The search for anomalies covers an almost unlimited number of potential patterns. Some of them may be real, but many emerge by chance alone
One of the most maligned ideas in economics is the efficient market hypothesis, perhaps because what is actually a rather technical statement about financial market returns is conflated with some entirely different claim about the superiority of free markets over government dirigisme.
The EMH has various forms, but in brief its message is very simple: an individual investor cannot reliably outperform financial markets. The reasoning is equally simple: money doesn’t get left around on the pavement for very long. If it was obvious that the stock market would rise tomorrow, investors would buy shares immediately and the stock market would rise today instead. Anything that could reasonably be anticipated already has been anticipated, and so markets instead respond only to genuinely unexpected news.
But the EMH has a problem: researchers keep discovering predictable patterns in the data, and such patterns amount to big piles of money being left on the sidewalk.The most famous of these is probably the “January effect”: that returns are particularly high in that month. The January effect was originally explained by investors selling shares in December for tax reasons, depressing prices. Whether or not this is true, the EMH says that other investors should stand ready to buy those cheap shares in December, and the January effect should simply not exist.
The existence of the January effect and countless other anomalies looks like a puzzle for the EMH. But it is really only a puzzle if the anomalies suggest profitable trading strategies. That will not be true if an apparent anomaly turns out to be pure coincidence. The search for stock market anomalies covers thousands of stocks, tens of thousands of daily returns, and an almost unlimited number of potential patterns to be examined. Some patterns may be real, but many emerge by chance alone.
For instance, one recent discovery is that an asset’s monthly return can be predicted by looking at the same asset’s maximum daily return during the preceding month. Did you have to read that twice? It’s a pretty obscure finding, and where there are so many such candidates to be identified as an anomaly, some will be pure coincidence.
But let’s assume that some of these patterns are real. That is a minor embarrassment for the EMH; and it becomes a major one if the anomalies persist after they have been discovered. Yet this seems doubtful. Burton Malkiel, author of A Random Walk Down Wall Street, noted in 2003 that the January effect had become a Wall Street joke, “more likely to occur on the previous Thanksgiving”. Elroy Dimson, another EMH expert, documented the reversal of a major anomaly – a tendency for shares in small companies to outperform the market – after it became known.
Strictly, such anomalies should not exist at all, but a pragmatic believer in the EMH would surely feel her faith confirmed by the observation that the anomalies turn to dust in the glare of publicity.
A new research paper by David McLean and Jeffrey Pontiff explicitly examines the idea that academic research into anomalies is a self-denying endeavour. They find some evidence of spurious patterns: if a given dataset suggests an anomaly, including subsequent data tends to erode it. But what is really striking is that after an anomaly has been published, it quickly shrinks – although it does not disappear.
The anomalies are most likely to persist when they apply to small, illiquid markets – as one might expect, because there it is harder to profit from the anomaly.
The efficient markets hypothesis is surely false. What is striking is that it is very close to being true. For the Warren Buffetts of the world, “almost true” is not true at all. For the rest of us, beating the market remains an elusive dream.
Also published at ft.com.