Money is flowing into so-called “quant” funds at a higher rate of late, largely driven by a pursuit of superior returns, which they’ve had over the last few years. It’s also probably driven by a fear of missing out; investors and investing institutions are primarily motivated by out-running the middle of the pack. Nobody wants to be the sick and old stragglers picked off by the trailing wolves. I mean that literally, though it’s often used in analogy.
Before joining the race away from the back of the pack, there is merit in looking at the underlying premises in quant investing. One is that with enough data, trends can be identified through correlational studies relating “signals” to market results. Each “quant house” (money manager applying the algorithms) tries to identify their own secret recipe of signals which will out-perform the other quant houses, as well as the more traditional investment entities who also rely on leadership skill, experience and judgment.
With the secret sauce of predictive indices, returns are projected given certain investing criteria and, boom, you have a company-specific mix of equities forming the fund’s investment strategy. And the mix changes. Fast. With new data come adjustments to optimize the new predictive curve. Millions of shares trade, with other automated transaction software, daily. Don’t bother the humans.
A second premise of this system is that the correlations represent causation. That’s the first fallacy you learn in the study of inferential statistics. The second fallacy, incidentally, is making the opposite statement, based on correlation alone.
The quant houses are in effect testing their ability to predict, real-time, betting other people’s money. Time will tell if causation is established, because there will be a building pile of, yes, big data, to show the net positive effects compared to other methods of investing. Maybe.
Because a third premise is that the experimenters (the quant houses themselves) are not causing the results based on their actions, which are of course part of the market data analyzed by their own algorithms. It’s the same problem that companies face when they decide to pay “above market” salaries. They act on that metric, which is a statistical measure of central tendency (“market” equals average or median within some range). By doing so, the central tendency moves a click to the right. If other companies make the same decisions, the right-handed migration of central tendency is significant. Next year, everybody tries to catch up, which they can never do.
These are not good premises upon which to bet my money.
Correlations are extremely useful, let’s be clear about that. Millions of successful drug therapies have been applied to good effect, but advertised openly that the “mode of action is unknown”, as per the Physician’s Desk Reference. These drugs were selected due to extensive, complicated correlation studies. But it’s very tempting to leap to causation. It lets you relax and rely on the predictive variables to get stuff done. And make money, of course.
Once relaxed into a false sense of predictive superiority, you can get “unexpected” events. Drug research companies call them “adverse events”. Business calls them “shocks”, like the dot.com bust, the Enron era scandals and the collapse of the financial markets due to “sub-prime” debt loads. (I love how economists and investors use euphemisms to paint crap with a pretty brush.) Without identifying all possible factors involved in a correlation (which is not possible in large, complex systems), some pretty significant “signals” will be missed. Once the unexpected becomes certainty, the existing paradigms are busted and must be rebuilt. Some duration later, another quest for differential information advantage will pop up when things settle down.
The fact that the quant houses, and the preponderance of automated transaction systems in general, are an increasingly large part of the value calculations in public markets, we can conclude that individual judgment is being relied upon less and less. To me, this is an argument to take money out of the quant houses (if that’s where you have it) and put it into judgment-based assets. By that I mean, assets being employed by people who have a long history of varied experiences inclusive of successes and failures that are relevant to the assets. Because these assets will be differentially better-applied over decades. This has been my field study conclusion of the investing options available to me, measured by the evidence of history.
That evidence includes the fact that the most significant market shifts of the last several decades were not predicted by investment algorithms, the most notable of late being the drastic drop in oil prices, at a time when everyone prior to the price plummet (including short-sellers) failed to see it coming. Same with the recent Presidential election. Same with the Great Financial Fiasco of 2008. One has to believe that there’s a “signal” related to politics somewhere in the quant house’s machinery, right?
But probably the most glaring omission of rationality in the quant house movement is that markets aren’t predicted. They are made. By people.