Jérôme Legras is a managing partner and head of research at Axiom Alternative Investments
If you think the time to worry about US banks’ solvency and capital has passed, think again.
No need to invoke the spirits of the 2008 GFC either. Just ask a Credit Suisse banker what they think of Silicon Valley Bank and the unlikely chain of events that transformed a regulatory loophole on government bonds into the failure of one of the oldest and largest Swiss banks.
With the Credit Suisse fiasco so recent, a proposed overhaul of US bank capital rules has not had the attention it deserves, but maybe launching the plans in the middle of summer was intentional? Purely quantitively, the impact is estimated by the Fed at about $170bn, or 16 per cent of current capital requirements. But this is not what really matters.
In a FT recent op-ed, Sheila Bair, a former Federal Deposit Insurance Corp chair and Basel Committee member, described what she thinks are the key elements of the reform and the rationale for them: the phasing out of internal ratings for credit risk calculations; the inclusion of some unrealised losses (so-called AOCI) for Category three and four banks (so >$100bn in assets); and changes to the operational risk framework to include banks historical operational losses.
Most of these changes have been discussed for years as part of the new Basel package. and make sense. The history of operational losses is a crude approach to assessing company culture risk, being largely a measure of litigation costs, but there is no credible other option available. The new market risk approach will lead to desk-by-desk model validation, more realistic holding periods based on liquidity, and a more consistent replacement for VaR called eVaR (expected losses above a threshold instead of the threshold itself). None of this is controversial.
But there is another change that is both totally unexpected and truly significant, because it has the potential to profoundly change the banking business. It is summarised in the following chart from the Fed:
In practice, there will be two different calculations: banks will compute:
-
a standardised credit risk RWA. This is the Basel standard, it does not use models and the Fed confusingly calls it the “expanded risk-based approach”;
-
a standardised full RWA approach which is effectively . . . the Basel 1 approach from 1988! The old Cooke ratio is resuscitated. Internal ratings-based estimates of credit risk will no longer be used to calculate capital requirements. The so-called IRB models are dead.
What justifies this spectacular jump in a time capsule to 1988?
Blair — and others obviously — offer some explanations as to why banks should not be allowed to model credit risk to calculate capital requirements.
First, it is argued that internal models failed and contributed to the financial crisis. This is the most intriguing of all arguments against IRB models — if only because they were introduced in 2008, a year after Household International nearly failed, the beginning of the GFC. Moreover, people arguing that IRB models caused or contributed to bank failures are usually unable to provide specific examples.
It also often said that, because of IRB models, banks did not have enough capital. This is also difficult to understand. The Basel II accord was calibrated precisely to keep the total requirement in the system unchanged. The requirement is why the exotic formula pictured below has a 1.06 coefficient at the end — a detail that meant the honourable Coventry Building Society was using the wrong calculations for 11 years!
The upshot is that as soon as IRB models became law, banks’ capital ratios did not change much.
A third criticism is that banks have an incentive to tweak models and reduce perceived risk to boost return on equity. Don’t get me wrong, bankers are always extremely tempted to optimise all they can and have a tendency to be slightly greedy (why would they be bankers otherwise). But it is not that simple. The governance of the implementation of the models is usually beyond a front officer’s grasp and skillset. More importantly, backtesting for severe outcomes is required. When models fail to forecast default probabilities, they have to be ditched or changed.
No model is perfect. In particular, loss-given-default and very low default probabilities are hard to estimate: read the EBA’s annual report card if you do not believe me. But, overall, the dirty secret of credit score vendors is that it is actually fairly easy to estimate default probabilities; much easier than to forecast GDP or inflation, which central banks use routinely.
We still need to dig deeper on this incentives issue. Because, you see, the Cooke ratio was not replaced by IRB models just because a French quant liked asymptotic single risk factor models so much she convinced the entire Basel Committee they would be cool to use. It was because of incentives!
Under the Cooke framework, lending money to a local bakery attracted the same capital requirement as lending to an unrated, large, government-owned, profitable utility with near zero leverage. Considering the difference in credit spreads offered on both loans, you can easily guess which loan produced the best ROE for the bank.
Internal models were introduced to stop the incentives in favour of high-risk lending. I was among those who warned against the pro-cyclicity of an approach based on ratings, where the compromise meant complex specifications that can lead to misunderstandings such as models being too lenient in downturns (a feature, not a bug).
Still, the credit market is evidently healthier with IRB models than without.
Estimating the right level of total bank capital in the system is an art, not a science, and we are not equipped with economic models smart enough to give us an accurate target. I don’t criticise the Fed for proposing an increase in total capital requirements. But I have learned, sometimes painfully, that wrong incentives can lead to disasters, in banking perhaps more than anywhere else.
It’s also forbidden in the US to use credit agency ratings to calculate capital requirements. So now what? Will banks simply be asked to navigate with a blindfold and compute capital requirements without any knowledge of the credit risk they take?
Maybe worse: financial crises tend to happen because regulatory loopholes create perverse incentives to add risks that are not adequately priced. In that respect, going back to the Cooke ratio of the 1980s may not be best idea we can come up with today.
Read the full article here