When The Junior Analyst And The VP Disagree About The Model, Who Is Likely To Be Right?*

Publish date:

Carrick Mollenkamp is a great reporter who currently owns one of my favorite niches: finding insider moles to bring to light behavior at big financial companies that is ambiguously squicky. Today he's got one that's close to my heart, and here it is: a junior analyst at Deutsche Bank disagreed about a technical modeling question with his VP.

Wait, what?

Well, here:

At a time when mortgage-backed securities were imploding and customers were fleeing the market, a junior analyst at Deutsche Bank AG protested when he was asked to alter the numbers in a spreadsheet to make a Deutsche security look less risky to ratings agencies, according to a person with knowledge of the matter.

The analyst, this person said, was asked by a mid-level Deutsche executive in late 2007 to make it appear that the investment would produce more cash than the bank actually expected at certain time points. ...

[Ajit] Jain had studied at the Indian Institute of Technology in New Delhi and joined Deutsche in June 2006, according to employment records kept by the Financial Industry Regulatory Authority. He joined the New York office in September 2007, when the CDO Group was struggling to find 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, according to people familiar with the matter.

These spreadsheets were cash flow models for some of Deutsche's CDO deals, which Deutsche gave to ratings agencies to rate those CDOs. They were complicated. How complicated? Here is a lovely detail:

Those spreadsheets were often so large and complex they could take several minutes to open on a computer, according to a person familiar with them. The spreadsheets involved fiendishly complex arrays of inputs and sophisticated calculations, involving everything from the default rates of the mortgages that backed the CDO, to when borrowers would pay off their loans. But one purpose of the spreadsheets was simple: to estimate how much cash the CDOs would generate at certain time points.

I had a computer like that once too! (Er, now.)

This sounds very unpleasant but I'm not sure. You could have two opposite theories:
(1) There was a right thing, which Jain did, and a wrong thing, which his VP told him to do, and then there was a cover-up; or
(2) There were two plausible things, and the VP thought his was right, and Jain thought his was right, and Jain was kind of a prima donna about it, and then there was an abundance-of-caution investigation that concluded that the VP was right or at least plausible (and more senior!).

Obviously (1) is a news story and (2) isn't. In some parts of the world this is called publication bias. Remember these numbers were cash flow projections - there was no "true" and "false" (I mean, there is, but only in hindsight, and my guess is that both sets of projections were "false"). There's only more or less plausible, more or less in accord with stated assumptions, more or less in accord with other projections used by DB. There is a lot that you could do that would look fraudulent - say, take internal projections that DB used to manage its risk and massage them for S&P, or claim to use certain assumptions but actually use different ones in modeling the cash flows. And there are other things that you can do that would look, just, messy - say, have one person assume one set of inputs and another assume another.

When Carrick Mollenkamp smells smoke there's often fire and certainly this sounds like naughty things went on. "Alter" is always a worse word than, like, "change," and "mak[ing] it appear that the investment would produce more cash than the bank actually expected" is sort of a no-no though who is "the bank" doing the expecting here? But my instinctual sympathies in the string of CDO scandals tend to be with the CDO structurers because things are just generally messy and there are often a lot of plausible ways to come up with projections for an obviously uncertain future.

And then you get hit by the publication bias. Lots and lots of emails get sent at investment banks. If a structuring team sends 1,000 emails saying "good job, everything checks out" and one dumbass analyst says half-jokingly "man, some of these bonds are dogshit," guess which email gets read at your criminal trial? CDO structurers, of all people, live statistically and stochastically - prosecutors and juries tend to take the worst facts as the most representative. You fuck one sheep!

Here, lots of people on two continents worked on these models. Let's say there were 20 of them. One thought something smelled kind of funny. Nineteen thought everything was fine. (Or are lying to you! Totally possible.) The one who has complaints gets the press. In most industries, when the junior guy on the team gives his perspective and is overruled by his boss, he either shuts up or eventually gets fired. In banking, he gets an investigation run by an outside law firm. This is why.

Deutsche Analyst Sounded Alarm When Asked to Alter Numbers [ProPublica]

* Obviously the analyst, right?