Quant Bloodbath Revisited, A Primer
Lehman Brothers Stategist Becomes The Sage of The Subprime Contagion Theory

While traders and market watchers got a breather from last week’s volatility over the weekend and took time to figure out what had happened, a theory of what had caused the disastrous results at many quantitative hedge funds solidified into conventional wisdom. The sage of the theory was Matthew Rothman, the global head of quantitative equity strategies for Lehman Brothers. His memo on “Turbulent Times In Quant Land” (linked here) quickly became the most read primer on what caused the damage, and he became one of the most quoted analysts on Wall Street. Both the Wall Street Journal’s Kaja Whitehouse and the New York Times Gretchen Morgenson relied on his report for their analysis.

Basicially, what Rothman describes is a three step process that undid the quants. First, fund managers who faced margin calls and losses from in the debt portfolios found themselves unable to sell those portfolios at what they considered reasonable prices. The market for these collateralized debt products had always been illiquid, and many were value according to models of their expected performance because their was no market to compare them to. With fears of subprime and collateralized debt obligations spreading in the market, there were few buys for these positions. So instead the fund managers began unwinding more liquid equity positions, buying stocks they had sold short and selling their long positions, and this was the second step in the quant bloodbath.

This sudden unwinding caused the prices of these stocks to “misbehave”—quant speak for when prices stubbornly refuse to obey the models they have worked very hard on. What was “supposed” to go up went down and vice versa.

"Wednesday is the type of day people will remember in quant-land for a very long time," Rothman told the Wall Street Journal "Events that models only predicted would happen once in 10,000 years happened every day for three days."

Since so many of the much-prized and closely-guarded secret quant models are very similar, the confusing signals from the unpredicted stock price movements caused a domino effect. The third step of the unraveling came when a few large quant funds began further de-levering and otherwise reducing risk by liquidating positions. This exacerbated the problems set-off by the earlier sell-off.

The result has been a blizzard of letters from hedge fund managers to their worried investors, emergency conference calls and now, from Goldman Sachs, a bailout of one of their largest quant funds. Nervous investors are now in a prisoners dilemma of sorts, fearing that large-scale redemption requests could topple the hedge funds with an old-fashioned “run on the bank” but worried that if losses continue and everyone else bails out, they might be left suffering an even worse fate of being the last man standing on a sinking ship while the lifeboats head for the horizon.

Comments

Posted by Commander Warf, Aug 13, 2007 1:45PM

check out this gem:

13:43 08/13 Lehman quant analyst: Friday was a good day
NEW YORK, Aug 13 (Reuters) - Quantitative strategies turned
in a good performance on Friday and reversed much of the losses
from the four days of last week, Matthew Rothman, global head
of quantitative equity strategies at Lehman Brothers, said on
Monday.
Speaking on a conference call, Rothman said that Friday's
price movements may have erased some 70 percent of the losses
for the week for the strategy.
That does not mean that every quantitative equity fund
erased 70 percent of their losses for the week, because funds'
borrowings could have boosted losses early in the week.

Posted by Calgary Schmooze, Aug 13, 2007 2:19PM

I think that 10000 is actually binary representation. Did Rothman actually say "ten thousand" or did he say "one zero zero zero zero years"? 2^5 = 32 yrs. (Never trust a guy who can count to 1024 on his fingers...)

Does anyone have any "out of the realm of the obvious" information out there? How about a study of fund performance against the Myers-Briggs personality types of the managers?

Posted by , Aug 13, 2007 2:23PM

That WOULD be cool. Good luck with that one.

Posted by roger, Aug 13, 2007 2:24PM

"Wednesday is the type of day people will remember in quant-land for a very long time," Rothman told the Wall Street Journal "Events that models only predicted would happen once in 10,000 years happened every day for three days."

"Rothman said that Friday's
price movements may have erased some 70 percent of the losses
for the week for the strategy."

this is unadulterated horse shit, both counts, and this alleged analyst/assclown is so obviously blowing smoke into any orifice he can find.

All this tells me is nobody has any idea what the fuck is going on, least of all lehman if they pay this wingnut a salary for this kind of fiction.

Posted by Bulging Bracket, Aug 13, 2007 2:27PM

CS: I'd expect that most managers are INTP/INTJ, especially the quant shops.

