William Black: "There Are No Real Stress Tests Going On"

From naked capitalism on William Black:

There are no real stress tests going on.

1) If you did a real stress test, as Geithner explained them, you wouldn't just have a $2 trillion hole -- you'd impose regulatory capital requirements of 50%. (FYI, the regulators have the power to set HIGHER individual capital requirements based on unusually large risks at a particular bank.)

2) You can't conduct a meaningful stress test without reviewing (sampling) the underlying loan files and it seems likely that the purchasers of securitized instruments (not just mortgages) do not even have the loan file data. Moreover, loss ratios vary enormously depending on the issuer, so even a bank that originates (or has purchased a bank that originates) similar product cannot simply take its own loss rate and extrapolate it to the measure the risk on the value of securitized credit instruments.

3) The regulators are overwhelmed because of personnel cuts (particularly heavy among their best, most experienced examiners that had worked banks that had engaged in sophisticated frauds. Buyouts were common, because more experienced examiners appear more expensive. This isn't true when you consider effecitiveness and productivity, but management didn't care about that. Treat what I write after the colon as hearing from me at my most serious and thoughtful: it is vastly more difficult to examine a bank that is engaged in accounting control fraud. You can't rely on the bank's books and records. It doesn't simply take more, far more, FTEs -- it takes examiners with experience, care, courage, and investigative instincts and abilities. Very few folks earning $60K are willing to get in the face of the CEO and CFO making $25 million annually and tell them that they are running a fraudulent bank and they are liars. FYI, this is one of the reasons why having "resident examiners" never works. The examiners don't even get to marry the natives. They get to worship God's annoited. Effective examination is good for you, but it is very unpleasant, ala a doctor's finger up your rectum. It requires total independence.

So, the examination force doesn't have remotely the numbers or the relevant experience and mindset to examine the largest banks with the greatest problems.

4) As Geithner describes the process, NO ONE can conduct reliable "stress testing." It inherently requires testing everything in every way any and all aspects of everything could conceivably interact. It also doesn't provide any meaningful output that can be operationalized (unless you want to force an enormous rise in minimum regulatory capital requirements, which he obviously doesn't want to do).

5) Examiners certainly can't A) do the stress testing that Geithner describes or B) evaluate the reliability of a large bank's proprietary stress test. If they were serious about constructing reliable stress tests, which they aren't, you'd require their analytics to be made public. You'd have the industry fund independent investigations by rocket scientists chosen by a committee selected by the regulators of the soundness of the analytics. You'd also have the industry fund competitions to rip them apart (a bit like we hire legit hackers to test security by trying to defeat it) and show where they produce absurd results. The geeks would have a field day (that would probably last a decade). There are probably zero examiners that have the modeling skills required to evaluate the most sophisticated stress test models. The concept that there are 100 examiners with these skills, suddenly freed up from all other duties, assigned to CONDUCT stress tests is a lie.

6) It is, however, possible to use even the less experienced examiners to conduct a far more useful examination of the quality and value of nonprime loans. My nightmare scenario which I fear is often true is that A) because the biggest originators of nonprime loans were mortgage bankers, B) because every large mortgage banker that specialized in nonprime loans went bankrups, C) because many of them went into Chapter 7 liquidations and even those that went into Chapter 11's had little incentive to hang on to files on mortgage loans they had sold to other entities -- the loan files on many nonprime loans may no longer exist. (My fervent prayer is that the loan servicers have tapes with copies of the underlying loan files, but I fear that this prayer will not be answered.) Under this nightmare scenario it will be extraordinarily hard to determine loan quality and losses and very hard to foreclose against borrowers that can afford attorneys (admittedly a minority) and that claim fraud in the inducement.

Yes, few examiners understand more exotic products. In my experience, nobody understands all the products. I certainly don't, and if I did I'm sure my knowledge would be out of date within weeks.

The problem is compounded by the fact that understanding how the product is actually used (CDS is a good example) v. how its proponents picture it as being used is essential. Understanding its sensitivity to credit and interest rate risk is well beyond the ken. Understanding the liquidity risks and interaction effects is out of the question.

Examiners rarely know that financial risks are not normally distributed, but have far fatter tails. (Nor do they understand why this makes truncating the VAR a recipe for disaster.) Examiners rarely understand that any econometric analysis undertaken during the expansion phase of a bubble will invariably find the strongest positive correlation with the worst possible business practices (because those practices maximize accounting fraud).

We were, 15 years ago, able to get some strong capital markets people and give them advance training. We sent them out on exams, they impressed the industry -- and they were promptly hired away from us at substantial raises