Sunday, April 11, 2010

How financial innovation causes crises

Original posted on Reuters by Felix Salmon:

Nicola Gennaioli, Andrei Shleifer, and Robert Vishny have a great new paper out entitled “Financial Innovation and Financial Fragility”.* It doesn’t break a lot of new conceptual ground, but it’s very thought-provoking, and it helps to codify in a formal way the serious problems with financial innovation. Their conclusion is spot-on, I think:

Recent policy proposals, while desirable in terms of their intent to control leverage and fire sales, do not go nearly far enough. It is not just the leverage, but the scale of financial innovation and of creation of new claims itself, that might require regulatory attention.

The idea here is that financial innovation is, by its nature, inherently and predictably dangerous. If something’s innovative, it’s new. And if something’s new, it’s untested. Meanwhile, a very large part of what we consider “financial innovation” consists of “improving” on existing securities, usually by creating a source of new supply for in-demand securities while also providing some kind of pick-up in yield.

Eventually, a test comes along: the world behaves in a way that no one had expected, and the new securities prove to be less attractive than the traditional securities they replaced. When that happens, demand for them plunges, their price falls dramatically, and enormous losses ensue. This narrative has been played out many times — look at CMOs and junk bonds in the 1980s, or CDOs and money-market accounts more recently. Or look back on eight centuries of financial folly, for that matter.

In order to make the model in this paper work, you just need to make a couple of very reasonable assumptions. First of all, there’s the assumption that investors aren’t perfectly rational; instead, they use what’s known as “local thinking”, and don’t consider every possible eventuality when buying securities. Secondly, there’s the assumption (which isn’t even necessary, it just makes the results stronger) that investors prefer safety over risk. The authors dryly note that it would be possible to model such an assumption by considering “investors who have lexicographic preferences with respect to particular characteristics of investments (e.g., AAA ratings)”. Quite.

The results are predictable. First, you get far too many of the new securities:

When some risks are neglected, securities are over-issued relative to what would be possible under rational expectations. The reason is that neglected risks need not be laid off on intermediaries or other parties when manufacturing new securities. Investors thus end up bearing risk without recognizing that they are doing so.

And second, you get a spike in tail risk:

Markets in new securities are fragile. A small piece of data that brings to investors’ minds the previously unattended risks catches them by surprise, causes them to drastically revise their valuations of new securities, and to sell them in the market. The problem is more severe precisely because new securities have been over-issued.

Finally, if you add leverage to this toxic mix, that only serves to make everything a great deal worse.

More generally, a lot of what’s going on here is that banks are creating “private money”, and that while economists have generally considered that to be a good thing, “security issuance can be excessive and lead to fragility and welfare losses, even in the absence of leverage and fire sales”. Private money, it turns out, is a bit like public money, but not nearly as robust: think of it as a hundred-dollar bill which, unbeknownst to the holder, can occasionally simply self-destruct.

There’s implicit support in this paper for government attempts to intervene in the market during a crisis: under the model, it’s entirely possible that “investors’ valuation of the claim is low not because it is unappealing per se, but because it is not the claim they wanted to hold”. In that situation, intermediaries — banks — can step in to arbitrage the difference, but they often don’t have nearly enough money to do so. If the government steps in with extra liquidity to buy up the now-undervalued securities, that can help to avert a systemic meltdown.

Left to its own devices, a market with financial innovations is very likely to end up harming investors, while still making lots of money for the innovators:

In our model, innovation benefits intermediaries who earn large profits selling securities, but hurts investors, who are lured into an inefficient risk allocation and suffer from ex-post price drops… Investors’ losses from risk misallocation may be so large as to eliminate the social value of innovation altogether.

Realistically, I see very little chance that any financial regulatory reform will do anything to prevent or even slow down the pace of financial innovation. But maybe, if enough investors read and fully grok papers like this, they’ll learn to stay away from securities they don’t fully understand, or which are so new as to be untested in the real world. But I’m not holding my breath.

*Yes, I read this paper the journalistic way, ignoring the mathematics. But dude, it’s Andrei Shleifer. Surely we can trust his math.

