Thursday, April 18th, 2019

Capitulation and the rule of 35

Everywhere we turn in global equity markets, there is the smell of freshly-roasted bear. Our asset allocation models in euros and dollars have moved decisively to increase their exposure to equities. They have still outperformed on the basis of the last 12 months. However, the year to date numbers are not good and the priority now is to reduce the scope for further underperformance by getting back to a benchmark weighting in equities and bonds.

If this sounds like capitulation, then so be it. Our probability curves illustrate how quickly sentiment has changed over the last month. By way of reminder, we calculate the probability of equities beating bonds using a series of different sample periods ranging from 13-52 weeks. The reason for the different sample periods is that we can never be sure ex ante which of them is going to give the most reliable signal. But if we take an average of many samples (like meta-data in pharma trials) the risk-adjusted returns always improve. However, the underlying data are also useful because they allow us to visualise whether short-dated samples are more bullish about equities than longer-dated samples or vice versa. An upward-sloping, convex curve is bullish indicator while a concave, downward-sloping indicator is bearish. In the last four weeks the equity curves in our US$ and euro asset allocation models have gone from mildly positive to vertical take-off.

The obvious danger is that the moment when bears capitulate could also be the moment when the rally runs of out of steam. We can’t rule this out, but we can test which sample periods have historically produced the best and worst returns and see how this relates to the current situation. In the US, our data – going back to 1995 – suggest that the worst sample period to use for switching between equities and bonds is 34 weeks. It scores worst in terms of absolute returns and second worst in terms of risk-adjusted returns. (Despite this, it is comfortably ahead of the constant 50/50 portfolio on both measures.)

It’s not just the US where this sample period seems to cause problems. Looking at the average of the ten largest developed markets (US, UK, Japan, Canada, Australia, France, Germany, Switzerland, Italy, Spain), the 35-week sample ranks worst in terms of risk-adjusted returns. Not only this, but the five worst sample periods all fall between week 29 and week 35. It’s not our style to identify a single optimum sample period, but if it were, we wouldn’t expect to find it here.

Nonetheless, the concentration of bad results over this time period suggests that there might be some sort of behavioural explanation. We deliberately chose ten countries which have been dominated by institutional investors for decades, because this result would be consistent with one of the main unwritten rules of institutional investing. This says that anyone is allowed to underperform for a single quarter or even two quarters in a row. But underperforming for three quarters in a row normally means that the individual fund manager doesn’t get much of a bonus and it may even result in the firm losing the investment mandate.

The result described above would be consistent with fund managers in general suffering two bad quarters, finding that the third quarter is also going against them and then cutting the offending position because their career risk is suddenly too high. It’s not a proof but it is an interesting result.

The implication for equities vs bonds is that our model is “panicking early”. It is only sixteen weeks since the US equity index bottomed. The model has been wrong for one quarter and is now trying to make sure that it is not wrong for a second. Managing downside risk is never glamorous. It doesn’t make the headlines like a big contrarian call but it does allow us to preserve our risk budget for future opportunities. If our work on sample periods is relevant, the most dangerous moment for global equities will come 29-35 weeks after the most recent equity trough, i.e. sometime in July or early August.

Leave a Reply

Full Blog Archive:


Other Recent Posts