The end of sell in May?

History suggests there’s a strong case for doing so: the market tends to do badly after May Day. Since 1966, the All-Share index has on average lost money in the six months from May Day to Halloween. Including dividends and inflation, returns have averaged minus 0.5 per cent in these periods. By contrast, they’ve averaged 8.2 per cent from Halloween to May Day.

This pattern isn’t a recent one, nor is it confined to the UK. Ben Jacobsen at Tilburg University and Cherry Zhang at Nottingham University Business School have shown that much the same has been true since the 17th century, and in almost all national stock markets since records began.

Worse still, investors face the risk of big losses in the summer months. Since 1966 there have been 11 May Day-Halloween periods when the All-Share lost more than 10 per cent in real terms. That’s almost one-in-five. But there have been only two Halloween-May Day periods when it lost as much.

All this, however, is not quite as convincing as it seems. This pattern has broken down recently. In the last two summers, the All-Share has risen. And it’s fallen in two of the last three winters.

Of course, this might be just luck. But there’s a reason to think otherwise. It’s that investors might have wised up. They’ve learned that shares prices have in the past fallen too much in the autumn and risen too much in the spring and have corrected their error. Perhaps they have learned to restrain the atavistic instincts that cause them to become over-optimistic as the nights get lighter and overly pessimistic as they draw in.

Any proven successful strategy – value, momentum, or whatever – is prone to behavioural risk: the danger that investors will wise up to it and so eliminate the mispricing. The stronger is the evidence in favour of the strategy, the more likely it is to be bid away.

However, even if we concede that 'sell in May' no longer works, it does not automatically follow that we should stay in the market. This is because there might be other reasons to sell.

Here, though, we must guard against a temptation. It’s easy to think we should use our judgment to answer questions such as: will fears of a trade war intensify or recede? Can stock markets withstand the likely rises in interest rates by the Fed and Bank of England? Will economic growth remain strong enough to support prices? And so on.

Such an approach is dangerous. 'Sell in May' worked so well for so long because investors were overly influenced by the seasons, becoming too optimistic in the spring and too pessimistic in the winter. Relying on judgment alone to tell us whether to sell now or not thus risks us falling into that error.

We need a more dispassionate approach. To do this, I took changes in the All-Share index in the six months to May Day and Halloween since 1996 and asked: what indicators, if any, can predict these?

We have several candidates here, as we know that there are some good lead indicators of returns.

For the six months to May Day and Halloween alone, however, most of these are weak. The link between the dividend yield and subsequent six-month returns, for example, is statistically insignificant. This is partly because there’s a high noise-to-signal ratio in such returns and because the yield works better as a predictor of longer-term returns. But it’s also because of the seasonality in returns in this period. Sometimes, the dividend yield has been high in the spring and yet prices have subsequently fallen – most nastily in 2008, for example.

This is disappointing for anybody wanting a reason to buy. The dividend yield now is above its long-term average. But a weak correlation between this and future returns means this isn’t as comforting as it seems.

We do, however, have some better news. Much the same is true of bearish signals. Foreign buying of US equities has been high recently, which is a worrying sign. But this too is statistically insignificant as a predictor of returns in the six months to May Day or Halloween. The same is true for the fact that the All-Share index is below its 10-month average.

There is, however, one lead indicator that is statistically significant for predicting returns for the six months to May Day and Halloween. It’s the global money-price ratio: the ratio of the OECD’s measure of the broad money stock to the MSCI word index. When share prices are high relative to the money stock, they tend to subsequently fall as investors rebalance their portfolios from equities to cash. With the money-price ratio still low (despite its recent increase) this points to slightly negative returns over the next six months – or, more accurately, a greater than 50:50 chance of the market falling.

Granted, this isn’t a strong signal: in noisy data, not much is. It is, however, a warning that even if we entirely discount the 'sell in May' rule (and I’m not sure we should) there is still a reason to be apprehensive about the market’s prospect for the next few months.