Models behaving badly

| January 20, 2012 | 1 Comment

By Photo-Fenix on Flickr

So it’s not just American banks playing fast, loose and fraudulent with figures. There’s a new ProPublica article out that raises serious questions about Deutsche Bank, picking up on a theme that was first reported on by Zero Hedge and by Dealbook, (who dropped it like it was hot), back in 2009.

The article highlights a junior Deutsche Bank analyst who was allegedly told to change the numbers in the financial models used to estimate how much cash would be generated by the mortgage-backed bonds that DB packaging up as CDOs. Ajit Jain was supposedly asked by mid-level management to pretty up the figures so that rating agencies would assign the CDOs a coveted AAA rating, which would allow Deutsche to sell them to a much broader base of investors.

Within a short time of his arrival, according to three people familiar with the matter, Jain raised questions about whether spreadsheets were being improperly altered. His complaints went to senior levels within Deutsche, including its legal and compliance departments [...].

And will you be surprised to learn that, in the end, no external scrutiny came of Jain’s complaints? No SEC investigation so far and the only related lawsuit, brought by M&T Bank, was settled and the court documents sealed. But regardless of the efforts to hide the truth, they cannot stop the outlines of the full picture from emerging.

There has been one central, unanswered question since the start of the financial crisis. “How could the ratings agencies get it so wrong by assigning the absolute safest credit ratings to investments that would turn out to be junk?” I’ve written previously about how captured the ratings agencies were by their bank-structurer clients. The incentive model alone goes a long way toward explaining the agencies’ failures. But were they also flat-out lied to on top of that, as is implied by the ProPublica piece? Unfortunately this kind of behavior seems all too plausible.

Aside from the question of whether there has been blatant abuse of these financial models, does the industry rely too much on them in the first place? What is a financial model, exactly? An extremely complex spreadsheet that combines some known data with a great deal of essentially unknown data for which various assumptions have been made.  The assumptions would ideally be based on historical patterns, but as we all know past performance is no prediction of the future. And because banks typically make up their own proprietary models (trade secrets, don’t you know) it’s difficult to find a neutral assessment on how dependable or robust they are.

Satyajit Das, prolific author and derivatives expert, discusses the malleability of financial modeling in his recent interview with Robert Johnson of the Institute for New Economic Thinking. The entire half-hour segment is wide-ranging and really worth your time but, on models, Das had this to say (emphasis mine):

People like me read the Brady report [from President Reagan's taskforce on the 1987 Black Monday crash] with great care, trying to work out what they knew. And we got very, very skeptical. But at the same time what we found is the academics went the other way. They basically said, “Oh no, it was just a faulty model and we can develop more elaborate models. And the problem, that I kept saying to these people, is: the more elaborate the model, the more the assumptions, the more the problems, the more the breakdowns and the more fragile it is.

 [...]

I almost found that the modeling, etc., was a way of dressing up certain things. And I always remember clients who actually would say to me, you know, what does this mean or whatever, and I would explain it, as a consultant I would help them look at it. And they were in awe of these models and I kept saying, “Forget about the models, forget about all of this. Why do you want to do it? What does it do? What economic purpose does it serve for you?” And if that makes sense, well OK, then we’ll fit the numbers around that.

Financial models are, by their very nature, extremely open to manipulation. How can you tell if a model’s been manipulated? Don’t even waste your time trying. A better question is: do you trust the person or the firm that has made it in the first place?

————

 

 

 

Tags: , , , , , , , , ,

Category: Banking

«
»