“If you think you can find out what you need to know by going to see the management of a company, then I have nothing to say to you.”
says Herbert Blank, a notable Quant, as quoted in an interesting article on quantitative investing last week (via Barry Ritholz). Well, Mr Blank, that is fighting talk. Tremble, for you have roused the sleeping Baruch, who manages a Fundamentalist tech fund and who may indeed want to see the management of a company. I have something to say to you: you and your kind are dinosaurs, Mr. Blank, and the meteorite just hit. We are the small mammals who will gnaw your stinking bones and grow to fill your ecological niche. Read on to find out why.
I have been thinking about Quants now (and trying unsuccessfully to blog my thoughts) for some weeks, and much has been written about them in that time. I think the August stockmarket conniptions we foolishly choose to call the “subprime crisis”, and which I and my fund negotiated so brilliantly, were not about subprime at all. Or rather, they were, but in an unimportant and dull way. The memorable thing about August was that it was a crisis in long short quant, when the majority of funds in that strategy blew up, wiping out billions and making liars of their marketers. Morgan Stanley dropped $390m in a day in August, smashing through its VAR limits. Goldman’s CEO whined about seeing “25 standard deviation days 6 days in a row” as his Global Alpha Fund dropped 23% and had to cut fees to get bailed out. For the Quants it was a “hundred year storm”.
Quant investing will never be the same again, and I think it is beginning to affect the whole hedge fund industry. US inflows fell sequentially for the 2nd quarter in a row this Q3, a 22% decline on Q2 inflows, in an industry that grew assets globally over 30% in 2006. I would bet (but do not know) that the shortfall is concentrated in the quants.
But people need to save, and what is lost by the Quants will turn up somewhere else. I submit that the coming years will see fundamental strategies gain, strategies that posit not mean reversion, but thrive on greater volatility; strategies which rely the ability of single managers and incorporate a human element of “quality” and judgement. Art rather than science. Strategies like mine.
What exactly has made quant investing such a powerful proposition? The promise of riskless, reliable returns, with sleep-filled nights. Quant strategies, be they simple or complex, often come down to the same thing. They are mean reversion strategies. Computers pore over teraflops (I think that’s the right word) of historical data to discover apparently stable relationships between securities that, over time, tend to hold true. A typical trade could be: veeblefitzer-maker stock ticker DUMB has hardly ever outperformed veeblefitzer peer DICK by 10% in any given period and vice-versa, over a 15 year sample. Running that trade over the past 15 years yields a positive return 75% of the time with an average gain of 7%. Next time DICK moves -10% relative DUMB, simply long DICK short DUMB. Screening the entire US market to find such correlations throws up tens of thousands of potential trades, 10s or hundreds of which could be triggered every day depending on the sensitivity of the parameters. Other, “factor-based”, strategies screen vast lists of stocks for parameters like PEs, indebtedness, volume, momentum; you go long the top quartile of the list, short the bottom quartile. These rules can be stress-tested in every enviroment that decades of intraday stock market data can throw at you. Trading is automatic, computer-run, and executed extremely quickly, cheaply and efficiently.
Consider the simple strategy positied by Khandani and Lo; if a stock outperforms a certain amount in a given period, short it, if it underperforms under the same conditions go long; they admit its over-simplicity, and call it a “contrarian” strategy. They backtested, and found that in 1995 the strategy would have returned 345%! This stuff really works. OK, it ignores important things like trading costs and liquidity which could knock 100-150% off that gain. But that remaining 100%-200% performance came with almost zero market risk — in the teeth of the 1998 LTCM débacle, Khadani and Lo’s strategy produced positive returns, even improving slightly, as the S&P 500 fell from 1,200 to 1000. The returns of quant funds were thus not only very big; they were steady, uncorrelated with the direction of bond, stock, and other investment markets; they constituted a wonderful “anchor” to a generalised investment portfolio.
In the world of the early 2000s, exhausted and deceived by the huge swings in the major indices and the eradication of value in individual stocks as the bubble deflated, the promise of the Quants to make you money in any market, almost all the time, must have been irresistible to clients; a slam dunk, as George Tenet said of something else at the time.
So it was great for the clients. But the business model was even better for the practitioners. The strategy is extremely scaleable. Computers do the work, and once you have done the original programming, and made sure everything is running, all that is left is counting the 2 and 20 you charge the punters. No armies of expensive and unbearable analysts from Harvard and Wharton MBA programs costing $200k/year with a $30k Bloomberg each, and minimal real estate to share your fees and eat up your overhead.
