The following table shows that while the returns on all three of these commodity indexes are highly correlated with each other, they have very low correlations with other asset classes, as shown in the following table:
GSCI Correlation with Other Major Asset Classes, by Currency
| USD* | AUD | CAD | Euro | JPY | GBP | |
| Domestic Equity | (.10) | (.20) | .26 | .14 | .14 | .10 |
| Domestic Bonds | (.06) | .16 | (.14) | (.33) | (.21) | (.26) |
| Global Equity | (.05) | (.14) | (.38) | .19 | .22 | .12 |
| Global Bonds | (.21) | (.03) | (.43) | .05 | .43 | .09 |
Given all these considerations, our preferred choice is the no sales-load Class D shares of the PIMCO fund (which tracks the DJ-AIGCI).
However, people who cannot access the no-load PIMCO shares (PCRDX) are still left with a difficult choice. Both the Oppenheimer and the Pimco funds’ "A" shares are still very expensive as index funds go, not only in terms of their operating expense ratios, but also due to the fact that they carry front end sales loads (which are almost never charged by index funds). Inevitably, this raises the questions about what is driving these high costs, and whether we can expect them to decline in the future.
On the one hand, the operation of a commodity index fund is quite different from that of a "normal" stock or bond index fund. Given the high costs that would be involved in holding physical commodities (transportation, storage, financing, etc.), commodity index funds instead hold a portfolio of commodity futures contracts. However, since these are leveraged instruments (that is, $1 invested in a futures contract gives you control over more than $1 of the underlying commodity), you don't need to invest the full amount of the money you have received (as the operator of the fund) in futures contracts. Both the Oppenheimer Real Assets Fund (QRAAX) and the PIMCO Commodities Fund (PCRAX) invest the remaining funds in government bonds. We prefer PIMCO's approach to this, because it primarily uses only real return bonds (U.S. Government issued TIPS). This is consistent with the role as an inflation hedge that commodities funds play in many portfolios. On balance, we believe that the additional costs inherent in operating a commodity index fund probably account for some of QRAAX and PCRAX’s relatively high (for index funds) expense ratios. However, until somebody (e.g., Vanguard or iShares) introduces a lower priced mutual or exchange traded commodity index product, we won’t know the extent to which this is the case.
On the other hand, the high front end loads charged by both QRAAX and PCRAX have nothing to do with the operation of the fund per se, but rather reflect the costs involved in distributing them (e.g., brokers’ commissions). We believe that, at this point in time, these high front-end loads reflect the difficult nature of the "sales process" for a retail commodity index fund. More specifically, we believe that most individual investors currently do not appreciate the potential diversification benefits offered by the commodities asset class, and as such are likely to regard it as a highly speculative (that is, risky) investment. Providing the education needed to overcome this initial investor resistance probably requires substantial amounts of a broker (or financial planner’s) time. Given this, Oppenheimer and PIMCO have probably had to charge a substantial front end load on their respective funds (QRAAX and PCRAX) in order to induce brokers and planners to spend the time required to sell these funds to their respective clients. Going forward, however, we believe that the average level of investor understanding of the benefits of investing in commodities will increase to the point that another company (e.g., Vanguard or iShares) will launch their own no-load mutual fund or ETF commodities index product. PIMCO’s decision to partially pre-tempt this step by making its no-load "D" shares available through mutual fund supermarket programs is very encouraging in this regard.
What About Natural Resources Stock Funds?
The high fees charged on QRAAX and PCRDX have logically led some investors to ask if they could substitute a lower cost natural resource (e.g., IGE) or energy focused (e.g., IYE) stock sector exchange traded funds (ETF) for either QRAAX or PCRAX and still receive similar diversification benefits for their portfolios (again, this assumes that an investor does not have access to PCRDX). GASFX would be another example of this approach.
Unfortunately, the answer as to which approach makes the most sense isn't clear-cut.
Over most periods of time during which all these funds have been in existence (and this is admittedly a small data set), the returns on the two commodity funds have pretty closely tracked the returns on the sector stock index funds (which invest in the shares of companies which produce the commodities, and products based on them). However, there have been periods during which the returns significantly diverged. Specifically, during the big stock market downturn we experienced in 2001 - 2002, there were periods during which the return on the commodity funds was positive (and therefore provided the kind of diversification benefit most investors seek from the commodities asset class), while the return on the natural resources stock index funds were negative (though by less than the negative return on the overall equity market). In short, during strong equity market downturns, it seemed that the equity market factor overwhelmed the commodity market factor in determining the return on the natural resources stock sector funds. As a result, people who invested in natural resources stock sector funds hoping for a diversification benefit from investing in commodities got a smaller one than they had expected.
From our perspective, the bottom line is this: the best solution for everyone will be the introduction of a no-load commodities index mutual fund, or, alternatively, a commodities index exchange traded fund. Until that happens, one is left with a choice between relatively expensive commodity index products, or less expensive, but also potentially less effective, natural resources stock sector index funds. As previously noted, the exception to this are the no-load Class D shares of the PIMCO fund. Unfortunately, they aren't available to all investors today. For a long-term investor without access to PCRDX, the sales load on either QRAAX or PCRAX could potentially be offset relatively quickly by the diversification benefits either of these funds would produce for a portfolio. On the other hand, we also respect those investors who, on principle, cannot bring themselves to pay the current front-end loads charged by these funds. For them, the choice is apparently either the lower costs and lower diversification benefits offered by natural resources stock sector funds, or waiting until a low cost commodity index fund product is introduced.
