A Rediscovered Asset Class: Commodity Futures

Introduction

I have long been interested in using commodity futures as an asset class but until recently there has been no simple way to implement a low cost diversified commodity futures investment into my portfolio.  I guess I could have bought futures and continuously rolled them over but this would be a nightmare to implement and the prospect of a truckload of pork bellies dumped on my front doorstep if I made a mistake wasn't too appealing. 

 Fortunately, the commodities investment landscape has changed for the better over the last five years or so.  Oppenheimer came out with the first collateralized commodity futures mutual fund in the late 1990s (Oppenheimer Real Asset Fund – ticker: QRAAX) which is based on the Goldman Sachs Commodity Index (GSCI).  The fund's expense ratio is rather high and some of the share classes require a front or back end load.  In addition, the Goldman Sachs index is heavily weighted towards oil and energy thus making it a bit less diversified than I'd like. 

About two years ago PIMCO came out with a commodities futures fund which is, in my opinion, the superior choice.  The institutional class shares (PIMCO Commodity Real Return Strategy Fund – ticker: PCRIX) carry an expense ratio of 0.74% and the fund tracks the Dow Jones AIG index (DJ AIG)  which I believe offers better diversification than the GSCI as will be explained below.  Another plus for the PIMCO fund in my opinion is that the collateral component is comprised of short duration TIPS (Treasury Inflation Protected Securities) rather than T-bills.  I believe that this offers superior inflation protection synergistic with the general movement of commodity prices which are positively correlated with inflation and negatively correlated with equities and bonds.  This should make for slightly better efficiency in a traditional equity/fixed income portfolio. 

Historical Study of Commodity Futures and Portfolio Efficiency

Okay, so where do commodity futures returns come from?  Robert Greer has written an excellent paper describing the source of returns from collateralized commodities futures: The Nature of Commodity Index Returns.  I would highly recommend reading this if you do not have an understanding of collateralized commodity futures.  Briefly, the returns are generated by the "insurance" premium derived from the risk taken by purchase of long futures or swap agreements, expectational variance (difference between the expected and actual price movements of the underlying commodity during the futures contract), rebalancing of the index components, and the return from risk free collateral (usually T-bills).  Interestingly, physical commodities as a group have actually lagged inflation but the components of collateralized commodity futures investment listed above have added several percentage points per year over and above inflation.  This is somewhat analogous to gold stocks which have outperformed physical gold by several percent per year over the last half-century or so due to fundamental differences in the investment itself even though both are based on the same physical gold. 

Since the PIMCO offering looks so promising I wanted to study it more closely.  First and foremost, I needed some idea of what the underlying commodity index returns have been historically.  Finding DJ AIG-like long-term commodity index futures returns is difficult since the DJ AIG data only goes back to 1991.  For the pre-1991 data I wanted to construct an index that resembles the DJ AIG commodity futures index as closely as possible.  The Dow Jones index is "liquidity weighted" amongst about 20 commodities.  This means that a group of commodities, such as oil and gas, that are more important to the economy are weighted more heavily than less economically important commodities, but not to the extent that any one group comprises more than one third of the index (as opposed to the GSCI which is almost 70% oil and gas).  The Reuters-CRB index, on the other hand, is equally weighted amongst a similar number of commodities.  The Chase Physical Commodity Index is liquidity weighted similar to the Dow Jones index but data only goes back to 1970.  In constructing a composite index, I chose to use the following for the time periods indicated:

1957-1969 - Reuters-CRB® Total Return Index (http://www.crbtrader.com/crbindex/ntotal_return.asp)

1970-1990 - Chase Physical Commodity index (http://www.iinews.com/site/pdfs/JAI_Summer_2000_Greer.pdf)

1991-Present - Dow Jones AIG Total Return Performance Index (http://www.djindexes.com/mdsidx/index.cfm?event=showAigHome)

 I managed to cobble together from the above an unleveraged, collateralized (with T-bills) commodities futures index going back to 1957. I then plugged the returns into my MVO (mean variance optimizer) program and generated the efficient frontier. The results surprised me. The numbers (all compounded and real - i.e. inflation adjusted) since 1957:

The correlation coefficient between the two is minus 9%. This means that, on average, when one index has a bad year the other will tend to have a slightly better than average year. The implications for portfolios are obvious: commodities can greatly lower the downside risk to an equity-heavy portfolio.

Here's a chart of the efficient frontier:

"1" represents the commodity futures index , "2" is the S&P500 index, "Rf" is the risk free rate (arbitrarily chosen as 1.5% for short term TIPS) and "A" represents the sweet-spot portfolio of about 60:40 S&P500:Commodity futures. The Y-axis is not labeled but is the annualized real return.

