Should we sell in May? The question is not just about market timing. It also raises two issues which investors encounter in other contexts.
Issue one is that the theories behind investment ideas are often general, while the implementation of them must be more specific.
The theory behind 'sell in May' is that investors’ appetite for risk is seasonal. As the nights get lighter and the weather improves in the spring we become more optimistic and hopeful, and so shares rise. April has for years been the best month for equities, with the All-Share index posting a total return in the month of 2.9 per cent on average since 1966. Such optimism, however, pushes prices too high with the result that returns are lacklustre in the following months. And then, as the nights draw in we get anxious and nervous. This causes shares to do badly, often falling to levels from which subsequent returns are good. Since 1966 the All-Share index has on average lost money in September but done well in the winter months.
Implementing or testing this general theory, however, requires us to buy and sell on particular days. An obvious way to do so is to sell on May Day and buy on Halloween. This isn’t just because these are salient dates. It’s because we’ve good historic evidence that this works. Since 1966 the All-Share index has given an average total return of 8 per cent from Halloween to May Day, but an average loss of 0.6 per cent from May Day to Halloween. This pattern is not confined to the UK, nor to recent years. Ben Jacobsen at Tilburg University and Cherry Zhang at Nottingham University Business School have shown that it has existed in almost all national stock markets since records began.
Last summer, though, this strategy was less successful: the index posted a total return of 5.4 per cent before inflation between May Day and Halloween. Much of this, however, was due to the market doing well in early May and in October. Shares did fall between June and September. That’s consistent with the general theory that appetite for equities is seasonal, even though it is inconsistent with one specific implementation of the theory. By contrast, “sell on May Day” has worked well so far this year: the All-Share index is 2 per cent down since then. But this tells us little about whether the general theory is still correct.
Which is an example of a common issue in the natural and social sciences (investing is applied social science). It’s the Duhem-Quine problem. This says that hypotheses can never be tested in isolation but only jointly with other hypotheses. In our case the general theory ('appetite for risk is seasonal') is tested alongside a specific one ('buy on May Day: sell on Halloween').
This problem afflicts all investors. The idea that value stocks – or defensives or momentum or growth or whatever – will outperform is a general theory. Which shares you actually buy is a specific implementation of it. Sometimes the general theory will work and the specific implementation of it will fail, or vice versa. When Russia invaded Ukraine, for example, many low-beta stocks (such as Polymetal) slumped, suggesting a big failure of defensive investing. But more intuitive implementations of defensive investing – such as holding big oil or big pharma stocks – succeeded, suggesting a success.
Our second issue is distinguishing between noise and signal. Between Halloween 2021 and May Day 2022 the index gave a total return of just 3.1 per cent, implying a loss after inflation. This means there has been only one Halloween-May Day period in the past seven years when returns exceeded their 8 per cent average, and that was the Covid-19 bounce of 2020-21.
How should we interpret this fact?
One possibility is that investors have wised up to the mistakes that caused them to under price equities in the autumn and over price them in the spring and so have eliminated the mispricing.
There’s plenty of precedent for this. John Cotter and Niall McGeever at University College Dublin have shown that many market anomalies disappear or weaken after they become better known. In the 1980s, for example, economists discovered that smaller stocks had for decades out-performed larger ones. Investors piled into them, with the result that since 1989 small caps have actually slightly underperformed. Maybe seasonal investing is going (or has gone) the same way as small-cap investing.
But there’s another possible reading. All investment strategies do badly sometimes. Even Warren Buffett struggled during the tech boom of the late 1990s; momentum investing did badly in 2009-10 and 2014-15 before roaring back; and so too did defensive investing in 2017-19. Maybe seasonal investing, similarly, has just had a short-lived bad time.
In real time, we cannot distinguish between these competing interpretations. As Georg Hegel said, the owl of Minerva (which brings wisdom) flies only at dusk. This is of course no help whatsoever to investors, but there’s no reason to suppose that the truth must always be useful.
This throws us back onto theory. Do we have strong theoretical reasons to expect the seasonal pattern of share prices to persist?
For this year, the answer is: no. There is always statistical noise around even robust patterns.
On average over the longer run, however, I suspect so. Seasonal changes in our moods are thousands of years old, as we see in the contrast between traditional May Days (when we celebrate hope and fertility) and Halloween or Samhain which is a time of fear, anxiety and preparation for hard times. It’s not obvious to me that investors are so rational and disciplined that they can be relied upon to over-ride such atavistic instincts.
