The latest hedge fund debacle is the meltdown in quantitative funds. In some ways, this meltdown is much more surprising than the sub-prime mortgage meltdown. Quantitative analysis was supposed to prevent these kinds of problems, at least, according to some of its advocates.
Quantitative investment analysis is the application of scientific method and technique to investment problems. Science gave us the industrial world, the wealth we have today. Why shouldn’t techniques that work so well in electronics and medicine work in investing?
Part of the problem is that the use of scientific method and technique does not guarantee that you will find anything important. For more than 30 years, the Committee for Skeptical Inquiry has been investigating claims of the Paranormal; that is, it has been investigating faith healing, astrology, ghosts and so on and has found… exactly nothing. There is no logical reason I know of why the world couldn’t contain ghosts. There is no logical reason why astrology and faith healing couldn’t work. It’s just when you look for evidence you don’t find any.
Similarly, there are many non-investment scientists who lack nothing in method and technique, who have spent their lives looking for something of value and found nothing. Forgive me if I don’t give you any examples, but some of them are friends of mine. Well, acquaintances. They found nothing because they were looking in the wrong places, of course. They had no good ideas. It would be remarkable if this wasn’t also true of some quantitative analysts. What is even more remarkable is that, right now, it seems true of so many.
The other part of the problem is how quants are trained and managed. Paul Wilmott, who runs a popular website devoted to quantitative analysis, has a section in his book Frequently Asked Questions in Quantitative Finance on how to get your first or second quant job. In it he notes, “If the hiring process is working well, the people seen by the bank will be roughly the same quality and from comparable backgrounds. Thus, you need to stand out in order to win.” But if everyone is of similar quality and comparable background, where are the new ideas going to come from?
There are two types of quantitative analysis and, therefore, two types of quants. One type works primarily with mathematical models and the other primarily with statistical models. While there is no logical reason why one person can’t do both kinds of work, this doesn’t seem to happen, perhaps because these types demand different skill sets and, much more important, different psychologies. I have only read one of Paul Wilmott’s books on quantitative investing, but judging on the basis of that book and his descriptions of other important quant work, including his own, the only thing he is interested in is mathematical models. In contrast, the Chartered Financial Analyst Institute’s book Quantitative Investment Analysis covers statistical modeling only, except for two odd excursions into the time value of money and basic portfolio theory. I guess they couldn’t find any better place to fit this material into their curriculum.
The quants who first came to Wall Street were mathematical quants and they brought with them, probability theory, differential equations and numerical methods. Good numerical quants live in an abstract world that only rarely touches the real one. Judging by what mathematicians tell me, it has a cold, austere beauty that you and I will never know. The ideas these people brought to Wall Street was revolutionary once, but it is not at all clear how many advantages these ideas still hold. Paul Wilmott offers a Certificate in Quantitative Finance, which includes instruction on the techniques that he believes work and the ones he believes do not. QuantLib offers open source software for modeling, trading and risk management. These techniques are clearly no longer rare. Without doubt, there are new ideas here, but perhaps not enough to keep the number of quants we have as gainfully employed as they want to be. In the real world of investing, there is no such thing as a permanent advantage.
Many of the later quants came from a statistical background. Statistical modeling demands knowledge of hypothesis testing, regression analysis, cluster analysis and other similar techniques. Good statisticians, and I have known many, are psychologically indistinguishable from accountants, except that statisticians are usually much more interested in finding the truth.
Judging by the job advertisements I have read and the questions headhunters have asked me, many employers, perhaps most, have no idea what kind of skills their quants should have. Investors must learn enough about quantitative analysis to know how to manage them. Or the quant must be taught enough about investing to manage himself and, perhaps, others. Both approaches are much harder than they seem. Which approach is safer depends on whether the investor or the quant knows himself better. It is usually the investor.
I have worked in several industries that perform quantitative analysis (although they always call it something else) and can state from personal experience that the quality of the work varies radically by industry. The best statistically oriented quantitative work is done in the pharmaceutical industry, thankfully. The worst is done in marketing research. The various parts of the investment industry fall in various places in the middle. In pharmaceutical research, there is a clear pecking order. Medical researchers give orders to statisticians who give orders to programmers and others. Putting quants in charge of the investment process is like having the statisticians give orders to the doctors. Would you swallow a pill developed by a quant with no medical knowledge? And if you wouldn’t, why would you place money with a quant with little or no investment knowledge?
Except, that in the beginning, this process worked. It worked and continued to work for some time because the ideas the numerical quants, and to a lesser extent the statistical quants, brought to Wall Street were novel and important. Just as important, perhaps, investment professionals almost always know less than they think they do. Unlike doctors, many of us in the investment industry have no clear area or expertise.
It is not clear that the recent debacle is a sign of impending doom. More likely than not, numerical quants will continue to have good months and bad. The less creative numerical quants will have more bad months than good months and, eventually, go on to do work they are better suited for. Unfortunately, as their theories become more nuanced, more difficult for us mere mortals to understand, it will become increasingly difficult to tell which numerical quants really have good ideas.
The same is not true of statistical quants. Or rather, the forecasting techniques statistical quants have brought into the industry, such as time series analysis, may be losing their potency, but that is not particularly important. Statistically oriented quants have another and much more important task to perform. As in medicine, the most important task a statistical quant can perform is to help investors decide what is true and what is not.
As long as a statistical quant is managed by an investor, it is not particularly important whether or not he has good investment ideas. In fact, investment ideas are a dangerous distraction and conflict of interest here. Several decades ago, Jane Martin, who would later become the head of the Managed Funds Association, told me she had confronted one of the industry’s first statistical quants, a CTA, and had asked him about his recent and oversized drawdown. Could the drawdown continue? Jane said that the man stared at her and said that it was a statistical impossibility. It wasn’t, of course. The drawdown continued and the quant eventually closed up shop. This was not an ignorant man, understand. According to Jane, the man had a PhD in statistics. The truth is much, much worse. He had turned into one of us.