How to explain the many seasonal quirks and oddities that crop up in the stock market? Why is it that over the past 20 years the stock market has risen four years out of five in the past two weeks of the year, and only one year in five in the 36th week of the year? Or why have the winter months (October to March) produced so much better returns than the summer months (April to September)?
According to calculations by Stephen Eckett, editor of the UK Stock Market Almanac (from which these fascinating statistics come), the difference in the two periods since 1970 has been a genuine phenomenon. If you only invested in the winter months since 1970, for example - that is to say you put your money into shares on 1 November each year and went back into cash the following 30 April - your portfolio would have grown from £1,000 to £29,500 over the 33-year period. That represents a compound rate of return of 10.5 per cent.
The reverse strategy, investing on 1 May and going back into cash on 31 October, would have seen your £1,000 portfolio more than halve in value, to just £456. In fact, there have only been five years out of the past 33 that the summer portfolio has managed to beat the winter one. That is equivalent to an annual rate of return of minus 2.3 per cent.
The scale and persistence of this seasonal difference in performance certainly surprised me, although it is fair to say that the size of the differential has been less marked in recent years than in earlier ones (and the strategy may not work in the current year, as the market was uncharacteristically strong from May to the end of September, before dipping in October).
In fact, I think it is dangerous to read anything too significant into these statistics, compelling as they appear to be. It is not difficult to find some strong and plausible behavioural reasons, related to the incentives of the fund management business, to explain, for example, why the last two weeks of the year tend to be good ones for the market.
This particular effect may also have something to do with another famous market anomaly, the so-called January effect, a discernible trend for the market to rise disproportionately in the month of January, something which so puzzled academics for a while that it still features in many finance textbooks.
Because the stock market is a discounting machine, if everyone now knows about the January effect, it would be no surprise to find smart investors who have all read the textbooks buying the market in mid-December in order to anticipate the January effect - thereby creating the end-of-year bounce one month earlier.
Most statistical curiosities about the stock market are, however, just that - curiosities, without any enduring significance. It is a well-chronicled foible of human nature that we are prone to see patterns - and give them meaning - even in data that has been generated in a purely random way. We make that mistake every day of our lives.
Any investor who relies on statistical rules of thumb to govern their actions is likely to end up regretting the habit. This is not to say, however, that the stock market is actually, as academics once confidently proclaimed, a purely "random walk", in which each day's price movement is statistically independent of the one that went before.
All the most recent academic evidence I have seen knocks holes in this idea. There does appear to be a statistically significant degree of predictability in the movement of share prices from one period to the next (which is not to say that this is easy to identify or exploit for profit). It also appears to be the case that the stock market has a "long memory", in the sense of being driven by real world undercurrents of varying intensity that can persist for many years.
There also seems to be force in a related notion of mean reversion, the idea that any market price series that moves significantly out of synch with its long-term trend (as measured by standard deviation in statistical terms) will in time revert to that mean - and usually overshoot in the opposite direction.
But even mean reversion remains a hypothesis, not a fact, in the sense that it tells you nothing about timing - in other words, when rather than whether trends will revert. To repeat the famous aphorism of John Maynard Keynes, "markets can stay irrational longer than you can stay solvent".
By doing sensible things, however, with mean reversion on your side, you can at least shift the odds in your favour.
Even then, bad things can still happen. It would be comforting to think that returns in the stock market follow what statisticians call a normal distribution, as you could then calculate the probabilities of outcomes with confidence. In practice, there are too many "outliers" - too many extreme days on both the plus and minus side of the equation - to form a normal "bell curve" shape.
That is why you get such extreme events as the October 1987 crash, when the stock market fell by 12.2 per cent in a day, and why fully 40 per cent of the positive returns from the US stock market in the 1980s came from just 10 trading days. The way the market behaves is more volatile and more lumpy than would be the case if returns were normally distributed.
The behaviour of the stock market is still so complex and also so dynamic that it is beyond the capacity even of modern computers to model its behaviour with any real accuracy.
Never lose sight of that fact when next you read of some statistical curiosity that sounds like it holds the key to future profits.
I can strongly recommend the UK Stock Market Almanac as a useful work of reference, but you should be sure that you treat the curious data pages as working hypothesis, not as gospel truth.Reuse content