James Hamilton of Econbrowser, in “CDOs: what’s the big deal?” weighs in on the question of what went wrong in the CDO market. He makes a point I haven’t seen stated as clearly anywhere else, namely, some CDO tranches may have been been likely to lose money from the get-go:
The benign view of CDOs is that they represent an important technological improvement that allows for better pooling of risk. I would characterize the main concerns as centering on whether in the process existing financial arrangements have accurately priced aggregate risk.
First let me clarify what I mean by aggregate risk. Some risks are inherently undiversifiable. One can understand that point most clearly with Robert Lucas’s elegant asset-pricing model. Suppose that the entire economy consisted of one big potato farm. All financial assets would then ultimately be nothing more than claims against future potato production, and there is no way they can credibly promise to deliver more potatoes than are actually available. If there is a bad harvest, people will be hungry, and no clever set of financial instruments can possibly insure you against that.
An example of the kind of aggregate risk I have in mind is thinking through what would be the consequences of, say, a 20% decline in average U.S. real estate prices, and what that might mean for default rates.
Let me next clarify what I mean by mispricing of this risk. I am not concerned about whether those who are bearing the aggregate risk (i.e., setting themselves up to be the guy who has the fewest potatoes when times are bad) earn a higher expected return to compensate them for the risk. Rather, my concern is whether they may have invested in assets with a negative expected return. For example, if you purchased an asset with an 85% chance of a 15% return (which you’ll earn as long as a recession does not occur), and a 15% chance of a -100% return (you’re wiped out if it does), then your expected return would be -2.25%.
The specific kind of example that comes to mind is New Century Financial Corporation, which was pushed into bankruptcy from a substantially more modest aggregate shock than the one I am concerned about here. The concern about mispricing is that if loans were extended that should not have been, the magnitude of both the real estate boom and its subsequent bust are amplified substantially relative to what they would have been with accurate pricing of aggregate risk.
But who would be so foolish to have invested hundreds of billions of dollars in extra risky assets with negative expected returns? The logical answer would appear to be– someone who did not understand that they were accepting this risk.
Now for purposes of discussion, Hamilton posits that the buyers of the riskiest CDOs might have ponied up hard dollars for a likely money loser. This is one of the reasons the so-called equity tranches of CDOs (and most asset-backed securities) are commonly called “toxic waste” or “nuclear waste.” They are dubious at best.
Now you might be asking, how could investors be so dumb? The buyers were institutional investors, after all. These guys are supposed to be pros.
There are two possible causes. One is that the sellers of this paper knew it was lousy, but also knew that with high yield paper in high demand, there would be buyers for it (let’s call this the “lipstick on a pig” theory, or Pig).
The second is that while they may have known they were hyping the paper, the sellers also had overestimated its value (this is the “believing your own PR” theory, or PR).
Now I’ve witnessed both Pig and PR in operation. In the early days of the OTC derivatives business (the early 1990s) one of the two biggest dealers was known among professional traders to take customers whenever they could. And we don’t mean take a little, we mean sell customers (generally big corporates) custom derivatives which were guaranteed money losers (and not small numbers either). How could they do that? They were structured to require no cash payment (mind you, that practice wasn’t unique to these bad deals, but you’d be amazed how the due diligence standards drop when no one has to cut a check).
Fooling yourself about the value of merchandise, or PR, also happens all the time. Consider buy-side M&A. Every study every done says most acquisitions fail (the estimates generally range from 60% to 75%) and the single biggest reason for deals not working out is that the buyer overpaid. Now admittedly, there are a lot of perverse incentives at work (everyone does better, including the acquiring CEO, even if the deal doesn’t turn out well). But it is amazing to watch how the participants will tweak the models to make the deal work. They fall under a peculiar spell, and act as if they believe that changing numbers in an Excel spreadsheet will influence reality. The financial model becomes more real than the business it is meant to represent.
Now if PR happens in M&A, with pretty simple math and line items in a spreadsheet that readily be compared with real world measures (target past performance, analyst report, forecasts by industry experts), imagine how easy it is to do with highly complex financial structures and pools of assets that are often heterogeneous?
Mind you, I’m not defending the people who designed and sold CDOs. But the abstractness and complextiy of these deals makes it way way too easy for everyone to con themselves. And here, I imagine the biggest con was in the way the rating agencies and packagers analyzed subprime risk. They used historical subprime data as a basis for forecasts, but those past subprimes had little in common with the paper being originated. Early subprime lending was done on a cautious basis, with a fair amount of borrower scrutiny, and more conservative loan terms (most importantly, lower loan to value ratios). If you have explosive growth in a risky asset category, almost without exception it is done by lowering quality standards. But that adjustment appears not to have been made.
In addition, even if the historical subprime data was applied to a comparable set of borrowers and deals, my impression is that the data was still questionable. The market started only in the mid 1990s and went into retrenchment in 1998, then came back in 2002 and grew rapidly. There is no data on how these loans perform in a serious down cycle like the 1991-1992 recession. And I have a sneaking suspicion that no attempt was made to create proxies for that severe a downturn.