Markets have increasingly become correlated, in what some commentators feel is a consequence of zero interest rate policy (ZIRP). Another way this gets described is "risk-on", and "risk-off" trades. Part of our investment philosophy is to seek out isolated, less-correlated investments to accumulate as core holdings, and also to trade around the edges of these positions to take advantage of non-correlated peaks and troughs relative to the rest of the market. As I have described previously, one way that we do this is through writing naked puts to acquire positions, and to sell off appreciated positions through covered calls, a strategy which also nets some time decay premium as well.
My goal in this article is to test several investment ideas:
To explore how concentrated positions in selected sectors including reinsurance and offshore drilling might function, relative to common ETF's listed above.
The main way I find relatively uncorrelated investments is through a correlation analysis. There are a few such tools freely available, such as at Vanguard. In the table below, I analyze the one-year correlation coefficients for three popular ETFs (SPY, TLT, GLD), a representative member of my targets sectors of reinsurance (Axis Capital Holdings Limited (AXS)), and offshore drilling (Atwood Oceanics Inc. (ATW)), and a mega-cap oil company Exxon Mobil (XOM).
|SPDR S&P 500 ETF (SPY)||iShares: 20+ Trs Bd (TLT)||SPDR Gold (GLD)||Axis Capital||Atwood Oceanics||Exxon Mobil|
|SPDR S&P 500 ETF||1||-||-||-||-||-|
|iShares: 20+ Trs Bd||-0.785||1||-||-||-||-|
|Axis Capital Holdings Ltd||0.456||-0.524||-0.175||1||-||-|
|Atwood Oceanics, Inc.||0.59||-0.544||-0.008||0.331||1||-|
|Exxon Mobil Corporation||0.806||-0.64||0.185||0.375||0.545||1|
Correlations higher than 0.80 generally indicate fairly close movements most of the time. Correlations close to 0 indicate lack of correlation, while a negative correlation approaching -1.0 indicates anti-correlation. Through experience I have found that most individual U.S. equities exhibit a correlation coefficient between 0.5 and 1.0 compared to a broad market ETF such as SPY.
Personally, I insist on a correlation coefficient less than 0.5 before considering initiating a large position in an individual equity. (This is because I already use passive index investing as the ballast of my portfolio strategy, at about 50% of my equity allocation. I don't view correlations of 0.5 to 1.0 as providing additional diversification, especially since I assume any correlations will increase during times of illiquidity).
Comments about this correlation matrix:
XOM is fairly closely correlated to SPY at 0.8 (not surprising since it is a significant component of the associated index).
TLT is fairly anti-correlated to SPY at -0.78 (which suggests owning TLT to a large degree merely offsets the movement of SPY. The same thing could be accomplished by either shorting SPY or holding a smaller position. Highly negative correlation is not good - what you want is closer to 0, whether positive or negative, in my opinion)
GLD is fairly non-correlated to SPY at 0.14, which indicates true diversification value in holding both.
The targeted companies of AXS and ATW are somewhat correlated to SPY, at between 0.45 and 0.59. In this particular analysis, the reinsurance stock, AXS, has a lower correlation than the offshore driller, ATW.
The reinsurance company is fairly uncorrelated to the offshore driller at 0.33, suggesting they might behave well together in a concentrated portfolio of a small number of positions.
My theory behind why a reinsurer such as AXS exhibits the lowest correlation to other equities is that the random nature of large reinsurance events provides inherent non-correlation to the stock market.
An analysis like this is exploratory in nature. Before acting on any of these findings, I would consider checking longer term correlations such as 3 and 5 year. However, as a wise statistics professor once taught me, over long enough time periods, inflation becomes a dominant factor in all correlations, and it becomes difficult to isolate or de-trend that, although one can attempt it by subtracting chained CPI from all of the asset classes. I did an MBA project attempting just that, which I hope to publish here if I can ever dig it up and re-run the numbers now that it is 8 years later. (By the way, the thesis behind the analysis was a way of having three tiered levels of portfolio complexity based on the degree of investor ambition, using a stepwise addition of 1-2 assets at a time to reach increasing levels of portfolio diversification).
The risks of relying on historical correlations is that they may not hold true in the future. As was noted during the financial collapse of 2008, equities and commodities, and indeed everything but government bonds, became unusually correlated, as people scrambled for safety and cash. Correlations have a nasty way of changing at exactly the times you want them to be upheld! Nobody said it was going to be easy to be an investor...
We have compared one year correlations of three commonly-held ETFs involving U.S. equities, gold, and U.S. Treasuries, and compared these to a commonly held stock, XOM, and a reinsurance company and an offshore driller. In this small analysis of a single time frame, the two least correlated assets were SPY and GLD. Among the individual equities, the reinsurance company exhibited the lowest correlation to the broad market and to the other two individual equities. Finally, because of the high anti-correlation (-0.78) of TLT to SPY, bond ETFs may not serve as effective a portfolio diversifier as GLD (0.14).
This has been an exploratory analysis, and not intended to draw any definitive conclusions. I hope it was helpful in generating ideas for you.