Thus it is that the Quants have come to dominate hedge fund investing. Some funds are huge, 10s of billions of dollars under management. The MIT article points to the classic case of:
James Simons’s Renaissance Technologies, which has earned an average of more than 30 percent a year since its founding in 1988. Like other quant funds, it is ferociously secretive. Still, so many investors have trusted Simons that the two funds under his management now total more than $30 billion. In 2006 alone, he earned $1.7 billion running the fund.
Renaissance, DE Shaw, Goldman’s Alpha Funds, a squillion others, and every multi-strategy house worth its salt — even my own beloved Swiss employers, all employ in all or in part these quant-based mean reversion strategies. They provide a phenomenally large proportion of the daily trading volumes on the major exchanges. I think it is hard to overestimate their importance. It is worth investigating whether their rise can be correlated with the otherwise mysterious lowering of volatility, the Great Moderation, which has marked the US markets since 2002-2003. Widespread statistical arbitrage may have had a huge dampening effect on individual stocks, taking the example of DICK and DUMB above, working to mitigate relative moves up or down beyond 10%.
Is the fact that volatility is going back up in 2007, here measured by the VIX index, correlated with the growing lack of firepower of the Quants? Can I partly explain the Great Moderation and its apparent demise by the Rise and Fall of the Quants?
Well, maybe not. My theory is speculative lumping of the first order, although it makes sense. What is true, however, is that these strategies became terribly crowded over the years, as more and more graduates from MIT entered the industry, learned the trade on some prop desk, and left to start their own quant funds. This hurt returns. While there are after all a great many possible quant strategies, probably an infinite amount, they all do pretty much the same thing, and probably on a lot of the same stocks. As more and more people short outperformers and long the unders, and the later adopters are happy to make 5% off the same trade where before you would make 10%, slowly, the opportunity disappears. Let’s return to Khadami and Lo’s fascinating experiment: their Contrarian strategy, which made 1.4% a day in 1995 (for 345% a year)and 0.57% in 1998, showed through continuing forwards backtesting (sic) a marked diminution of returns over time. By 2006 it was barely making 0.15% a day before costs and trading friction. The 50%-100% years were gone, replaced by 5% and 10% years; this is not what the punters bought into. So you lever up 5x, 10x, and we get our 50% years back, but now at greater risk.
It’s clear that what happened in August was that one or more large funds cut leverage. In light holiday trading it must have had a disproportionate effect on prices, leading funds with the same positions to suffer losses, and given the rather hysterical atmosphere of those weeks (would capitalism survive?), to pull their exposure too, creating an awful cascade effect, what one could call a “quant squeeze”. Mean reversion didn’t just stop, the reversion reversed, ie it zoomed off in precisely the direction it was not supposed to go in. Strategies that would normally make 0.2% a day lost 5% and 10% a day. And the leverage! A 5% loss with 5x leverage is of course 25% of your capital, and for some it was clearly even worse; entire funds wiped out in a single session . With the 8x leverage typical of Quants at the end of 2006, Khandani & Lo’s Contrarian strategy would have dropped 25% in 3 days. The dry language of their article belies the carnage and the terrible velocity of what Quant managers must have gone through that week. The horror, the horror. As it was, once everyone was out of the market, mean reversion snapped back again. Those few who had the nerve to stay invested saw their losses made up and sometimes more, but most everyone else ended August with a huge drawdown.
I think no endowment manager, Fund-of-Fund manager or other fund selector can look at Quant in the same way after this. The greatest single thing about markets is that they adapt, and the overcrowded quant market, with its replicable strategies, is on the wrong side of this basic truth. Quant investing today must be like fighting the Borg, who are able to upgrade their defences to any weapon after the first 2 or 3 shots are fired.
The MIT article again:
Gregg Berman created commodity-trading systems at the Mint Investment Management Group. In the mid-1990s, he says, a good algorithm might trade successfully for three or four years. But the half-life of an algorithm’s viability, he says, has been coming down, as more quants join the markets, as computers get faster and able to crunch more data, and as more data becomes available. Berman thinks two or three months might be the limit now, and he expects it to drop.
“How long will your strategy last?” is the question the investor should be asking the Quant guy who comes to pitch, but at the same time they have to understand that the Quant guy is going to have absolutely no clue what the answer is. Both of them will likely find out when returns diminish, or, if they use leverage, in a more abrupt, nastier way. Successful Quant investing no longer comes down to the excellence of a particular strategy, but the ability of the programmer to stay ahead of the curve. The backtesting becomes meaningless. Investors still want the riskless returns that the Quant guys promised, but these returns only really existed for a short time. It must be like that, for otherwise we would have created a money machine, which cannot exist. The market is the Borg; you will be assimilated, and lose your edge.
