if their risk profile warrants - whether or not markets are under stress. "We try to make our judgements based upon our own assessment of a bank's risk profile. Obviously, the politics shouldn't get into the supervisory judgements," he says.

But banks won't be reassured by the words of one regulator - however senior. National supervisors are in charge of applying Pillar II in their own jurisdiction. Will they each make the same judgements and exercise them in a consistent way? "No chance," says the head of Basel at a large US bank in London.

He says the scant guidance provided so far suggests wide gulfs from one country to the next. In the UK, the Financial Services Authority appears to be moving towards a mechanistic approach in which capital additions under Pillar II would be calculated in line with a set of published thresholds. Australia is adopting a more flexible stance in which only qualitative guidance will be provided, while the Hong Kong regulator will insist on a 10% minimum level of capital, rather than the 8% required elsewhere, and will then use a model to work out what additions are needed, if any. In Japan, banks have been used to a set 8% capital level with no capital additions, and moving away from that will be slow and painful, the banker says. In other words, Pillar II means different things to different regulators.

Still, a little inconsistency is likely to be forgiven as long as banks are able to enjoy some kind of capital relief - and Deutsche Bank's annual results presentation offers one ray of light. During a call with analysts on February 7, the bank highlighted that the introduction of Basel II has boosted its Tier I capital ratio from 8.6% to 9.4%, generating roughly EU^R2.4 billion of spare capital, according to one analyst. However, the New York-based Basel head notes that Deutsche Bank may not be free to deploy that capital elsewhere. "It looks like Deutsche has some kind of benefit from moving to Basel II. But does that mean they're going to be allowed to reduce the ratio back down to where it was? I don't know. I really don't know what's going to happen," he says.