However, while May Day is a strong sell signal (as perhaps are high investment trust prices), we also have a strong buy signal. Figures from the US Treasury show that in the past 12 months non-Americans have been net sellers of over $50bn (£27.74bn) of US shares. In the past, such selling has been a signal that global investors are too pessimistic and hence a predictor of good returns on shares in the following 12 months. For example, weak foreign buying of US shares in 2003 and in 2008-09 led to rising prices in the following 12 months.
The obvious way to adjudicate between these competing signals is to use a regression equation to combine the signals coming from May Day and foreign buying of US shares.
This equation works better than the simple 'sell on May Day' rule alone. For example the 'sell on May Day, buy on Halloween' rule has worked in seven of the last 20 May Days. This might not seem impressive, but many of its 13 failures were small ones in which the All-Share index rose only a couple of per cent, while some of its successes were spectacular; it saved us from a 23 per cent loss in 2002 and a 30 per cent loss in 2008, for example.
However, our combined signal has succeeded 13 times in the last 20 Mays. For example, it told us to stay with shares in May 2009, which gave us a 19 per cent rise.
What’s more, of the seven failures, five came because foreign buying of US shares merely reinforced the 'sell on May Day' signal; the most recent of these failures was last year. On the 10 occasions when the foreign buying signal over-rode 'sell on May Day' and told us to stay with shares, only two proved to be mistaken - those in 1994 and 1999.
And here’s the thing. In the past 12 months, foreign selling of US shares has been so strong as to more than offset the May Day effect. Post-1994 relationships between the May effect, foreign buying and All-Share returns point to the index rising more than 10 per cent in the next six months.
Of course, there have always been reasons not to sell in May; it is such reasons that have often caused shares to be over-priced in the spring and thus to subsequently fall. This time, though, those reasons have an unusually strong basis in statistics.
