« "Life After 'Rational Expectations'?" | Main | Paul Krugman: Looters in Loafers »

Monday, April 19, 2010

Did Goldman Sachs Use "Computational Complexity" to Hide "Lemons"?

The case against Goldman Sachs is not air tight, and law professors have been busy figuring out how the company might defend itself against the charge that it failed to disclose all the relevant information about the CDOs it was selling. One idea noted by Erik Gerding is that so long as investors knew what assets were in the CDOs they purchased (as opposed to who constructed it), they had all the information they needed to make an informed decision.

As he notes, "The question in this case is whether you should be told that that this gambler betting against you selected the cards in the deck." His original answer was that knowing who picked the cards is unimportant so long as you know what cards are in the deck, but he is now reconsidering that conclusion:

Investigating Goldman: How you can hide a lemon in plain sight, by Erik Gerding: In a previous post, I noted that the SEC's case that Goldman failed to disclose the Paulson & Co hedge fund's role in selecting the collateral might be weakened, because the investors themselves likely were told which assets went into the SEC [I think this should be CDO]. To rehash my metaphor, it isn't as critical to know who selected the deck when you know the cards in the deck. Evidently, I was not alone in this conclusion -- see some of the reactions of other law professors in the NY Times.
Offline, a reader pointed me to a paper that convinced me that I might be completely wrong about this. The paper's conclusions may turn out to have pretty significant implications for this case. Here's the insight, an October 2009 paper by a team of computer scientists and an economist at Princeton argues that the parties that structure a CDO may be able to hide lemons -- that is assets that it knows are subpar -- in the collateral of a CDO through carefully structuring the CDO. The investors in the CDO may find it impossible to detect these lemons even when the collateral is fully disclosed and the investors are sophisticated and have significant computing resources.
Why? The paper (Arora et al., "Computational Complexity and Information Asymmetry in Financial Products") argues that the complex structuring of derivatives can create "computational intractability." In layperson terms, finding the "lemons" can become an inordinately difficult mathematical problem. Unless an investor has unlimited computational power, it may not be able to "solve" the problem and detect the lemons. It's the same problem that occurs with trying to decode computer messages protected with a certain level of encryption. The structuring of the deal functions as a kind of encryption to camouflage the bad assets.
This means that the party that both selects the collateral and structures a complex derivative (like a synthetic CDO) has a potentially insurmountable information advantage over its counterparties. ... (In an amusing twist, the paper argues that "even Goldman Sachs" wouldn't be able to detect the lemons).
What could this mean for the Goldman case? Many things. First, we shouldn't assume that when the investors (or ACA, the collateral manager for that matter) knew what the collateral was that they could easily detect any lemons. Arguments that ... investors need to rely on the proper incentives (or at least the disclosure) of both the party that selected the collateral and the party that structured the deal gain a lot more weight.
Second, how the deal was structured (not just how the collateral was selected) may prove to be crucial. From the SEC Complaint, it ... is unclear if Paulson played a role in structuring the deal. Did Goldman structure the deal to "hide" the Paulson-selected assets[?] Unfortunately, based on the conclusions of the Princeton paper, detecting this hiding is subject to the same intractability problem. Unless there is some "smoking gun" evidence -- e.g. loose-lipped e-mail correspondence, testimony from Goldman employees. Even the absence of a smoking gun doesn't detract from the first point -- that the investors wouldn't be able to detect lemons even if the collateral was fully disclosed to them.
Third, this insight means that an already complex case may require even more expert witnesses -- let's see if the Princeton team gets a call. ...

    Posted by on Monday, April 19, 2010 at 12:33 AM in Economics, Financial System | Permalink  Comments (21)


    Feed You can follow this conversation by subscribing to the comment feed for this post.