It’s September, which poses the question: do markets learn? I ask because this month has historically been (on average) the worst for UK equities. Since 1966 investors have lost an average of 0.59 per cent in September, which just edges out June’s average 0.37 per cent loss to make it the worst month of the year.
The fact that this August has seen shares fall is no comfort here. If we look at September’s returns after August has seen losses, returns are still negative on average, at minus 0.07 per cent. The fact that August has been a bad month does not, therefore, mean September will be a good one.
Of course, differences between September’s returns and those of some other months might be just luck. But there is little chance that this is the case for the difference between September’s returns and those in the best months, December, January and April.
There’s a plausible reason for such a pattern, pointed out by Mark Kamstra at York University in Toronto. It’s that as the nights draw in in the autumn we get more anxious and depressed. This reduces share prices to such a low level that they do well in December and January. As the days get longer in April, our mood improves and so shares get another leg up.
But, but, but. In theory, this shouldn’t happen. If investors anticipate bad times in September, they should sell beforehand, causing prices to drop so that September’s returns are around average. Equally, if they anticipate rises in December or April, they should buy before then, again bidding away the better returns.
If investors learn, seasonal patterns should disappear and so September should see average returns – which means a small rise. But do investors learn?
There’s some evidence they do. In the mid-1980s investors realised that smaller stocks had done well for years. So they piled into newly created small-cap funds with the result that prices of smaller stocks rose so much that they underperformed for the next decade. After Paul Marsh and Elroy Dimson at the London Business School pointed this out, however, their prices began to rise again.
Also, John Cotter and Niall McGeever at University College Dublin have found that most “anomalies” in UK equities – ways of predicting individual equity returns based on price history or accounting data – tend to diminish significantly after academic researchers have identified them. That echoes work by David McLean at the University of Alberta and Jeffrey Pontiff at Boston College. They studied 97 different anomalies in the US market and found that their profitability dropped sharply after researchers had publicised them.
Learning, however, does not always happen. Heiko Jacobs and Sebastian Muller tried to replicate McLean and Pontiff’s results, and found that they only held in the US. In most other markets, they found, deviations from full market efficiency stayed strong even after they had been pointed out in academic journals.
And there is little evidence so far that learning has eliminated the September effect. In the past five years the FTSE All-Share has on average lost 0.6 per cent in the month, and has fallen in three of the past five Septembers.
This warns us that there are obstacles to markets wising up. One of these is that it is dangerous to short a stock. Volatility is so high that even if you are right eventually, you might have to put up a lot of cash in the meantime to cover rises in the price.
A second problem is that the September effect works because our tastes change: we are more risk-averse at the end of the month than at the beginning. But we are terrible at predicting such changes – which is why we pay too much for houses with swimming pools or for convertible cars in the summer. In good times, we can’t imagine bad. As Harvard University’s Matthew Rabin has shown, we tend to project our current tastes into the future: when we are tolerant of risk, we imagine that we will remain so.
Thirdly, success in trading depends on anticipating what others will do. The question is not: have I wised up to the September effect? It’s not even: have others wised up? It is: do other traders believe that others have wised up? And perhaps there are even higher degrees of derivation. When you are trying to anticipate what others will anticipate yet others will do, a lot can go wrong.
As if this is not enough, we have another problem. Even if the September effect has disappeared, it’ll be many years before we can say so with confidence. There’s a lot of volatility around that average loss of 0.59 per cent – so much so that even if this average persists there is still a one-in-six chance of the market rising 5 per cent or more in the month. We’ll need more than one or two good Septembers before we can declare the September effect dead.
If all this sounds inconclusive, it should. Stock markets are inherently uncertain places in which there are few general rules. In an uncertain world, only madmen seek certainty, and only charlatans offer it.