The silver lining to synthetic CDOs

One of the more thought-provoking bits of the Shleifer paper on financial innovation is this part of the model:

Optimism about the profitability of the new claim at t = 0 encourages the intermediary to over-invest in an unproductive activity, eventually triggering a loss… Investment in A occurs only if new securities can be engineered, so financial innovation bears sole responsibility for unproductive investment. It can be argued that the expansion in the supply of housing in the last decade was an example of such inefficient investment needed to meet the growing demand for securitization of mortgages.

To put it another way, it was the excessive and irrational demand for collateralized debt obligations which caused all those Miami condos and Phoenix tract homes to be built in the first place.

That makes sense to me, but it raises an interesting question about the damage caused by synthetic CDOs. Here’s Jesse Eisinger and Jake Bernstein, from their investigation of Magnetar:

By helping create investments it also bet against, the hedge fund was actually fueling the market. Magnetar wasn’t alone in that: A few other hedge funds also created CDOs they bet against. And, as the New York Times has reported, Goldman Sachs did too. But Magnetar industrialized the process, creating more and bigger CDOs.

Several journalists have alluded to the Magnetar Trade in recent years, but until now none has assembled a full narrative. Yves Smith, a prominent financial blogger who has reported on aspects of the Magnetar Trade, writes in her new book, “Econned,” that “Magnetar went into the business of creating subprime CDOs on an unheard of scale. If the world had been spared their cunning, the insanity of 2006-2007 would have been less extreme and the unwinding milder.”

Certainly the banks and investors who ended up on the long side of the synthetic CDO trade ended up losing lots of money to people like Magnetar and John Paulson who were on the short side of the trade. But in some ways the Magnetar-driven boom in synthetic CDOs was actually preferable to the boom in RMBS-based non-synthetic CDOs which preceded it.

Think about it this way: both CDOs and synthetic CDOs resulted in losses for investors on the long side. But In the world of CDOs, demand for paper ended up creating a disastrous building boom which diverted resources from more productive activity, skewed local tax revenues, and created the precondition for the wave of foreclosures which is likely to continue for the foreseeable future.

In the world of synthetic CDOs, by contrast, demand for paper just ended up making a bunch of shorts extremely rich: all the other real-world repercussions of the CDO market were actually avoided.

I’m not saying that the world of synthetic CDOs was a good thing. In fact, I’ve explained why I think that it was harmful. But the point that investors started moving from CDOs to synthetic CDOs marked the point at which the housing bubble stopped growing: the move played a significant role in ending the real-world housing insanity. If banks could create synthetic CDOs out of thin air, they no longer needed to encourage subprime originators in the Inland Empire to give $600,000 mortgages to itinerant strawberry pickers, just to keep their channels full.

When talking about credit default swaps, the material out of which synthetic CDOs are made, people often get very upset that you can have more CDS outstanding on a certain name than there is of the underlying instrument. But just think how much better off we would be if the amount of real-money subprime lending had never boomed at all, and if all the financial speculation on subprime mortgages had been confined to synthetic CDOs, all of which referenced a relatively small handful of subprime deals. We wouldn’t have had nearly as much of a housing boom, we wouldn’t be stuck with crumbling suburbs, we wouldn’t have a foreclosure crisis, and we would have invested our money in much more productive things than real estate for most of the last decade.

Of course, it would have been much harder to find people like John Paulson to take the short side of those trades: you needed a bubble to attract the hedge funds who fueled the synthetic CDO boom. But I still think it’s reasonable to consider synthetic CDOs to be less harmful, at the margin, than their real-money counterparts.

All that said, synthetic CDOs did make it much easier for banks, in particular, to take on enormous amounts of highly-leveraged exposure to the subprime market, by holding on to unfunded super-senior tranches. That was a particular problem in the case of Citigroup. When the likes of Citi and Merrill Lynch got out of the moving business and started going into the storage business, they were creating a lot of systemic risk where none had previously existed — and the rise of synthetic CDOs made it much easier for them to do so. As I say, synthetic CDOs were indeed harmful. But were they more harmful than normal CDOs? I’m far from convinced.

No comments: