The quant revolution in academic finance began about 60 years ago, and it came to Wall Street in force about 30 years ago. It has been blamed for every disaster since, which is not entirely unfair.
Most of the innovations during this period have been quant-linked, and in many cases were pure quant. That means quant models affected all Wall Street events, including disasters. More positively, quants have increased the size, speed and power of finance, which is an enormous net benefit, but it does make disasters more significant for the economy as a whole.
Like in much of finance, the massive positives that come with quant also entail some additional risks. Furthermore, from each disaster, quants learn lessons that improve models going forward (you would think everyone would learn from disaster, but the evidence suggests quants are the exception, as only the rare non-quant is able to retain learning from bad economic events).
Let us not forget there were plenty of financial disasters before quants showed up on Wall Street, and the subsequent disasters (including the current one) had plenty of help from non-quants. The call to throw out quant models and go back to “basics” will not reduce the incidence of disasters, but will do a lot of damage.
I have been stewing about this for a while, and the spark that set of this essay was reading Scott Patterson’s The Quants: How a New Breed of Math Whizzes Conquered Wall Street and Nearly Destroyed It. The book is not very accurate, but it does me the favor of putting the anti-quant position in precise terms.
I am not qualified to review the book because I am mentioned in it, know most of the characters and have worked at three of the firms. I am not going to discuss any specifics about those people or firms, which should not be taken as confirming or denying accounts in the book, just as a “no comment.” What I am going to do is tackle the thesis of the subtitle — that quants “nearly destroyed” the financial system.
We have to begin with the definition of “quant.” Patterson is catholic on the subject. Blackstone co-founder Stephen Schwarzman qualifies when an example of lavish excess is required, as does Dick Fuld of Lehman Brothers when it’s time to show erratic, detached-from-reality behavior. Neither of them, however, qualifies as a “quant” by the usual definition, and neither is mentioned elsewhere in the book.
On the other hand, Benoit Mandelbrot, Nassim Taleb and Paul Wilmott are exonerated from quant original sin, despite solid quant credentials, including writing some of the most important quant books.
All the quants profiled extensively in the book run leveraged portfolios and were trained in either pure mathematics or mathematical finance. Structured product quants are mentioned only briefly, but are also blamed indirectly. Derivative pricing quants are absent from the story altogether.
Structured products and derivatives pricing are the areas where you tend to find more physicists. That’s interesting, because Patterson’s two intellectual complaints about quants are they trust mathematics over reality, or believe in efficient markets, neither of which are occupational hazards of physics.
The first near-disaster Patterson describes is Black Monday — October 19, 1987 — when the stock market fell nearly 25% in one day. He blames the crash on portfolio insurance, a common but not universal opinion. But speaking more generally, quants certainly brought the ideas of dynamic replication and index arbitrage to Wall Street, and built program trading operations on it. Those programs issued trade orders in a quantity that nearly overwhelmed public exchange processing.
Had things gotten a little worse, it’s possible that some exchanges would have failed, and that trading settlement would have been disrupted.
Nevertheless, I don’t think Black Monday supports the book’s thesis. Quants had been working to improve transparency and trading infrastructure for years before 1987, precisely because they feared this kind of event. It wasn’t the quant algorithms that failed; they actually made money (or in the case of portfolio insurance, reduced losses investors otherwise would have suffered). It was the old infrastructure that nearly broke down, with its embedded opaque frictions that entrenched non-quant practitioners exploited for monopoly profits.
There was plenty of blame to go around. The public stock markets could not keep up with the processing, adding to the confusion and panic, and nearly lost transactions, which would have led to major disaster; the public futures markets nearly failed to open for fear of counterparty defaults; the telephone and telex trading systems were overwhelmed (people could not get their brokers on the telephone, or put timely confirms on the telex), and information flows were hours behind the facts.
Following Black Monday, it was the quant algorithms that survived; the old ways of communicating, matching and settling trades were discarded in favor of rational, quant-based systems. This was creative destruction, the march of progress, the kind of disruption new ideas bring; not disaster.
Granted, the destruction was violent, and could have been worse. The crash might have been avoided if quants had been more patient. If we had really been in peril of a complete financial meltdown, there also might have been a case for slowing innovation even more than it had been already. But Patterson’s book presents no evidence for this.
Exchanges have failed several times in the past, delaying settlement and changing rules to bankrupt their customers instead of their clearinghouses. None of these failures have caused a major disruption in the financial markets, much less a complete meltdown.
Certainly, if the crash had been a larger disaster, individual investors might have fled stocks — but by 1987 we had robust alternatives like mutual funds for investment, and private equity and leveraged buy-outs if stock got too cheap. There is no reason to think that a slightly worse crash would have caused serious long-term disruptions to the capital markets.
People tend to compare disasters to the alternative of steadily-rising, non-volatile markets. That’s unfair, because we never get the latter for extended periods. Dynamic economies are characterized by bubbles and crashes. Even if quants really do cause a major crisis every decade, we don’t know that’s worse than the alternative.
When Genius Failed
The next quant brush with mass destruction occurred in 1998, when Long-Term Capital Management failed. As with the Crash of 1987, there are various theories about the ultimate cause, but there’s no doubt high-leveraged derivative positions held by hedge funds and dealers were at least an important ingredient, and possibly the main culprit.
While quants did not invent leverage, they tend to use it more than quals do. The reason is that highly precise quant analysis more often results in lowering the risk in a strategy than in raising the expected return.
A stock-picker using qualitative analysis, for example, might average 10% return per year with a standard deviation of 20%. A quant pairs trader might average 2% per year with a standard deviation of 1%. By levering up 10 to 1, the quant has twice the expected return as the qual manager, but only half the risk. Half the risk, that is, if you ignore the leverage risk.
And the leverage inflation does not stop at the strategy level. Most qual strategies are correlated with the stock market, and with capital assets in general. The quant strategy will not be highly correlated with anything (except leverage costs), and therefore volatility can be reduced further by running many quant strategies using the same capital pool. Less volatility, but more leverage.
Leverage risk begets liquidity risk, since small losses can be magnified to the point that they force trading, which sucks out all the liquidity; this, in turn, causes larger losses and more forced trading, which can lead to frozen markets, fund blow-ups and financial institution failures, without any large underlying economic cause.
In 1998, as in 1987, quant innovations disrupted a financial system built for slower, less leveraged, less sophisticated trading; like accidents caused when early automobiles drove on a road system built for foot and horse traffic. LTCM caused dramatic improvements in counterparty credit risk management, as well as leverage, funding and liquidity risk reporting and management.
Disaster was averted in 1998 when the Fed arm-twisted an agreement among LTCM’s counterparties to forgo collateral until the positions could be unwound in an orderly manner. Liquidity slowly returned to the market, leverage was taken down, and nobody died (reputations and fortunes were injured, however).
Suppose this had not happened, and LTCM had blown up. The LTCM trades, which were widely shared among hedge funds and prop desks, would have lost additional money. More funds would have blown up, and dealers would have been hit by prop trading losses, along with counterparty credit and prime brokerage losses. It is widely assumed that Lehman Brothers would have failed, and possibly other institutions.
Undoubtedly, this would have been painful, but it would hardly have destroyed Wall Street. In fact, doesn’t the notion that LTCM’s modest-sized hedge fund (by today’s standard) was going to bring down the world sound a little ridiculous?
Any problems would have happened during economic good times, and the damage would have been far less than what we have survived since 2007. We might have avoided, or at least reduced, the Internet bubble and the real estate bubble. We would have cleaned out a lot of financial deadwood, and hastened the adoption of risk management improvements.
Here, as with the Crash of 1987, I may be wrong. The damages might have been worse than I estimate, although Patterson’s book advances no new evidence or argument on that score.
More importantly, the old system of counterparty leverage and credit risk management had to change sometime. If LTCM hadn’t triggered a crisis in 1998, something else would have forced one some other time; and there’s no reason to believe that crisis wouldn’t have been worse than what actually happened.
By the way, this is not intended as criticism of the regulatory effort to avoid the crisis. At the time, no one knew how much damage an LTCM default would have caused. Information systems were not available to net and aggregate exposures scattered among many dealers and funds (deliberately so, in the case of LTCM, in order to disguise its strategies).
The Fed could not have been certain positions weren’t 10 times the size they actually were, and in the time frame available for decisions, it certainly had no capability to run Monte Carlo simulations or stress tests to estimate probable damage. Moreover, even a survivable disaster with Lehman and a dozen hedge funds failing was something most regulators would prevent, if possible. The Fed also deserves credit for moving decisively to attack the market deficiencies that led to the crisis.
