Correlation, decorrelation and sector selection
Andrew Clark (pictured), chief index strategist for Thomson Reuters Indices and Lipper, draws on second-order variability analysis to predict which US sectors will decorrelate first, assuming the current market rally has legs…
In a recent Reuters article, Bill O’Neill of Merrill Lynch was quoted as saying, “Correlation and volatility are very high, and it is very difficult for asset allocators to diversify. Making tactical calls in this environment is a fearsome challenge.”
This article shows which US sectors will decorrelate first, assuming the current market rally has “legs”, and briefly addresses the related matters of volatility and tail risk hedging.
In the two tables below we show the past two months’ sector correlations, using the Thomson Reuters US Sector Indices (using the Thomson Reuters Business Classifications).
Table 1A: US Sector Correlations – 7 October, 2011, to 6 December, 2011
Source: Thomson Reuters Indices
Table 1B: US Sector Correlations – 7 October, 2011, to 6 December, 2011
Source: Thomson Reuters Indices
All the values indicates significant correlations among all the sectors. Similar results were true for the past two years, except the correlation values tended to be smaller, and in some cases much smaller.
Using second-order variability analysis (2VD), it is possible to estimate which sectors will decorrelate first. Second-order variability analysis gives an analyst access to the information underlying current local correlations, allowing them to “see into” the reasons for correlations and to separate short-term from long-term correlations. With this information the analyst can infer the direction correlations (and decorrelations) may go.
The four sectors most likely to decouple first from the pack are Energy, Financials, Healthcare, and Technology. Energy, Financials, and Healthcare will more than likely continue to have some correlation with each other, but correlation with other sectors will decrease over time. When Technology decouples, it will have decreasing levels of correlation with all the other sectors over time.
The remaining sectors – Utilities, Telecommunication, Non-cyclicals, Cyclicals, Industrials, and Basic Materials – will probably remain correlated with each other for the foreseeable future.
As to when the decorrelation may start, February could show the first signs of the decoupling. By the end of the fitst quarter or sometime in the second quarter, further decorrelation should be occurring.
One of the implications of our analysis is that asset allocators can look forward to getting back to their old job of selecting investments and start to build well-diversified portfolios. And while there is concern about the sustainability of the current rally, if the eurozone can continue to make progress in resolving its problems, policy and policy shifts will stop dominating market thinking, which will depoliticise the markets. This would be good news for all concerned.
Regarding volatility, there have various press reports on tail risk hedging and how it has dampened investors’ returns, both in October 2011 when the market rose 20 per cent, and during the current rally.
The same 2VD work used above indicates that of the sectors that may break away first, Technology’s tail risk hedges probably need to be maintained, while the tail risk hedges of Energy, Financials and Healthcare can be pared. For the non-breakaway sectors, tail risk hedging could be reduced, since the chance of extreme events in these sectors is less than those in the breakaway sectors.
- By Category
- News from other sites
- Special Reports
- Partner events