Murray Stahl’s Horizon Kinetics 3Q 2014 Commentary
This year’s commentaries review some of the surprising ways in which scientific-seeming or rule-based approaches to investing, which are now the norm and implemented via exchange-traded funds (ETFs) and index-based mutual funds, are foiled in practice by the social science reality of the fluid marketplace. A formulaic approach can work for a while, until a sufficient number of additional investors apply it. Their aggregate actions impact the supply/demand balance, valuations change, and the formula can no longer work. In reviewing some popular building blocks of the asset allocation model of investing, we have, hopefully, demonstrated:
-That an emerging markets index probably does not contain much in the way of emerging markets exposure, so much as exposure to large, relatively mature companies, many of which are exporters that, economically, are really global companies, not local. That the historical excess returns recorded by emerging markets indexes are probably not a reliable set of figures. That a non-indexed approach, or a different form of index, could better capture the local-economy potential of emerging markets.
-That private equity index returns are probably not as uncorrelated with the stock market as the historical results suggest. That a semantic correction from the common-use term ‘private equity’ to the more financially accurate term ‘leveraged equity’ might better capture the pricing risk inherent in private equity investing. Looking forward, private equity investors, providing the small slice of equity collateral that backs the substantial borrowing in these acquisitions, are essentially leveraged backers of low quality debt (or debt that can rapidly become lower quality) during a period of historic low interest rates. What might happen in several years if interest rates for those private equity companies are meaningfully higher? The private equity limited partner is taking on serious refinancing or duration risk, something that bond investors are now loath to do.
In the asset allocation framework of investing, investors who buy short-maturity bonds in order to avoid duration risk may also be purchasing private equity in the belief that they are diversifying their portfolios away from interest rate risk.
In this quarter’s commentary, we continue to explore the question of whether one actually achieves, though a typical asset allocation approach, meaningful diversification at all.
The Absence of Sectoral Diversification in the S&P 500® Index
There was a time when institutions could achieve some measure of diversification by employing a variety of money managers with varying sectoral exposures at different points of time. That as a tactic, with diversification as an end, has become difficult to a startling degree, for reasons that the following statistics make clear. Table 1, below, records the correlation coefficients with the S&P 500 of various industry-specific ETFs between December 31, 2007 and September 30, 2014.
Table 1: Correlation Coefficient with the S&P 500 Index
|IYW||iShares US Technology||0.903|
|RTL||iShares Retail Real Estate Capped||0.733|
|RTH||Market Vectors Retail||0.864|
|BJK||Market Vectors Gaming||0.810|
|IYH||iShares US Health Care||0.825|
|ITB||iShares US Home Construction||0.680|
|IYT||iShares Transportation Average||0.870|
Source: Bloomberg using monthly returns,
Horizon Kinetics Research
The iShares U.S. Technology ETF (IYW), as an example, and which does not even contain exactly the same technology stocks as the S&P 500® Index (“S&P 500” or “S&P”), has a correlation coefficient of 0.903, which means that its price behavior over almost seven years is 90% synchronous with the S&P 500. Another ETF, the iShares U.S. Retail Real Estate Capped, which holds REITs, or real estate investment trusts, that own retail properties and is part of the NAREIT real estate index, has a correlation coefficient of 0.733. Market Vectors Retail, an ETF which includes actual retailers like Target, comes in at 0.864. The iShares U.S. Healthcare ETF figure was 0.825, and so on.
The various industry sectors in this table were chosen on the belief that their particular business characteristics were distinct enough that they would be rather uncorrelated with the S&P 500. Not having much success with broader industry sectors, how about more narrowly drawn industries sub-sets, such as the Market Vectors Gaming ETF, which is comprised solely of casino companies, which are unusually volatile. The result: 0.810. How surprising that even this very narrowly constructed index has such a high correlation coefficient with the S&P. This means that one can sell all of one’s investments in constituents of the S&P 500 and replace them with gaming stocks, and still have a correlation of 0.810 with the S&P 500. It is an incredible statistic when you think about it.
Another interesting point is that the correlation coefficient over the time period in question of the Market Vectors Gaming stocks with the S&P 500—which is 0.810—is almost the same as the correlation coefficient of the iShares U.S. Healthcare ETF, which is 0.825. Is it possible—and this is a rhetorical question—that after a sober, realistic assessment of the available data, the aggregation of participants in the efficient market has arrived at the collective conclusion that the business fundamentals of gaming relate closely to those of the healthcare sector, the largest components of which are the likes of Johnson & Johnson, Amgen, and United Healthcare?
To lend a somewhat less abstract character to this line of reasoning, we’ll focus just on the holdings of the Market Vectors Gaming ETF, ticker BJK. A popular strategy in the modern era is to diversify equity exposure internationally by allocating some proportion of a portfolio to emerging markets. Here is the exercise: to calculate the emerging market exposure of BJK in the most conservative manner possible, since we know it has a very high correlation to the S&P, by eliminating U.S. companies. We will eliminate companies such as Las Vegas Sands and Wynn Resorts, and consider them to be U.S. companies since they trade in the U.S., ignoring entirely the fact that the preponderance of their earnings comes from emerging markets and that most of their assets trade as emerging market stocks on emerging markets exchanges. Going though each holding in this way, as depicted in Table 3, the remainder of the holdings in BJK are obviously emerging market stocks, and these remaining companies account for 51.7% of the net asset value of the ETF. Despite this effort, and despite the concentration in one very volatile industry, and which indeed is concentrated to a very great extent in the city of Macau, the correlation coefficient drops only from 0.810 to 0.666
The iShares Mexico ETF has an S&P 500 correlation coefficient of 0.844; MSCI Japan: 0.734. Here is a good one: Wisdom Tree Japan Hedged Equity (DXJ), in which all the currency risk was hedged out, was able to achieve a correlation coefficient to the S&P 500 of 0.679, which is slightly higher than the Europe and emerging markets ETFs.
In this context, one can easily understand why gold has become a popular asset allocation move. The SPDR Gold Trust ETF (GLD) has a correlation coefficient with the S&P 500 of 0.081, using the same December 31, 2007 to September 30, 2014 period. Of course, these figures are heavily time-dependent. If starting from GLD’s November 18, 2004 inception date to December 31, 2007, the correlation coefficient with the S&P 500 SPDR is 0.188.
There is a solution to this problem of convergent correlations. First, though, some exploration of why this phenomenon that is confounding