The Guns of August
This brings us to the quant equity crash of August 2007. Here the charge of quant-complicity is more explosive. Unlike 1987 and 1998, the 2007 crash was followed by severe and extended financial system crisis and economic contraction. However, Patterson does not blame quants for the entire mess.
In his account, the initial cause is subprime mortgages. Losses in these instruments caused a quant hedge fund or prop trading desk to take down quant equity positions, as these were the most liquid positions available. These trades forced stock prices in the opposite of quant-expected directions, causing losses in all quant equity strategies.
Since these strategies use high leverage, a small loss can force large position reductions. For example, suppose a fund with $1 billion in net asset value holds $6 billion in long stock positions offset by $6 billion in short stock positions. If the long positions move down by 0.1% and the short positions move up by the same amount, the fund loses $12 million, or 1.2% of assets.
To maintain the original leverage ratio, this fund has to sell $72 million of long positions and buy to cover $72 million of short positions. So a move that would cost a $1 billion long-only equity manager $1 million, and would generate no trades, costs the quant equity manager $12 million and generates $144 million of trades. These trades add to the trades other quant equity funds are doing, which might move the stocks more than 0.1%, generating a new larger wave of trades.
Note that this argument does not require all the quant funds to hold the same positions, as is frequently claimed. At any given time, some stocks are going to go up, and some are going to go down. Any successful strategy must buy more of the former and short more of the latter.
Quant strategies are successful, so they are more likely to be long the stocks that go up than the ones that go down, and more likely to be short the stocks that go down than the ones that go up. There may be very little overlap between any two quant equity funds, but we know overall there will be more funds long a winning stock than short, and more funds short a losing stock than long. So if all quant equity strategies delever at the same time, prices will move in the economically wrong direction, and all long-term winning strategies will lose money.
The quant delevering forced the high-frequency traders out of the market, as they could not keep up with the flows. This removed the main source of liquidity in the modern stock market, and thereby exaggerated the price effect of additional selling. Patterson cites two events that turned the tide: Goldman Sachs Asset Management injected $3 billion into a quant equity fund to avoid selling, and Cliff Asness of AQR Capital Management began relevering his funds, which induced other quant funds to follow. The book claims that if these events had not happened, the subprime financial crisis might have been worse that it was.
Here is an alternative interpretation. As the subprime crisis was building, it gradually became clear that risk capital was going to get scarce and volatility was going to rise. Both of these meant all investors should take down leverage.
Quant equity was the first to get the message. It took its pain quickly, in four days, and early. It didn’t ask for a bailout, pressure regulators to change rules for its protection, rail against mark-to-market accounting while reporting wildly unrealistic values for its assets, blame market manipulators, or threaten to take down the financial system if its demands were not met.
No quant equity manager was paid a bonus for losing money. The quant equity strategy delevered, accepted its losses without whining, and went on to make the money back quickly and remain a stable investing strategy — even as much of the non-quant financial world was collapsing.
What if things had been worse? If the losses had persisted, funds would have faced investor redemptions to add to the selling pressure. It’s possible the entire quant equity strategy would have been wiped out, with depleted assets returned to investors. But other than quant equity managers and investors, who would have cared?
Remember, the stock market went up while quant equity was crashing; delevering funds were buying to cover shorts as much as selling to reduce longs. These funds are so carefully balanced with respect to all market factors that withdrawing them from the market makes almost no difference.
The August 2007 crash did not even disrupt the market. It’s true that stock market volume was high (August 9th set an all-time record for volume). But that record has been broken 151 days since; in fact, the volume that day was only a little more than the average volume since.
The market was volatile on the 9th: the S&P 500 high was more than 3% above its low. However, there had been many previous days as volatile, and the most volatile day of the crash was only a third higher than average volatility since.
This is not to say that quant equity managers, plus all managers of levered portfolios and all equity investors, did not learn from the crash. The markets on those days provided an important new data point for strategy design, calibration and risk management. Strategies are better as a result.
The Apology Question
I believe that the financial problems described in Patterson’s book, and others as well, resulted from the collision of quant innovations with entrenched inefficiencies. When discussing this with the author he asked one question that resonates with me: “Why has no quant apologized for the damage done?”
It’s no answer to say I believe the long-run good quants have contributed outweighs the short-term pain (although it’s true), because it’s still appropriate to apologize to the people sacrificed on the altar of progress. Moreover, no one elected me to mete out rewards and punishments, so what right do I have to decide Wall Street should be quantified and all who resist should be smashed?
