I have long been sceptical of statistics masquerading as research. The stock market is littered with it. It is the process of taking history, averaging it and making some observation about the past (observing some coincidence) and then declaring that as something inciteful that will persist. It is the rather lame process of using history as your forecast.
But there is nothing prophetic about the past. It is simply what happened.
For instance - the average return from the All Ordinaries Index over any 12 month period using month end numbers is 5.77 per cent. Here is a plot of the 920 12-month returns from the All Ordinaries since 1937 on a month-by-month basis. So there is one red bar for the return over the year from January 1937 to January 1938, one red bar for the return from February 1937 to February 1938, and so on. They are all plotted on the chart. The highest 12 month return since 1937 was 80 per cent and the lowest was minus 45 per cent.
This is the chart of all the 12 month returns from the All Ords since 1937 plotted.
Now if the average of 5.77 per cent was a realistic expectation you would notice a large bulge in the returns around 5.77 per cent. In other words, most returns would be around 5.77 per cent, with a few outliers of course, but basically you'd hope for a lot of small positive returns around 5.77%.
But what you get of course is nothing of the sort. What you get is 920 12 month returns plotted on a chart from highest (+80%) to lowest (-45%) and, as you can see, you get an almost linear progression of returns from plus 80 per cent to minus 45 per cent. There is no grouping of returns around 5.77 per cent, there is no tendency to small positive returns, there are just 920 red bars stacked from highest to lowest with no gravitation towards 5.77 per cent at all and, I am afraid to say, this tells you that while the statisticians and financial product marketers might average the 920 numbers and present the average as an expectation, you are fooling yourself to believe it.
Equities are a risky asset class and over the next ten years you are not going to experience an ‘average’ return based on the past, you are going to experience ten completely unique annual returns that could be a disaster or glorious, and because of that, if you want an ‘expected return’ to base your expectations on, forget it, there isn’t one, because equities have good years and bad years and there are good stocks and bad stocks. In the next ten years you are going to experience whatever returns your own personal stock picks return and ‘the market’, if it is relevant at all, will deliver ten unique annual returns that will have nothing to do with an ‘average’ from the past.
The bottom line is that statistics (adding up the numbers and noticing some coincidences) is not research it’s just valueless number crunching. It’s the sun spot syndrome…the past is random, but random patterns that happen to exist do not bring order to the future.
This is why I turn off when an analyst writes “In eight of the last 11 years the market has rallied in the year of a Democrat election win”. Really? Go and do some work and stop wasting our time.
The most obvious example of past statistics being used to predict the future is “Seasonal charts” which we can produce on Thomson Reuters. These are designed to identify the seasonal trends in anything.
This is a chart of the seasonal performance of the All Ordinaries index (white) going back to 1982 (35 years of averaging the All Ords performance on a calendar basis). The orange line is the All Ords index so far this year (orange). You can see that, if the market does move in seasonal patterns then May and June are two of the worst months along with October and November. July is a great month as is December. The market this year (orange) has been following the same trend.
Now we start to reinforce these statistical observations with a few stock market idioms:
“Sell in May and go away”, “Buy before July”, the “Santa Claus Rally”.
Of course, for this chart to have any credibility you have to be able to explain why these seasonal trends occur and from that logic determine if the same seasonal factors will recur every year. Then it might be worth following. Otherwise it’s just a statistic because, let’s face it, there is bound to be a ‘best month’ and a ‘worst month’ even if the data was completely random.
Reasons for this seasonal trend persisting?
- One thought is that the international institutions dump the big banks ahead of the dividends because they don’t get the franking but take advantage of the franking chasing Australians pumping the stocks up before results and dividends.
- The banks go ex-dividend in May/June and Oct/Nov knocking up to 3% off the major bank stocks twice a year, so of course the index falls. If you notice the regular May and June fall is only about 1.5-2%, the ex dividends could do that.
- Because its self-perpetuating. The more a trend is observed the more investors will believe it and perpetuate it by selling ahead of the weak months and buying ahead of the recovery months.
- Tax loss selling may be a factor in June in Australia and December in the US. But it’s a bit of a lame excuse.
- In July a lot of new money arrives in Australia and then the same process happens in January in the US as people do their tax returns in June and December, decide to make make contributions to their pension (super) before the end of the tax year (a tax fiddle) and that money arrives in July and January in the hands of fund managers and is invested pushing the market up for a month.
Whatever, it happens and in this case I think it is becoming self-fulfilling, in which case the market should now drift into the end of the financial year and then rally in July.
We can do seasonal charts for any sector where a seasonal logic exists - but it’s pointless for a random unconnected bunch of stocks (the industrials index for instance).
The most commonly quoted season chart is that of the Bank sector:
This is the seasonal chart of the S&P 500:
Seasonal chart of Telstra:
SELL IN MAY AND GO AWAY
Sell in May and go away – We have seen the worst May since 2013, here it is again. In the last 20 years $100 invested in the S&P 500 from November to April would have turned into $343 but $100 invested in the S&P 500 from May to October would have turned into $98.50. Perhaps more interestingly $100 invested from November to April in the S&P 500 from the 1929 Crash to 2017 would have become $4270 whilst $100 invested over the same period from May to October would only be worth $257. The S&P 500 return in the last 6 months has been 10.8% which is the annual average prompting some fund managers to cash it in for the rest of the year.