The weekend edition of the Wall Street Journal contained a very thought-provoking excerpt of its reporter, Scott Paterson's, new book, "The Quants."
Based on the lengthy piece in the Journal, this is one book that promises to be worth reading.
The review provides a synopsis of some of Paterson's reporting results, primarily from the viewpoint of a secretive quantitative proprietary hedge fund known as the Process Driven Trading group, within Morgan Stanley.
Describing the group, Paterson writes,
"Instead of looking at individual companies and their performance, management and competitors, they use math formulas to make bets on which stocks were going up or down. By the early 2000s, such tech-savvy investors had come to dominate Wall Street, helped by theoretical breakthroughs in the application of mathematics to financial markets, advances that had earned their discoverers several shelves of Nobel Prizes."
Sound familiar?
It should. Paterson's description of Morgan Stanley's PDT group is almost identical to the image and composition of another infamous hedge fund, John Meriwether's Long Term Capital Management. That group, which was largely composed of an exodus of talent from then-independent Salomon Brothers, plus a helping of Nobel Laureates, including Myron Scholes and Robert Merton, attempted to exploit the same sorts of arbitrage opportunities in global securities markets.
LTCM didn't pan out as expected, shutting its doors in bankruptcy after only six years. And causing concern that its failure would bring down global capital markets.
What Paterson's teasing sample reveals is that, well, there really is nothing much new in financial markets.
Back in 1987, the then-experts in trading and risk management were convinced "portfolio insurance" strategies would avoid catastrophic losses. They didn't. Instead, identically-designed risk management systems all dumped the same types of securities amidst a sudden downdraft of equity prices, causing a torrential selloff.
Then again, in 1998, LTCM triggered the same phenomenon. Of course, that time, the underlying hedges were much, much more sophisticated than those of 1987. Back then, the big fad was lightning-fast trading of baskets of S&P components to exploit minute, fleeting price differentials.
By 1998, with tremendous advances in computing power and speed of data communications, LTCM's traders were able to construct far more elaborate, incomprehensible hedges which unwound in almost completely-opaque fashion. Opaque even to their designers.
Now, thanks to Paterson's brief excerpt in Saturday's Journal, we see that the quants were at it again. It's always the same. Bright young 20+ and 30+ year olds who've never seen a market meltdown apply the latest mathematical and physics advances to series of securities market data.
They may have read the fundamental papers on portfolio theory from the 1950s, but probably not. They may be familiar with the pesky, annoying details of minimum assumptions underlying the existence and behavior of liquid securities markets, but probably not.
What they are familiar with is how to take many series of data and extract measures of variance and covariance. Nevermind that what they 'discover' are temporal relationships of abstruse complexity underpinned by then-existing market environments.
To the latest generation of quant trading jockeys, it's all a straightforward application of math to securities markets prices.
The field you never see plumbed nor represented on the hot Wall Street hedge fund trading desks is..... catastrophe theory. The body of work exploring what happens when there are profoundly discontinuous changes in environments which cause rapid and extreme dislocations in outcomes of some system.
From my own background in statistics, I can assure you that the sorts of statistical methods most commonly applied by quants tend to require assumptions of continuity in pricing inputs and behaviors.
Precisely the conditions which send prices moving discontinuously are what tend to be unquantifiable. Especially when those conditions are triggered and amplified by a few dozen hedge funds operating similarly-based quant trading systems.
I've written about this before in posts under the 'Risk Management' label.
What happened in 2007-08 was, it appears, nothing more than the latest version of the application of the most current mathematical methods to series of securities prices which are, in reality, not naturally-occurring forces of nature, but the varying outcomes of human behaviors.
It's telling that, at the end of Paterson's excerpt, Peter Muller, head of PDT, has only two options: hold or fold/sell. He opts to sell.
That's another common element of quant strategies. They assume plentiful capital, or some sort of calm, VAR-based orderly loss of position values which allows a methodical unwinding of desk positions. But that's not what happens in a market meltdown.
Instead, position values melt away and erode capital, triggering risk alarms, margin calls and the need to sell positions in order to conform to capital requirements.
Ah, those pesky day-to-day operating assumptions and details.
If only....if only.....
By now, anyone my age or older, or even a decade younger, should understand that simply because hedge funds are run by people with names like Cliff Asness or Peter Muller, or John Meriweither, with pedigrees from Goldman Sachs, Morgan Stanley or Salomon Brothers, hardly means they will survive the test of a market collapse.
Sure, they'll no doubt do very well in a reasonably calm, or even frothy upward-moving market. Or a prolonged, gradual market decline.
But the sort of market conditions which are the specialty of Nassim Taleb, the market composed of sudden shear forces, seems to cripple these quant hedging strategies every time. Every decade, in fact.
No, there really doesn't seem to be much new in financial markets when it comes to advanced risk management, complex hedging strategies and quant models. They work really well- until they don't. Then they fail spectacularly, in concert, and turn ordinary market downturns into market panics.
Subscribe to:
Post Comments (Atom)
5 comments:
I will be interested in reading the book. However, from what I have heard from very good sources, PDT was flat for 2007. Thus, with this huge reported losses in the beginning of August, they still did not lose money over all of their strategies for the whole of 2007. Considering the many billions of dollars PDT has reportedly made Morgan Stanley, and its profits year after year after year, having one, really, really bad week hardly can be compared with the failure of LTCM.