Posted by roger, Aug 13, 2007 2:31PM

why is your website SOOOOOOO slow uploading posts? Takes like 2 minutes, that is why there are so many dupe posts on your board. You need a bangladeshi webmaster to fix that for you? I have the number.

ps- Rothman further went on to confirm, with ahigh degree of confidence, that 84.7% of statistics are made up on the spot, and that 93.8% of all the public will assume that the title "global head of quantitative equity strategies" means his imaginary statistics are much more accurate than the average man on the street's imaginary statistics.

Posted by Lee D, Aug 13, 2007 2:33PM

It should almost go without saying that if Gretchen Morgenson buys it, it's probably not 100% accurate.

Posted by , Aug 13, 2007 2:35PM

Wingnut. Nice, rog!

Posted by jt, Aug 13, 2007 2:36PM

anyone know what "data" (the fictional kind?) this joker used to come to these conclusions? Agreed with Roger, in the realm of rank speculation and complete malarkey this guys is really pushing the envelope.

Posted by Calgary Schmooze, Aug 13, 2007 2:38PM

BB: I'd expect that too. Which only heightens my concern that the other 97% of the population has no clue how these people tick. We know the general communication ability of that group too.

All of this is starting to look like a bad Muppet Labs experiment, with Dr. Bunsen Honeydew as the manager and talking head and Beaker as the trader.

Posted by , Aug 13, 2007 2:49PM

yes, his numbers are for the most part in the correct range. it is calculated using certain factors tracked by market participants. The 1/10000 years is a calculation derived from the historical distribution of these factors. And yes, the factors that have moved the most have retraced 70% (or more) now.

Posted by anon, Aug 13, 2007 2:51PM

Does anyone have a precise example of one of these "factors?"

Posted by , Aug 13, 2007 2:53PM

value, growth, etc. are ones that are commonly tracked by popular models such as Barra.

Posted by , Aug 13, 2007 3:04PM

the stats are simple. Take all the stocks in the Russell 1000 and calculate, for example, their FCF/P. Now rank high to low the 100 companies by their FCF/P. Calc an avg return for the top and bottom 10% (decile).

When he says they've retraced 70% he's just saying that the top-bottom decile spread has comeback in general for the factors he's tracking.

Rothman is a smart guy... egotistical, for sure, but smart.

Posted by nm, Aug 13, 2007 3:05PM

"The 1/10000 years is a calculation derived from the historical distribution of these factors"
Love that calculation with only ~100 years of data.

Posted by , Aug 13, 2007 3:07PM

Its more than enough significance statistically to calculate that interval within a certain confidence

Posted by MG, Aug 13, 2007 3:10PM

JC - I hate to join the grammar brigade on the dealbreaker message board, but your high school should consider rescinding your diploma based on this sentence:
"The market for these collateralized debt products had always been illiquid, and many were value according to models of their expected performance because their was no market to compare them to."
I would suggest "valued" in the place of "value." The second "their" should be "there," and prepositions shouldn't end sentences.
Not to be picky, but three glaring errors in just one sentence makes for tough reading.

Posted by , Aug 13, 2007 3:11PM

"Its more than enough significance statistically to calculate that interval within a certain confidence"


Basically, the model works perfectly until it falls flat on its face.

Posted by EE, Aug 13, 2007 3:11PM

yea, what anonymous 3:07 said

Posted by , Aug 13, 2007 3:13PM

Carney was a lawyer everyone knows they dont copy edit their own stuff thats what the steno pool is for (or whatever the modern day equivalent is. secretary i guess)

Posted by roger, Aug 13, 2007 3:13PM

""The 1/10000 years is a calculation derived from the historical distribution of these factors"
Love that calculation with only ~100 years of data."

I love that an alleged "Quant" (i.e. cannot tie shoes) will state that something that's almost supposed to happen once in 10k years happened three days in a row, but doesn't conclude that the model is FUBAR. Just a quirky 72 hours in the space-time continuum, should be smooth sailing for another 9,999 years now."

I can't believe people believe this shit. Especialy some of YOU guys who should know better.

Posted by Series7.5, Aug 13, 2007 3:19PM

the point the odds are overwhelmingly for a shift back to normality, which the models predict quite well. what the models will never account is how to position on the assumption that you can NOT stay liquid forever until you regain the baseline.

all the models upstairs here have SCREAMING buys or sells on right now like "go mortgage your home (if you can find a lender) to put on this trade, and i swear you will retire a billionaire if the fund doesn't go belly up first."

