Are REITs “Just Small-Cap Value Stocks”?

Here’s a comment that was much more common when I started this job in 2006 than it is today: exchange-traded equity REITs, people used to say, are essentially “just small-cap value stocks,” implying that they didn’t offer any significant diversification benefit that you couldn’t get simply by holding a mutual fund or ETF focused on the small-cap value segment of the stock market.  (Some described REITs instead as mid-cap value stocks, or maybe SMID—small & mid-value.)  The analysis was pretty simple: most REITs have comparatively small (or mid-sized) market caps, and generally can be characterized as value-oriented companies rather than growth-oriented companies…et voilà!

Of course it isn’t so simple: diversification benefits come from underlying return drivers, not from either market cap or the value/growth distinction.  In fact, investors’ attention to both market cap and the value/growth distinction comes from the stunningly important empirical results regarding what are collectively called “the Fama-French factors” (after the most widely recognized researchers, Nobel prize-winning economist Eugene Fama of the University of Chicago and Dan French of Dartmouth University, though many others have contributed)—but here’s the thing: the Fama-French research never had anything to do with diversification!

To see the danger of using market cap or the value/growth distinction rather than underlying return drivers to think about diversification benefits, Table 1 shows the average correlation between monthly returns for exchange-traded equity REITs and monthly returns for nine segments of the market: large-cap growth, large-cap aggregate, and large-cap value;  mid-cap growth, mid-cap aggregate, and mid-cap value; and small-cap growth, small-cap aggregate, and small-cap value.

Table 1: Correlation between Equity REITs and Stock Market Segments, 4/1991 – 3/2016

You’ll notice that the highest correlations between REITs and the broad stock market are in the small-cap value and mid-cap value segments, meaning that there’s a kernel of truth in the old observation: small- and mid-cap value stocks are, indeed, more like REITs than are large-cap stocks or growth stocks.  But that’s a far cry from saying that “REITs are just small-cap (or mid-cap) value stocks”: in fact, even those segments have displayed very low average correlations with REITs of just 73.8% and 73.0% respectively.

Now compare the correlations among the segments of the stock market, as shown in Table 2, paying special attention to the correlations between the segments that are most like exchange-traded equity REITs (small-cap value stocks and mid-cap value stocks) and those segments that are least like REITs:

• Mid-cap value stocks have more in common with mid-cap growth stocks (correlation 75.4%) than they do with REITs (73.0%).
• Small-cap value stocks have much more in common with large-cap value stocks (83.4%) than they do with (73.8%).
• Small-cap value stocks have much more in common with small-cap growth stocks (84.2%) than they do with REITs (73.8%).
• Mid-cap value stocks have much, much more in common with large-cap value stocks (95.5%) than they do with REITs (73.0%).

Table 2: Correlations among Stock Market Segments, 4/1991 – 3/2016

In other words, even comparing REITs with the segment of the stock market that is most like them makes it clear that non-REIT stocks are more similar to other non-REIT stocks than any segment of the non-REIT stock market is with REITs.

You may wonder whether those low REIT-stock correlations are a data artifact left over from, say, the early days of the “modern REIT era,” or from the dot-com boom/bust, or from the Great Financial Crisis.  Nope.  Using the DCC-GARCH empirical methodology to estimate the whole historical series of correlations between exchange-traded equity REITs and each segment of the broad stock market makes it clear that the finding still holds—in fact, it’s even more powerful now than it has been historically.  Table 3 shows the latest estimate of the correlation between exchange-traded REITs and the different segments of the broad stock market:

Table 3: March 2016 Correlation between Equity REITs and Stock Market Segments

And Table 4 shows the latest DCC-GARCH estimates of the correlations among stock market segments.  Mid-cap value stocks had current correlations of 93.7% with mid-cap growth stocks and 95.4% with large-cap value stocks compared to just 62.0% for MCV with REITs, while small-cap value stocks had current correlations of 83.7% with large-cap value stocks and 92.7% with small-cap growth stocks compared to just 56.4% for SCV with REITs.

Table 4: March 2016 Correlations among Stock Market Segments

The bottom line?  Diversification benefits come from combining equity investments whose returns respond to different sets of underlying drivers, not from combining small-cap with large-cap stocks or combining value stocks with growth stocks.  The reason that exchange-traded equity REITs have provided such powerful diversification benefits relative to investments in non-REIT stocks is that REIT returns are driven by the real estate market cycle whereas the returns of most non-REIT companies in the stock market—small- or large-cap, value or growth—are driven by the much shorter and much different business cycle.  To achieve diversification, the most important characteristic of REITs is simply the first two words in their category: real estate.

Et voilà.

Technical notes:  I used 25 years of monthly returns, from April 1991 through March 2016.  For large-cap stocks I used the Russell 1000 indices, for mid-cap stocks I used the Russell Mid-Cap indices, and for small-cap stocks I used the Russell 2000 indices; all of those indices are associated with value and growth sub-indices.  DCC-GARCH is the Dynamic Conditional Correlation model with Generalized Autoregressive Conditional Heteroskedasticity developed by Nobel prize-winning economist Robert Engle [2002], which has been shown (see for example Case, Guidolin & Yildirim [2014]) to provide superior estimates of current correlations.  To estimate the parameters of the DCC-GARCH model I used the longest period available, which was January 1986 through March 2016.