Bailout

Shleifer vs Milken on financial innovation

By Felix Salmon

Yesterday I went to a seminar at NYU where I heard Andrei Shleifer defend his paper on how financial innovation causes crises. At the same time, Mike Milken published an op-ed saying that financial innovation is a wonderful thing, and that what we really need to worry about is too much leverage and too little assiduous underwriting. “Over the long run,” he writes, “the best way to maximize profitability is not to increase leverage, but rather to analyze credit properly”.

This is true, but Shleifer’s point is that it becomes much harder to analyze credit properly if you’re constantly trying to analyze new products without any real-world past history. In the world of credit, innovation generally consists of taking risky stuff, waving some kind of diversification and/or overcollateralization magic wand, and ending up with something which is (a) meant to be safer, and (b) much more difficult to analyze on a fundamental basis: you end up having to use models instead. And models have a tendency to break.

The Milken Institute’s Glenn Yago, in the comments to my initial post on the Shleifer paper, lauds CLOs as wonderful innovations, in contrast to CDOs, on the grounds that they outperformed CDOs during the crisis. But that’s ex post — and the fact is that during the crisis the prices of CLOs dropped sharply, if they were traded at all, entirely in line with what Shleifer’s paper would predict. The point is, says Shleifer, that when people bought triple-A-rated CLOs, they wanted something completely riskless. As a result, the mere realization that there’s a possibility that they won’t get paid out in full is sufficient to see them collapse in price, given the way in which such instruments tend to have been oversupplied by the market.

“The amount of financial innovation is incredibly sensitive to how people think about risk,” said Shleifer. “That’s the point of the model.” In boom times, people think locally, don’t think about tail risk, and pile in to innovative financial products, which banks are happy to pump out in essentially unlimited quantities. In crunch times, people get much more cautious, don’t trust models any more, and flee to the safety of traditionally risk-free products like Treasuries.

This is all true pretty much regardless of the actual product in question. In this rebound, CLOs have obviously outperformed CDOs, since there’s real value in CLOs, while a lot of CDOs are simply going to zero and staying there.

The point is that when Milken says that “credit research should go far deeper than ratings” while at the same time lauding the innovation of collateralized loan and bond obligations, he’s trying to have his cake and eat it. You can do deep credit research on a plain-vanilla unsecured bond or loan. But when you start adding bells and whistles to it, and burying it within much larger and more complex securitization and passthrough structures, deep credit research becomes much, much harder — to the point at which, with CDO-squareds, it’s essentially impossible. At that point, you’re forced to rely on models and Monte Carlo simulations and the like.

In this crisis, model risk went, pretty much overnight, from something which no one spent much time worrying about to something which everybody was terrified of. And so good CLOs sold off alongside bad CDOs. And that exacerbated the crisis. That’s endemic to financial innovation: it’s a real bug in the system. And no one at Milken seems to be willing to admit it.

Full post as published by Macroprudential on April 15, 2010 (boomark / email).