Posted by Bulging Bracket, Aug 13, 2007 3:21PM

The 1/10,000 stuff simply demonstrates how bad people instinctually are at statistics, even the exceptionally numerate ones. Dynamic, chaotic systems don't behave how our instincts expect nor how we are trained to expect with normal/poisson distributions.

Unfortunately systems that are inherently chaotic, dynamic, and with many stable points are those that we apply statistics to the most, since we understand them the least. Weather/climate behavior is essentially identical to the stock market, with 1/x year events happening with much greater frequency. This comes partly from model problems and partly from chaotic distributions that clump together and give "rare" results at certain stability points.

Just as building a levee to withstand a 10,000 year flood guarantees that next year is 10,001, a financial model that relies on these very bad things not happening is funcitonally equivalent to sending your account details to the Nigerian spammers. Don't apply linear, static abstractions to non-linear, dynamic, chaotic systems. When backtesting, look at it's behavior under 6, 7, and 8 sigma events, since that's when problems in your model will crop up and destroy your fund. No one wants to run a conference call that starts off with you saying you're down 30% on the week, so plan for "rare" events.

Posted by , Aug 13, 2007 3:26PM

3:13, most quants of course realize that the 1/10k years is garbage. Nobody believes that historical distributions are the correct forward looking distributions--those are impossible to know.

Instead the history is used as a comparison to illustrate how rare an event is compared to historical realizations. 1/10k years is not a number people believe in, but something that is used to show how it's an event that is incredibly far off from what has been seen so far.

Posted by anon, Aug 13, 2007 3:27PM

isn't Barra basically a COTS product? you mean these schmucks aren't even pretending to come up with their own models? It should come as a huge (HUGE) surprise then that theres so much 'overlap'...

Posted by Bulging Bracket, Aug 13, 2007 3:28PM

Series7.5: this is why only a few people make money at the bottom. You have to have gone all cash 4 weeks ago to take advantage of this. Bargains only come when most people can't (and don't want to) buy.

Shoritng the QQQQ in 98 or 99 was an easy call, you were just really bloody likely to go broke. Fortune had a great screaming sell piece on real estate speculators in Vegas in the Summer of 05, but things only started to blow up a year later and shorts on builders and lenders only really paid off 18-24 months later. It's hard to stay short unless you're already really, really rich.

Posted by Series7.5, Aug 13, 2007 3:39PM

it reminds me of a guy i worked with, massive outsize longs in homebuilders right through '05. started selling in '06 but when he gets his '05 bonus ... dude goes out buys a HUGE new vacation home in miami.

*shrug*

anyway, it's like livermore says, you make money by sitting. but for a lot of these guys the seat is just too hot. you other get shaken out of your positions and lose 50%, or in the vast majority of times you sit tight and lose even more then crumble. tough business! i wonder what kind of leverage jesse was employing. i'm guessing not 10x!

Posted by anon, Aug 13, 2007 3:43PM

"value, growth, etc. are ones that are commonly tracked by popular models such as Barra."

Hi Anonymous 02:53 PM. Unfortunately I don't have Barra. Do you know if there is there something similar on Bloomberg? Also, does "value factor" in Barra refer to the ratio between a basket (bucket?) of value stocks versues the broader market? Or is it the ratio between a basket of value stocks versus a basket of growth stocks? I', guess Barra produces all of these kinds of factors? Does Barra calculate factors as price ratios or price differences? Finally, are these reports saying that risk factors such as value vs growth saw 6 or 7 standard deviation moves last week?

Posted by lancey, Aug 13, 2007 3:57PM

BB
are you lumping poisson and 'normal' together?

do you even understand your regurgitated codswallop.

one thing is for sure. you don;t understand statistics. on one hand you get angry at a 240% return because it is too extreme then say that everyone else doesn't understand that extreme events are not that rare.

shorting in 98 was an easy call? ok bro. were you short?

Posted by , Aug 13, 2007 4:03PM

lighten up francis

Posted by c, Aug 13, 2007 4:47PM

did stat arb funds own cdos? that sounds strange to me. (fund managers who faced margin calls and losses from in the debt portfolios found themselves unable to sell those portfolios at what they considered reasonable prices.)

Posted by c- student, Aug 13, 2007 5:10PM

4:47 you are missing the middle step.

