Thursday, September 12th, 2019

Hyper-stability is destabilizing

This week’s commentary on many of our equity sector models mentions that fact that the recommended weights and the rankings are largely unchanged over the last four weeks. We note that this level of stability is unusual and can be the preface to a period of significant change. The price action in the US this week suggests that this was a prescient call. We now have headlines like “Quant Crash”, highlighting that many of the individual stocks which had performed very strongly over the previous three and six months have sold-off aggressively this week and that many of the previous losers have seen a surge of interest. There are even comparisons to August 2007, when a similar rotation took place. What does it all mean?

First, we are suspicious of any attempt to try and tie this in with any macro-economic story. It is true that yields on the US 10-year Treasury have suddenly stopped falling, but proving that this caused the rotation is quite another matter. There have been many occasions when a much larger movement in yields has not caused a significant rotation at a factor or a sector level. November 2016 is a good example. Furthermore, any serious analysis of the macro-economic causes of factor rotation would use a multi-variable, not a single variable approach. To call this is a coincidence has all sorts of derogatory overtones, but that’s what it is.

Second, bearing in mind that we are less than a week into this episode, we need to be cautious of the parallels with August 2007. Harlyn does not run the sort of factor models that are the basis for these headlines, but your analyst was doing exactly that, some 12 years ago. The important lesson of August 2007 was that the losses incurred during that month were largely eradicated by the middle of Q4 2007. The investors who really lost money were those who changed their process or reduced the amount of capital they allocated to it. This is not the death of momentum or growth or low volatility as attractive investment factors.

Third, we think the best explanation for the sudden increase in rotation is the fact that it was so quiet beforehand. We can’t test this for factors at the individual stock level, but we can look at the amount of activity recommended by our sector models (i.e. the gross change in recommended sector weights over the last four weeks) and compare that with our 24-year history. Just to be clear, this is not a measure of stock market volume; it is an internal measure of the changes our model thinks are necessary on a week to week basis. As of last week, this number was in the third percentile of all observations – i.e. very low. We have not adjusted this time-series for seasonality, because there is no significant seasonal pattern in the data, even though there may be in stock market volumes. We see the same effect in our Eurozone model, 1st percentile, and in Japan, 3rd percentile, but not in the UK, 41st percentile, which is probably because of Brexit and the volatility of sterling.

Finally, going back to August 2007, there are many strategists (us included) who would argue, with hindsight, that this episode marked the beginning of the late-late cycle in the last equity bull market. We have a sample size of one and we don’t even know if the current episode qualifies yet, so any sort of statistical analysis is impossible. Nonetheless, for asset allocation purposes, this may be the most important lesson. Time will tell.

Leave a Reply

Full Blog Archive:

Other Recent Posts