Reflecting on an interesting few days at the NAPF Investment conference in Edinburgh this week, I thought there were a couple of interesting observations relating to equities and volatility.
The equity return distribution is flattening, and tails widening
Professor Paul Marsh was the first to make this observation on the first day of the conference, followed by several of the high profile panelists in the session billed as “what’s hot and what’s not when it comes to pension investment”.
I think when this was mentioned it was in relation to the long term, lets say, 5 or 10 year equity returns as opposed to the short term.
Personally, I would challenge the assertion that the underlying distribution itself is changing, we just happen to have recently experienced an outurn that sits in the left tail of the distribution. Any analysis of long-term equity returns, such as this nice piece , shows that there is significant variability of returns even over the longer term, and this is driven by the presence of occasional large downside events which dominate the distribution. There is often a misguided notion of “time diversification” based on a combination of the misunderstanding of the mathematics of compound returns, combined with an anchoring to equity mean reversion. From a statistical viewpoint of course, the problem is that there is just not a sufficient number of independent data points to properly test the hypothesis that the experience of the last 5 or 10 years does indeed belong to the same distribution, against an alternative hypothesis that the distribution has changed.
Equity volatility mean reverts relatively quickly after peaking
This was a claim made by Professor Marsh and backed up by some quite interesting data. He’d looked at a number of instances of spikes in the VIX index, and concluded that on average it took 106 days post peak for the VIX to return to its long term average level (of around 20), and that the half-life of this process (the time taken to retrace half of the distance from the peak to the long term average) was a mere 11 days : however bad things might seem, they get better quickly.
I would challenge this conclusion, on a coupe of fronts. Firstly, the main problem with looking at the data in that way is the fact that its only ever obvious in hindsight that a peak has been reached. A quick look at the VIX shows that at any given high point it can go even higher, and remain above its original level for a long time. In fact it is often argued that going long the VIX as it makes a new high is a profitable trading strategy. The reverse (of going short the VIX as it makes a new high) would have historically lost a lot of money : in that case the observation that the post-peak “reversion” to the mean is relatively fast won’t be of much comfort.
Properly adjusting for survivorship reduces the historical equity risk premium
Many long term equity data studies (including previous versions of the yearbook compiled by Professor Marsh) suggested the historical equity risk premium – the ten year average annual compound return of equities in excess of bonds – was somewhere north of 4%. Professor Marsh argued, seemingly quite logically, that going back 100 years, the Russian and Austrian equity markets were a significant part of the world total (around 12%), yet the Russian market experienced a total loss following the 1917 revolution and the Austrian market took many decades to recover its value. Many studies, concentrating as they do on the US market data set, are implicitly biased toward the most successful market. Allowing for these other failures, Professor Marsh argues, reduces the historical equity risk premium to somewhat less than 4%.