Many funds that own CDO's chose not to liquidate their illiquid debt portfolios, and instead liquidated their equity portfolios to meet margin calls / de-lever. Many of these funds had longs in quality stocks with good fundamentals and shorts with the opposite characteristics. The long names fell and short names rallied.

Separately, stat-arb funds whose models are watching and trading around these events could not deal with events.

Is this clearer?

Posted by Ben Disraeli, Aug 13, 2007 5:15PM

Theres 3 kinds of lies- lies, damned lies, and statistics.

Posted by Bulging Bracket, Aug 13, 2007 5:17PM

Lancey: People who have taken a few stats courses can understand systems based on basic distributions like normal, poisson and others.. say 301 level stats. No, I'm not saying they're the same, but that people with a superficial undersanding of the math can understand them and get to an instinctual level but can't handle more advanced and complex systems. You obviously can't understand somewhat complex posts.

Excluding a catastrophic event from your model and claiming that you're a genius because a catastrophic event happened making you a billion dollars are flip sides of the same fallacy. Too many traders and model developers will claim to be geniuses when a 6 sigma event happens in their favor but say that it wasn't their fault if a similar or slightly more likely event goes against them. This is the wrong approach to events that are perceived to be unlikely and the same psycholgy behind buying lottery tickets.

Being short in 98 and 99 was a very easy call. Rather a few people made it, even momentum players, because valuations were insane and all the metrics sucked. Most of the people who got short blew up, because they were right too early and couldn't handle a worthless stock going from $400 to $600 before it went to $0. Most of the people who were long blew up as well because they were wrong too late - they were the greater fool. In a manic bubble, it doesn't pay to be the lone sane person or the one eyed person in the land of the blind. You want to be fool n-1, maybe fool n-10, to lock in your gains ad take advantage of the mania.

Posted by Calgary Schmooze, Aug 13, 2007 7:33PM

If we're going to start bringing back 98-99 references, can we at least start referring to some appropriate talking head as the "Henry Blodget of 2007"?

Posted by sykes who, Aug 13, 2007 10:16PM

someone give this bulging bracket a book deal (or at least a db writer gig), i like that fool n-1/10 comment, well said.

Posted by , Aug 13, 2007 11:02PM

Bulging Bracket is takling out of his ass too much lately. All this news about quant blow ups seems to be bringing out the idiot in a lot of people who are categorically NOT quants.

Posted by , Aug 13, 2007 11:30PM

ta-ta-ta-da
I made money during the quant blow up!

HAS ANYONE HEARD OF RISK MANAGEMENT OUT THERE?

Posted by c, Aug 14, 2007 5:30AM

c-student
i understood what the story said. what i'm questioning is that a bunch of stat arb / equity market neutral funds actually held CDOs or exposure to them. most of these guys wouldn't touch an interest rate swap so i find it hard to believe they've been piling into cdo tranches.

Posted by french, Aug 14, 2007 9:22AM

Marketwatch :"Goldman Sachs, an investment bank that's also the second-biggest hedge-fund firm, said its $9 billion Global Alpha quant fund is down 27% so far this year, with more than half of those losses coming last week.

calm down, gs is just an bank running an hedge fund, like anyone on the street

Posted by french, Aug 14, 2007 9:22AM

Marketwatch :"Goldman Sachs, an investment bank that's also the second-biggest hedge-fund firm, said its $9 billion Global Alpha quant fund is down 27% so far this year, with more than half of those losses coming last week.

calm down, gs is just an bank running an hedge fund, like anyone on the street

Posted by London, Aug 14, 2007 9:27AM

Friday and Monday saw a big reversal of much of the market moves that hurt the quant funds so much Tuesday-Thursday of last week. Does anyone know how how many funds de-leveraged on Wednesday and Thursday, thus locking in most of their losses? Have many bounced almost all the way back--perhaps all these double-digit declines in August have gone closer to flat for the month for many funds?

Posted by , Aug 14, 2007 9:35AM

lots have locked in losses. some big ones have gone to cash. rief gained 4% back on Friday and 2.5% on Monday. so the answer is all over the spectrum.

Posted by lancey, Aug 14, 2007 9:41PM

BB said

"Most of the people who got short blew up, because they were right too early and couldn't handle a worthless stock going from $400 to $600 before it went to $0"

ok you are an accredited mo.

people blew up because a stock went 50% against them. if that is the case they blew up because they had too big a position.

point is most of what you know comes from regurgitating taleb.

and you don't understand it

you are my bitch

Post Your Comment