Speaking for myself, and I think most quants, it never occurs to me to apologize, because I don’t think of my job as managing the financial system. My job is to design the best technical solutions I can to advance the interests of my investors or employers — or, when I’m trading my own money, my interests.
I understand that the collective result of many quants working like this is to re-engineer the financial system, which creates winners and losers far beyond my sphere. I have no idea whether this could lead to global catastrophe, beyond a mild mystic faith that knowledge is better than ignorance, progress is better than stagnation and good engineering leads to good outcomes.
I am sorry that people lose their jobs, their homes and their retirement savings, but only in the same general way I’m sorry people get sick or lose friends. A trade or model of mine might have caused someone economic distress, just as a sneeze of mine might have transmitted a virus that caused someone to get sick, or an article I wrote might have caused two friends to have an argument that caused a rift between them. If you worry about the unknowable consequences of your actions, you can never do anything.
For a different reason, I can’t apologize for being a financial quant. I chose the field in part because I enjoy it and the money’s good, but also in part because I believe I do long-run good. I wish I could do that good without hurting anyone, but I don’t know how.
I do make serious efforts to minimize harm, and think through effects as far as I am able (which is unfortunately not far). More importantly, I try to communicate as much as possible: through classroom teaching, writing books and articles and speaking at conferences, as well as through learning, reading and listening. I think this is the most important thing I can do to minimize the problems of innovation.
I do see room for apology in my faith in engineering, but it’s a conditional apology. I have no proof, or even strong argument, that technical innovation is always good. If someone sincerely preferred the pre-quant financial system, I can apologize for helping destroy it, without his permission, without even considering his interests. I don’t mean I would have done anything different had I known of his views, but I can still apologize for actions that harmed him.
However, I don’t believe it’s fair to pick and choose among innovations, saying you like your iPod but hate CDOs; you’re glad there’s Viagra but not ETFs; antilock brakes are good but high-frequency trading is bad; or even, sticking to finance, saying index funds are sensible but derivatives are weapons of mass destruction. Innovators are not auditioning for your approval so that you can have a pleasing matched set of changes.
It’s even less fair to pick and choose among effects of innovation, saying you like it when quants make the stock market go up and interest rates and oil prices go down, but don’t like it when they move prices in the opposite direction. Therefore, my apology for supporting innovation is reserved for the small group of people who reject all innovation — none of whom have asked for my apology, and few of whom are likely to read Patterson’s book.
I once argued with Roger Lowenstein, author of When Genius Failed, about Steve Miller, who presided over some tumultuous turnaround situations. I said he had done the right things, and Roger countered with, “Maybe so, but he enjoyed himself too much while doing them.”
In that spirit, I can offer a more general apology for enjoying myself while being a quant, and for making money at it. If my innovations made me rich (slightly) and happy (very), and made you poor and miserable, I’m genuinely sorry. Not sorry enough to give away my money and cry all the time, but sorry enough to think about ways to help people dislocated by financial change. Unfortunately, any solution I come up with will be further quant innovation, which will cause new disruption, and may help you at the expense of creating a new set of victims. Quant innovation is all I know.
Quants are a form of innovator, and like all innovators bring creative destruction to marketplaces. That always entails pain while it improves the world. In addition to the true pain, you hear screams from entrenched interests who want to reverse their losses in the competitive marketplace through legislation; and you see everything bad that happens to anyone blamed on the new thing, because it is new.
Not every quant idea proves to be good. Putting numbers on a good idea makes it more precise, testable, actionable and improvable. Putting numbers on a bad idea makes it more impressive, powerful, dangerous and opaque to non-quants.
Overall, quants have made markets better, not worse. In 1987, in 1998, in August 2007 (and certainly during the credit crisis that began soon after), quants were no threat to the financial system. Now, as at every time since 1980, quants are essential to fixing old problems and making new things better.
Article first appeared on RISK PROFESSIONAL - April 2010. Reprint courtesy of GARP and Aaron Brown
About the author: Aaron Brown is risk manager at AQR Capital Management and author of The Poker Face of Wall Street and A World of Chance: Betting on Religion, Games, Wall Street (with Reuven and Gabrielle Brenner). Mr. Brown serves on the Editorial Board of GARP and is a member of the National Book Critics Circle.