Your post suggests that just because they suffered some huge losses when there were 10+ sigma factor movements implies that they were unaware of the risks. I would argue that the ability to break even in a year with such a big event causing such losses shows very good risk control. Additionally, my understanding is that PDT has continued to make money in 2008 and 2009. Making money in the market collapse of 2008 goes against your conclusions at the end of the posting.
In general I do agree with the basic position of your post. Far too many quants do not deal with massive kurtosis and price discontinuities properly or effectively. However, using PDT as a poster child of this is deeply flawed and is simply not reflective of the facts.
Thanks for your comment.
Because yours is anonymous, I think you can understand that I place zero reliance on your hearsay evidence of PDT's profits, losses, etc., for any of the years you mention.
To be honest, I'm not really interested in whether or not PDT managed to limp back to breakeven after their apparently large losses in 2007.
Nobody knew how long the discontinuity would last. They got lucky when the Fed stepped in to prop things up.
Then lent PDT's parent, Morgan Stanley, enough money to keep it solvent.
Doesn't sound like shrewd management to me.
Further, I'm not really sure how one prices discontinuities properly or effectively.
I'm not trying to be sarcastic. Not totally, anyway.
Really, if it were feasible, don't you think some of the brighter mathematically-inclined whiz kids would have done that?
Then, again, perhaps a few have. But we won't hear about them, because they've wisely remained modestly-sized and proprietary.
-CN
In this business, there are many reasons why one may need to post anonymously. That being said, you are correct to be skeptical of what is posted, anonymously or not. However, that does not mean that the information posted is incorrect or cannot be at least partially verified from other sources.
This field is littered with examples of traders that make small amounts only to lose very big amounts later. During this same time period, it is well documented that a proprietary trading desk in Morgan Stanley's fixed income division completely blew up costing Morgan Stanley many billions. According to published reports, this desk's loss was approximately an order of magnitude larger than the PDT's reported loss during this time period. Further, that desk had nothing like PDT track record of historical profits. This mortgage related loss cost many people their jobs including Zoe Cruz, Morgan Stanley's President.
The issues that you raise about the Fed bailing out Morgan Stanley in the end of 2008 have nothing to do with PDT, its profitability or its positions. While 2008 financial collapse is certainly related the the mortgage meltdown, the events in the equity market in the beginning of August 2007 had significantly subsided by the end of the month.
As for dealing with massive kurtosis and price discontinuities, there are many steps that can be taken. Sadly, most people look at the 1% worst case of recent performance and assume that they are looking at tail events. When things are relatively well behaved, this is about a 2.5 sigma event. In financial segments, 10+ sigma events seem to occur at least once or twice a decade. Simple stress testing for this magnitude event and putting together an appropriate action plan to ensure that corresponding losses can be tolerates is a basic first step.
As for PDT's decision to liquidate a major portion of the positions in one of its strategies in 2007, it had nothing to do with the availability of capital. It was simply a risk control choice. By selling and locking in known losses, PDT was ensuring that it could continue to keep trading since it had merely sustained a bad but fully acceptable loss in the face of a significant discontinuity. Furthermore, they were able to significantly unwind their positions since they were trading basic stocks which are very liquid, and their positions were quite manageable relative to liquidity.
A central issues which I hope the book will address is leverage. PDT was exceptionally well capitalized relative to the marginal risks it was adding to Morgan Stanley's books. Many other groups were not, and consequently they were forced to liquidate in order to cover losses on massively leveraged books. This is the lesson of LTCM, and it also applies to the long-short equity market in August 2007. Of course, the financial meltdown of 2008 is the prime example of this.
Again, my core point is that there are huge numbers of examples of trading desks, divisions, and firms taking huge bets with insufficient risk controls and negative optionality which can cause catastrophic losses. Again, one can find examples of this within Morgan Stanley at the same time period of 2007. PDT is simply not one of those cases.
Thanks for your followup comment.
As I have read my post, it isn't clear to me that I singled out PDT, so much as responded to Paterson's choice of it as the excerpt from his book.
My comment regarding Morgan Stanley's rescue by the Fed is, I think, germane, because money is fungible. Over what timeframe did PDT lose how much money, before making it back? I don't know. Perhaps Paterson's book will divulge this, perhaps not.
I agree with your 'core point.' I don't think we actually disagree on much, other than your focus on PDT. I only mentioned it because Paterson did, and my interest is in its portrayal as a trading shop which, like so many, seemed unprepared for the evaporation of the correlations on which it based so much, if not all, of its strategies.
You didn't address that aspect of my post. But I think it's really my core point.
Additionally, in the midst of a rapid collapse of equities, often exacerbated by the forces to which Paterson alludes, historic, typical price moves become irrelevant.
A friend involved in trading at Citibank in 1987 told me that, twice before, in the summer of that year, prices had 'gapped,' but up, not down. So fills came at much higher prices than anticipated.
Nobody complained then. Only when they gapped down and clearing took days did traders find it sobering.
-CN
I also believe that we mostly agree. Too few trading groups have actively attempted to model the potential impacts of discontinuities or temporal break downs of historical correlations. This has repeatedly caused problems in the past, and is only becoming a bigger problem as the worldwide markets become more and more interconnected in ways that are not obvious.
Post a Comment