Tuesday, October 29th, 2013

Hurricanes Hardly Ever Happen

As a young analyst, one of my formative memories was wandering around a deserted City of London on the day of the Great Storm of 1987. The next working day was Black Monday. So I freely confess that I get nervous whenever a big storm paralyses the transport system in the South East of England. The link between meteorological and market hurricanes is hardwired into my brain.

This gives us an excuse to discuss tail-risk – the extreme outcomes (sometimes good but usually bad) which are not described by the normal distribution of investment returns. This blog is also a response to an excellent article on tail-risk written by Sebastian Page of PIMCO  (www.cfainstitute.org/learning/products/publications/cp/Pages/cp.v30.n3.5.aspx) the abstract of which is as follows. “Asset allocation is evolving into an approach based on forecasts driven by macroeconomics and risk factor diversification. The dynamic nature of markets requires both secular and cyclical investment horizons. In addition, investors should look beyond volatility as a measure of risk and explicitly estimate the risk of tail events.”

We completely agree with the focus on protecting portfolios against the large drawdowns associated with tail-risk and we also welcome the idea (fully discussed in the article) that the correlation between asset class-returns is prone to sudden, non-linear changes. Where we disagree is the idea that the macroeconomic forecasting or risk factor diversification can help.

On the first point, let us use yesterday’s weather as an analogy. Unlike 1987, the UK Met Office forecast this storm very early and got its timing and central path right to within about an hour and 20 miles respectively. Nonetheless the disruption to the rail network was greater than forecast because the train companies underestimated the number of trees which would fall across the tracks. Service was suspended for 24 hours – four times the original estimate of six hours. This is the nature of tail-risk. Even if you forecast the circumstances correctly, the sample size is too small to make reliable projections of the eventual loss.

The debate about risk-factor as opposed to asset-class diversification is still in its infancy, and may take a long time to cross over from academics to investors. First, investors cannot hedge risk-factors with same liquidity and simplicity as asset classes. Second, diversification is often not an effective strategy in a crisis, when there are only two types of asset: safe and unsafe. In 2003 a typical well-diversified portfolio produced a monthly drawdown which was 3.5 standard deviations worse than the average five-year return to the prior peak (Harlyn data). In 2008, the equivalent figure was 6.9 standard deviations. The winning strategy was to make sure the portfolio was concentrated in safe assets, not diversified across loss-makers. In one sentence this is the reason why Harlyn developed its probability-based approach to asset allocation.

Leave a Reply

Full Blog Archive:


Other Recent Posts