Footsie’s moves this year have strengthened my prior belief that equity returns are seasonal. The market’s big rise in April and fall in October fits the rule proposed by Cherry Zhang and Ben Jacobsen: 'sell on May Day; buy on Halloween'.*
But which stocks should we buy? My table gives us some clues. It shows price changes for some FTSE sectors from May Day to Halloween and from Halloween to May Day since 1987.
Several things stand out here.
First, all but one of the 27 main FTSE sectors does better in winter than in summer: the sole exception being utilities. Seasonality is not, therefore, driven by one or two unusual sectors. It’s widespread.
| Price changes since 1987 | |||
| Difference | Summer | Winter | |
| Construction | 8.8 | -2.7 | 6.1 |
| Electronics/electrical | 8.8 | -1.7 | 7.1 |
| Travel/leisure | 7.2 | -1.7 | 5.5 |
| Engineering | 6.8 | -0.9 | 5.9 |
| Media | 6.5 | -1.5 | 5 |
| Chemicals | 5.9 | -0.7 | 5.2 |
| Support services | 5.2 | -1 | 4.2 |
| Mining | 4.8 | 0.2 | 5 |
| IT hardware | 4.8 | 1.3 | 6.1 |
| General finance | 4.7 | -0.1 | 4.6 |
| All-share index | 3.1 | -0.2 | 2.9 |
| Telecoms | 1.5 | 0.4 | 1.9 |
| Pharmaceuticals | 0.7 | 1.7 | 2.4 |
| Tobacco | 0.1 | 2.9 | 3 |
| Utilities | -0.1 | 1.9 | 1.8 |
| Source: Datastream. Winter is October 31 to April 30, summer is April 30 to October 31 | |||
Secondly, the most seasonal sectors tend to be riskier ones. In some cases, such as general financials and IT hardware, the risk is market risk. In others, however, it is cyclical risk: construction, electronics and engineering.
By the same token, the less seasonal sectors tend to be more defensive ones such as tobacco, telecoms and utilities.
The main exception to this pattern are banking stocks. Despite being risky, these are not especially seasonal.
What’s more, there’s not much sign of seasonality fading. If we look only at data since 2000, we see a similar pattern to my table. There’s a winter premium in all 27 main sectors and big premia for construction, electronics and engineering: for the All-share index it is 2.1 percentage points.
This summer largely fits this pattern. We saw big falls in general financials, engineers and construction while some defensives such as beverages and pharmaceuticals actually rose (although tobacco and telecoms did badly).
You might find it surprising that investors haven’t wised up to stock market seasonality and thus eliminated it. Perhaps, though, it’s not so surprising. Evidence for seasonality has ambiguous implications. On the one hand, it could warn us not to buy in the spring or sell in the autumn. If we heed this message, we would indeed eliminate the seasonality. But we could read the evidence another way. It tells us to buy in April because other investors will soon bid up prices then, and it tells us to sell in the autumn before others do so. If investors make this inference, seasonality will persist and even increase.
My table shows something stranger. It shows that many sectors, and the All-Share index itself, have fallen on average in the summers since 1987: in fact, the same is true since 2000 too. This is weird. Even in the best of times, shares must be riskier (ex ante) than cash and so they should outperform it on average as compensation for this risk. For many sectors, however, this doesn’t happen. There is no risk-reward trade-off during the summer. That defies common sense and basic financial economics.
The most obvious explanation for this is that investors misjudge risk. As the nights get lighter in the spring they become too optimistic and too willing to take risks and so drive share prices up too high. Cyclical stocks especially therefore become overpriced. The opposite happens in the autumn: darker nights make us too pessimistic, which drives prices down too far.
This theory fits the fact that defensive stocks tend to be less cyclical than others: this happens simply because their prices are less sensitive than others to changes in investors’ perceptions of risk.
None of this of course means that cyclicals are certain winners for the next six months. Their great average gains in the winter months hides the fact that they have sometimes fallen. The construction and engineering sectors have both fallen four times in the last 30 winters, a failure rate of 13 per cent.
What it does mean, though, is that the chances of gains are considerably higher over the next six months than they usually are.
This poses the question: why do so many stock-pickers ignore this evidence when many of them are keen to trade on data that often has much less predictive power? The answer, I suspect, is that although the evidence for seasonality is strong, it is unconventional and thus less credible than the evidence about company fundamentals that stock-pickers usually rely upon. And this raises the old question allegedly posed by Maynard Keynes: is it better to be conventionally right than unconventionally wrong?
*The saying 'buy on St Leger day' (mid-September) never had much scientific backing, and it came a terrible cropper in 1987, when shares crashed in October.