Commodity futures returns have only slightly lagged the S&P500 since the 1950's and actually were less volatile (lower std. deviation). You can greatly decrease the volatility of a stock-dominated portfolio by adding commodities without sacrificing returns. Take a look at these 10 portfolios on the efficient frontier:

 

One thing jumps out at me from this table. The "sweet spot" on the efficient frontier is located at about 11-12% std. dev. This is where returns are highest with the lowest volatility. This "sweet spot" portfolio contains no risk free asset! To me this means that you could have kept a hefty equity allocation to shoot for higher returns but accomplished better risk reduction by substituting commodities for bonds or cash.  As mentioned above, PCRIX is a collateralized commodity futures fund similar to what I have modeled except that the futures position is collateralized with short/intermediate TIPS rather than T-bills. I substituted TIPS (1.5% coupon) for T-Bills but found little difference in the efficient frontier portfolios.

There is a major downside to commodities mutual funds, however, in that  they are very tax inefficient and should probably be kept in a tax deferred account if at all possible.

What about adding other asset classes to the portfolio?  How do they affect the efficient frontier?  I added gold stocks (PM equity) and REITS to the MVO calculations (data since 1972). The following efficient frontier was generated:

1972-2003 geometric returns (annual real)

SP500                 ------  ------  ------  ------  ------  ------  ------
Commodities           0.0701  0.1401  0.2102  0.2802  0.1130  ------  ------
EAFE                  ------  ------  ------  ------  ------  ------  ------
Small Value           0.1303  0.2606  0.3908  0.5211  0.6567  0.8644  1.0000
REITS                 ------  ------  ------  ------  ------  ------  ------
TIPS(1.5%)            0.7676  0.5352  0.3029  0.0705  ------  ------  ------
PM equity             0.0320  0.0641  0.0961  0.1282  0.2303  0.1356  ------


Standard deviation    0.0320  0.0640  0.0959  0.1279  0.1599  0.1919  0.2238
Arithmetic mean       0.0388  0.0626  0.0863  0.1101  0.1298  0.1379  0.1383
Geometric mean        0.0383  0.0606  0.0821  0.1028  0.1187  0.1220  0.1169

For some reason, even though geometric returns were used by the optimizer to calculate the efficient frontiers, the asset class returns were printed out by my MVO program as arithmetic rather than geometric. For volatile asset classes (i.e. PM equity) this overestimates the true compound annual growth rate (CAGR).  Anyway here they are:

1972-2003 arithmetic returns (annual real)
-----------------------------------------
                           return     std
-----------------------------------------
1    SP500                 0.0715  0.1776
2    Commodities           0.0702  0.1969
3    EAFE                  0.0826  0.2331
4    Small Value           0.1383  0.2238
5    REITS                 0.0921  0.1741
6    TIPS(1.5%)            0.0150  0.0000
7    PM equity             0.1349  0.3749
-----------------------------------------

1972-2003 asset class correlations (annual)
---------------------------------------------------------------------------------
                                1       2       3       4       5       6       7
---------------------------------------------------------------------------------
1    SP500                 1.0000 -0.1783  0.6138  0.6956  0.4986  0.0000 -0.1303
2    Commodities          -0.1783  1.0000 -0.0556 -0.2539 -0.1483  0.0000  0.4710
3    EAFE                  0.6138 -0.0556  1.0000  0.4383  0.3224  0.0000  0.0661
4    Small Value           0.6956 -0.2539  0.4383  1.0000  0.8537  0.0000 -0.1633
5    REITS                 0.4986 -0.1483  0.3224  0.8537  1.0000  0.0000 -0.0716
6    TIPS(1.5%)            0.0000  0.0000  0.0000  0.0000  0.0000  1.0000  0.0000
7    PM equity            -0.1303  0.4710  0.0661 -0.1633 -0.0716  0.0000  1.0000
---------------------------------------------------------------------------------

It looks to me like a low-fee collateralized commodities product does more to enhance a domestic equity-dominated portfolio than any other asset class I've looked at including bonds, cash, international stocks and bonds, REITS, precious metals, etc.  It is interesting to note that the addition of PM equity, while also based on a commodity, does increase the efficiency of most of the portfolio combinations in conjunction with a diversified commodity futures product so I think that it is worth owning both.  REITS don't show up on the efficient frontier, likely because of their high correlation with small cap value (ScV) and since ScV has generated better returns it gets the nod from the optimizer.  Note also the conspicuous absence of the S&P500 and EAFE along the efficient frontier.

