Being a Chicken Is a Low-Risk Business

Chris Dillow speaks:

Stumbling and Mumbling: Chebyshev and chickens: Mr Viniar can't be that stupid. He knows financial returns are not normally distributed, but have fat tails; extreme events are more likely than a bell curve suggests. More likely, he's thinking of Chebyshev's inequality. It says that in any data sample, no more than 1/k2 of the values are more than k standard deviations away from the mean. On this view, a 25 standard deviation event is a one in 625 probability.

Does this make Mr Viniar look less stupid? No, however we cut it, [Goldman Sachs's] Mr Viniar is talking wibble. He just under-estimated risk.

There's one very stupid way of doing this. Imagine you're a chicken. Every day, the farmer feeds you. After a while, you figure: "My returns from the farmer are pretty stable, as I seem to get roughly the same amount of corn every day. Being a chicken  is a low-risk business." The following day, the farmer breaks your neck.

Any hedge fund who took default risk (say by holding CDOs) and boasted about its Sharpe ratio based on post-2003 returns would have done much the same as this; you can't measure default risk by looking at past returns. I had assumed that no-one was stupid enough to fall for this trick; one reason why I was underwhelmed by Taleb's The Black Swan - which laboured similar points - was that I thought he was just stating the bleedin' obvious. But perhaps I was wrong.

Not only were they massively exposed to subprime risk, but they didn't know that they were massively exposed to subprime risk, and they don't seem now to be able to coherently explain why they didn't know that they were massively exposed to suboprime risk.

Comments

[David Lloyd-Jones]

Chris says, quite rightly, that Teleb was just "stating the bleeding obvious." Exactly because what Taleb says strikes me as obvious, I have raved about him ever since I found him, including in one or two posts here.

I was seven years old when I figured out that everything written down is contradicted by something else written down, so that the chances cannot be more than 50-50, at the very best, of anything you read being true.

In the years since I have found it extremely unusual for anything I read to be even faintly plausible, so when I find somebody like Taleb writing down the obvious, it is always something I treasure.

[Benjamin]

David, apparently there is a 50-50 chance that that Taleb is speaking hogwash.

[archer]

First, recall that Viniar is merely the CFO of Goldman, not a trader. All you have to have to be a CFO is four-function math.

Second, I still prefer John Dizard's theory in the FT earlier this week: this is a simple agency theory problem. Normal distributions apply in markets most of the time, except when they don't, which seems to happen roughly every ten years. But you get to charge big fat annual fees. So money managers have nine years of high income and a tenth year resume put.

But third, a lot of these guys still drink their own Kool-Aid. Robert Merton, the Nobel prize winner of LTCM fame (or more accurately, disrepute) still thinks their models were correct and they were victims of a perfect storm. No responsibility for the fact that they were short vol in every market known to man (and therefore taking bets that were highly correlated under the right circumstances) and had such massive positions that even in the absence of a crisis they'd have trouble working them down. Oh, and their models (to the extent they had them, they had started trading big in markets in which they had no experience, like risk arb) had at most 10 years of data.

[Roger Bigod]

The observation of "fat tails" was around a long time before LTCM. I recall reading this in a discussion of option pricing not long after Black-Sholes became popular. The authors noted that many commercial options dealers used a correction factor for OTM options, because empirical data suggested more extreme data than the Gaussian curve predicted. It's also implicit in Mandelbrot's work. IIRC, he used a data set of market prices as an example of a fractal time series.

[anne]

http://economistsview.typepad.com/economistsview/2007/08/paul-krugman-wo.html

August 17, 2007

Paul Krugman: Workouts, Not Bailout
Edited by Mark Thoma

Paul Krugman has a proposal to deal with the consequences of the mortgage market meltdown:

NY Times: ...According to data released yesterday, both housing starts and applications for building permits have fallen to their lowest levels in a decade, showing that home construction is still in free fall. And if historical relationships are any guide, home prices are still way too high. The housing slump will probably be with us for years, not months.

Meanwhile, it's becoming clear that the mortgage problem is anything but contained. ... Many on Wall Street are clamoring for a bailout — for Fannie Mae or the Federal Reserve or someone to step in and buy mortgage-backed securities from troubled hedge funds. But that would be like having the taxpayers bail out Enron or WorldCom when they went bust — it would be saving bad actors from the consequences of their misdeeds.

For it is becoming increasingly clear that the real-estate bubble of recent years, like the stock bubble of the late 1990s, both caused and was fed by widespread malfeasance. Rating agencies like Moody's Investors Service ... seem to have played a similar role to that played by complaisant accountants in the corporate scandals of a few years ago. In the '90s, accountants certified dubious earning statements; in this decade, rating agencies declared dubious mortgage-backed securities to be highest-quality, AAA assets.

Yet our desire to avoid letting bad actors off the hook shouldn't prevent us from doing the right thing, both morally and in economic terms, for borrowers who were victims of the bubble.

Most of the proposals I've seen ... are of the locking-the-barn-door-after-the-horse-is-gone variety: they would ... have been very useful three years ago — but they wouldn't help much now. What we need at this point is a policy to deal with the consequences of the housing bust.