Which is why I disagree with Rick Bookstaber, who thinks the existence of the Quants threatens the functioning of the system itself: he thinks the Quants’ tools are “interactively complex” — we don’t understand them — and “tightly coupled” to other parts of the market — spillover in subprime creates havoc in stocks through the demise of the Quants. Similarly, and unsurprisingly, Buttonwood (in a column this week which covers much the same ground as this post) fundamentally misunderstands the situation, when s/he writes of the Quants in August:
If it were just a few hedge funds. . . losing money, it might not matter. But the funds had become too important. Rather than adding stability, as marketmakers are supposed to do, they added volatility. . . regulators should also reflect that markets are less stable than they assumed. The presence of leveraged traders such as quants at their heart means conditions can now turn , at the flick of a switch, from stability to panic.
Both are wrong. If we recognise that Quants’ strategies almost always depend on mean reversion, we can go a long way in understanding their effect. Linkage extended from subprime to quant investing, but then the chain of disaster stopped (helped by large dollops of money from Helicopter Ben). August was the market functioning normally, clearing out a set of stale, unstable strategies which depended on a huge amount of leverage to stay relevant. I was there; sitting on 80 or so equity positions, staring at my Bloomberg, employing my old-fashioned directional strategy, I did not see anything in August that I would not have expected anyway, given the climate of risk aversion the subprime issue had created. If you think the financial system is coming to an end, you would expect Cisco and some dodgy tech stocks to sell off, which they did. I was made aware of the quants’ cataclysm by rumours and press reports. I did not notice it in the trading of my stocks, though admittedly, these reports probably also contributed more risk aversion. Yes, there were some strange moves in individual names. At one point small caps looked like they were outperforming large caps, which is not normal at times of risk aversion, but that was it, and it only lasted a few days. It was market neutral positions that were being liquidated, ie, for every long that was sold there was an offsetting short that was bought. Directionally, which is what I believe Bookstaber and Buttonwood refer to when they discuss systemic risks, there was no impact from the Quant meltdown. If I am right, by the way, when I say that one impact the quants did have in stocks was in dampening overall levels of volatility (selling stocks that rise and buying those which fall) from 2003 to 2007, then nervous nellies like Bookstaber and Buttonwood will in fact miss them when they have faded back into the obscurity that beckons; Quants’ absence will increase volatility and the magnitude of directional movements in markets and stocks.
So where does the punter get his uncorrelated kicks from while at the same time as reducing his exposure to quants? Well, it gets a lot harder, but he knows he’s got to find strategies that can provide 15%-20% a year and avoid leverage. Only equity will really do. I see no alternative but going back to basics, finding good, individual managers who do the work, and who are able to exploit their fundamental knowledge of a sector or a market to produce solid, consistent numbers. It also involves paying up for them, as these guys are scarce. Increased volatility will make the choice harder, and to make it worse, whole sectors in the market seem to have become more toxic recently, including financials, construction, cyclicals and consumer non-discretionary; markets are going up, but have become narrower, confined to commodities, infrastructure, healthcare and yes, technology. So the fund pickers will have to become more selective, and will have to finally do some work.
They will need a combination of long only and long-short strategies to succeed, but if we live in a world of increasing volatility, one thing they will likely have to look for is the ability of managers to benefit not from mean reversion, but from the breaking of bounds, of standard deviations. This is what I do, for I am the anti-Quant. I screen the markets not for stocks which are “stepping out of line”, but for stocks which can do something special, defy the conventional wisdom, and beat the consensus. I am an elitist, while the quants are socialists; I am a splitter, looking for differences, while the quants are lumpers, trying to see the similarities. On this site we are public in our admiration for the theories of Nasseem Taleb, of the idea that we live increasingly in Extremistan, where Black Swan events — of which the blow up of the Quants was certainly one — become more and more common, and more and more meaningful. The more we move in that direction the less effective mean reversion strategies become.
This, more than anything, is why I think the Quants and Herbert Blank are toast for now. Their sin is epistemological. Of course, when the strategies become less popular and mean reversion trades a lower proportion of daily volumes, then of course in that ironic way the market has the strategies will become more effective. However, I think we have a long time before that happens. It is interesting that Andrew Lo, co-author of the paper linked to above and professor to hundreds of financial engineering graduates of MIT now staffing dozens of quant funds, has sold his own quant fund Alpha Simplex and exited the market. Until then, Fundamentalist strategies which depend on qualitative assessments of individual stocks and investment themes, and which have never really gone away because they are not easily replicable but are created by the actions of particular personalities and individual processes, will predominate. That implies, Mr. Blank, that you had better start dusting down your DCF models and boning up on PEs and EPS forecasts; if you want tips on how to get the best out of meetings with CEOs and CFOs, put your email address in a comment and I will get in touch.