Now for something very intriguing. We will look at rolling 5-year compound annual real returns of three asset classes: collateralized commodity futures, ScV, and the S&P500.  ScV data consists of the Barra 600 value index since its inception in 1993 and prior to that consists of the middle 1/3 book-to-market sector ("blend") of the small cap market as defined by Fama and French.  Note that there are ScV indices which are more value weighted but none is available to the small investor unless he or she has a financial advisor with access to DFA funds.

The 5-year correlation coefficients are a whopping -0.69 for ScV:commodities and -0.31 for S&P500:commodities. These are much more negative than the one-year correlations as we saw above.  Here are the 5-year rolling annualized real (inflation-adjusted) returns ending in the year shown in the first column of the table:

If you look through the table you'll see that almost invariably when equities have had a negative real return commodities kick in with a positive return. In fact, ScV and commodity futures have not both had a negative 5-year return at the same time.  This is truly remarkable.  Note especially the brutal bear market of the mid 70's. Here commodity futures had above average returns while equities got killed. It seems that the diversification benefits for the long-term investor are even better than what it appears from the annual data.

Just to show that the 5-year data is not a fluke consider rolling 3-year CAGR's (real) for Commodity Futures and ScV. Again, when one has been negative the other has always been positive. You can't find a much better complement to ScV than commodities, IMO. For those of us who own a lot of ScV this is great news indeed.

PCRIX Tracking Error

Trying to determine the tracking error for PCRIX is not easy because it is tracks two very different indices: the DJ AIG commodity index and, supposedly, the Lehman TIPS index which is the collateral component. I can't find the full data series for the Lehman index but I think that a pretty good proxy is Vanguard Inflation-Protected Securities Fund (ticker: VIPSX) since it has an effective average maturity and duration that is very close to that of the Lehman index. In fact, I did find 6 mos. worth of data (2004) for the index and it has a correlation coefficient of 0.9997 with VIPSX.

I simply added the returns from the DJ AIG index to the returns from VIPSX after adding back Vanguard's 0.18% mgmt. fee and compared the sum to the actual returns from PCRIX:
 

Here's the monthly tracking error in chart form:

This chart shows the monthly net return for PCRIX vs. that predicted by simply adding together the DJ AIG index return and VIPSX minus expenses:

As you can see, the sum of the two indices tracks PCRIX pretty well. The average monthly tracking error is -0.08% or about -0.95%/yr. This is pretty close to what you'd expect since the expense ratio for PCRIX is about 0.74%/yr.  Since we only have about one and a half years of data the 95% confidence limits would be about -2.9% to +1.1% so the jury is still out on the magnitude of fund tracking error.  I will update this when we have more monthly data.

In an attempt to discern how much an allocation to commodities might enhance the ever-elusive portfolio safe withdrawal rate (SWR) I performed much the same analysis that was used for the EAFE (http://raddr-pages.com/Retire%20Early%20Web/International%20Diversification%20Improves%20SWR.htm) a few weeks back.

First the bad news. I was able to find data for the CRB index (presumably spot prices, equal weighted?) dating back to the 1920's. Commodities helped little, if any, in the survival of 1929-era portfolios. This is not surprising since this era was wracked by deflation and hard assets didn't fare much better than stocks.

The other severe bear market of the last century which started in the late 60's was a whole 'nother story and the news is good. This bear market was characterized by severe inflation which is favorable for commodities. Here are some representative SWR's for the post-1950's era with commodities data included (commodities collateralized with T-bills, expense ratio = 0.7%/yr.):

Clearly, commodities would've helped enormously in this bear market. I think that this is a better reference period than the 1920's since we were on the gold standard then which favored deflation when the economy went bad. On the other hand, since we went to fiat currency in 1971 the economy has been fighting inflation. Deflation is not as likely to happen again since the government can print money to get us out of economic funks. This, of course, was not an option in the gold standard era. So, barring another unlikely era of deflation, a dollop of commodities should go a long way towards inflation-proofing our portfolios and raising the amount of money we can safely withdraw in retirement.

Just for grins, TIPS (really Ibonds, 2.5% coupon) are substituted for cash in the table below. There was very little improvement in the SWR's. Commodities were much more effective.
 

Certainly no rational person would retire with a portfolio split roughly equally between ScV and a commodity futures fund but it is clear that a portfolio split about evenly amongst multiple asset classes including commodities would've fared well in the past and probably will in the future.

Disclaimer:  I have accumulated a substantial position in PCRIX as of mid-2004.  I will update my holdings on this page if they change significantly.

 -raddr

Last edited: 12/28/2004

[ Home ]