Consider a borrower who can't meet his or her mortgage payments and is facing foreclosure. In the past, ... the bank that made the loan would often have been willing to offer a workout, modifying the loan's terms to make it affordable, because what the borrower was able to pay would be worth more to the bank than its incurring the costs of foreclosure and trying to resell the home. That would have been especially likely in the face of a depressed housing market.

Today, however, the ... mortgage was bundled with others and sold to investment banks, who in turn sliced and diced the claims to produce artificial assets that Moody's or Standard & Poor's were willing to classify as AAA. And the result is that there's nobody to deal with.

This looks to me like a clear case for government intervention: there's a serious market failure, and fixing that failure could greatly help thousands, maybe hundreds of thousands, of Americans. The federal government shouldn't be providing bailouts, but it should be helping to arrange workouts. ...

The mechanics ... would need a lot of work, from lawyers as well as financial experts. My guess is that it would involve federal agencies buying mortgages — not the securities conjured up from these mortgages, but the original loans — at a steep discount, then renegotiating the terms. But I'm happy to listen to better ideas.

The point, however, is that doing nothing isn't the only alternative to letting the parties who got us into this mess off the hook. Say no to bailouts — but let's help borrowers work things out.

[bakho]

Were these instruments really so repackaged that the risk was hidden from fund managers? I find that hard to believe. (Then again, some of these people believe a lot of stupid things like revenue generating supply side tax cuts.)

I am hearing that a lot of Japanese investors got taken because they were borrowing money at lower interest rates to buy higher interest rate packages in the US. Is this legal in Japan? Are we going through cycles where nothing bad happens, so we relax the rules in a way that promotes bad behavior?

Maybe now is a "teaching moment" to educate the American public about the benefits of good regulation?

[kid bitzer]

isn't this flatly fallacious?

"no more than 1/k2 of the values are more than k standard deviations away from the mean. On this view, a 25 standard deviation event is a one in 625 probability."

the first sentence says that *no more than* one in 625 events will be a 25-s event.

for all distributions. no matter how flat or otherwise weird.

the second sentence says that a 25-s event is actually a 1/625 likelihood, in the distributions under discussion.

but chubby checker's theorem says that 1/625 is the absolute *maximum* likelihood of such an event, surveying over the whole universe of distributions. not that it *is* the likelihood, for every given distribution.

what dillow concludes from that is something far, far stronger.

[Slocum]

"I had assumed that no-one was stupid enough to fall for this trick; one reason why I was underwhelmed by Taleb's The Black Swan - which laboured similar points - was that I thought he was just stating the bleedin' obvious. But perhaps I was wrong."

But the problem, of course, is that investment strategies that fail to take in account unlikely (but far from impossible) events will do better, most of the time, than those that conservatively factors in unlikely events.

Your strategy may have, say a 25% chance of going bust in the next decade, but that's OK for a fund manager -- because a 75% chance of being lucky enough to run the table and pocket enormous sums of cash and be considered a prince of Wall St is a good bet. And then even when the black swan event does happen, you've probably already pocketed a lot of cash AND you really won't be blamed anyway, since plenty of others will be in similar straits and many of your customers will accept the explanation that your strategy was perfectly sound but that nobody could have been expected to predict the financial equivalent of being struck on the head by an asteroid. Which is what Viniar is claiming -- and I don't think it's stupidity, it's spin. Although it's possible that he believes his explanation--sometimes you can deceive others more effectively if you're deceived yourself.

The incentives just aren't there for planning for black swans. The highest rewards and the greatest glory go to those managers who bring in the highest returns during 'normal' times at the expense of leaving their funds exposed to black swans.

[Paul G. Brown]

May I make an Appeal for Technical Rigor, please. And tell a story.

All of these appeals to statistical arcania ('Chebyshev's inequality', 'high kurtosis distributions (aka fat tails)') are, bluntly, misguided.

We are not reasoning about a population (equity prices) based on the kind of uniform random sample you are required to have to employ said machinery. You have, instead, the entire series _up_ _to_ _this_ _point_ _in_ _time_, and you know that the phenomenon being examines is NOT independent of time. Swans and chickens aside, this is basic stats.

Now my story.

Several years ago I tried (an failed) to make various quantitative managers understand why they should use my company's brand of data management software in their systems. I talked to a lot of very smart people (physicists with PhDs from good schools whose non-native birth made it impossible for them to work in a national lab).

Among other wine fueled stories I heard was that the best performing fund on the street at the time (late 1990s) built it's trade strategies around kabbalistic divinations.

When I opined that this sounded a lot to me like a random process, and noted that random selection processes (throwing darts at the WSJ equities page) had a pretty good track record, well, that's the point at which I think we lost the deal.

[Thomas]

In the past, the bank that made the loan would often have been willing to offer a workout, modifying the loan's terms to make it affordable, because what the borrower was able to pay would be worth more to the bank than its incurring the costs of foreclosure and trying to resell the home. That would have been especially likely in the face of a depressed housing market.