What About Commodity TRAKRS?
There is one other vehicle that tracks the Dow Jones AIG Commodities Index, but it is an unusual one. "Total Return Asset Contracts" were recently launched by Merrill Lynch, and are perhaps better known by their brand name "TRAKRS". Technically, they are neither mutual funds nor exchange traded funds: they are futures contracts, but of a very special type. Unlike typical futures contracts, they can be through a brokerage account, and do not require a separate futures trading account to be set up (although some reports suggest that other brokers may be reluctant to do this, given that TRAKRS are a Merrill Lynch product).
The reason TRAKRS can be held in a brokerage account is that unlike a typical futures contract, no leverage or margin calls are involved. Individual investors must post one hundred percent of the contract’s value when it is purchased. The value of the commodity TRAKRS fluctuates in line with the total return on the Dow Jones AIG Commodity Index. The current commodity TRAKR contract is traded on the Chicago Mercantile Exchange (www.cme.com) and expire on June 28, 2006, when they are settled for cash (presumably, Merrill will introduce another contract at or before this date, to enable investors to maintain their position in this asset class). TRAKRS can also be sold before maturity. TRAKRS are not treated like other futures contracts for tax purposes, and instead become eligible for capital gains treatment after they have been held for more than six months. Because they are futures contracts, TRAKRS pay no dividends; the only taxable event occurs when they are sold or expire.
Another attractive feature of TRAKRS is that, because they are futures contracts, they carry no annual operating expense charges. There are, however, other costs involved in owning them. First, there is a brokerage commission when they are purchased (similar to the brokerage commission one pays when buying an exchange traded fund). Second, due to the structure of the contracts themselves, TRAKRS typically trade at a slight premium to the underlying index value. This premium has been estimated to be about three percent, on average.
How, then, would you evaluate the trade-off between the PCRAX mutual fund and the commodity TRAKR? There are a number of issues involved. The first is the relationship between the front-end load on PCRAX and the combined brokerage commission and price premium on the TRAKR. Let’s assume (unrealistically, but for the sake of illustration) that the sales load equals the brokerage commission. The question then becomes what discount rate should you use to convert the three percent price premium to an annual equivalent fee (analogous to a fund operating expense charge) over three years? Logically, the rate you use should reflect the opportunity cost of that money - that is, the rate you could otherwise earn on the TRAKR premium charge. To keep this example nice and tidy, let’s assume that this discount rate is what you would expect to earn if you invested that three percent in PCRAX. The breakeven discount rate that would make your three percent estimated TRAKR premium equal to the 1.24% annual expense charge on PCRAX is 11.5%. If you expected to earn more than this each year on PCRAX, it would appear to be a better deal than the TRAKR, assuming that the front end load on the former was equal to the brokerage commission on the latter.
But of course, this assumption isn’t true - the front end load on PCRAX is probably much higher than the brokerage commission you would pay to buy the TRAKR (though because the latter is a futures contract, you should check on the size of that brokerage commission in advance). This means that the breakeven opportunity cost is actually much higher. For example, if the difference between the sales load and the brokerage commission reduces the effective TRAKR price premium to 2%, the breakeven discount rate rises to over 38%. In this case, it seems like the TRAKR is a cheaper way to gain exposure to the commodities asset class. However, this assumes a three year holding period. If your expected holding period is shorter -- remember, the initial TRAKRS contracts mature in June, 2006, and if the three percent premium continues (even on the shorter maturity contract), the breakeven discount rate will be lower, and PCRAX may still be a better deal.
On the other hand, what about the case where you can purchase the PIMCO "D" shares without a front-end load? In this case, the TRAKRS don’t look like such a good deal. If you assume the brokerage fee to purchase the TRAKRS equals 0.25% (twenty-five basis points), the breakeven rate of return falls to only 7.00%. If you expect the Dow Jones AIG Commodity Index to increase by more than this much per year over the next three years, you would be better off investing in the PIMCO "D" shares and giving TRAKRS a pass.
The bottom line is that TRAKRS are new, different, but potentially very interesting products. In addition to commodities, TRAKRS have been introduced that track gold as sell as the Euro/U.S. dollar exchange rate. If you don’t have access to the PCRDX shares, and are looking for a cheaper way to invest in the commodities asset class, at least for the next two and a half years, TRAKRS may make sense if you don’t mind the additional operational hassles that investing in them entails. For more information, visit www.trakrs.com.
What About Gold?
The long anticipated launch of a U.S. gold-based Exchange Traded Fund finally happened in November, and quickly attracted over $1 billion in assets. Trading under the ticker GLD, and with an expense ratio of just 0.48%, the new ETF resembles similar offerings already available in the U.K., Australia and South Africa. The ETFs are designed to trade at a price equal to ten percent of the prevailing price for an ounce of gold. In addition, they are backed by an amount of physical gold equal to ten percent of the notional physical volume represented by the ETF. For example, if the total value of the ETFs outstanding represent 1,000 ounces of gold, the shares would be backed by 100 ounces of physical gold. Supporters of this new product claim that it is much cheaper to own gold this way, because you avoid many costs associated with storing and safeguarding the physical product (e.g., gold coins you directly purchase and hold in a bank safety deposit box). Detractors claim that because the ETFs are only fractionally backed by gold there is still a large difference between this new financial product and, for example, having a pile of gold coins in your safety deposit box.
We also have concerns about this new product, but they are of a different nature. First, as described in this month's letter to the editor, there is a significant difference between the source of returns from owning a physical commodity versus owning a futures contract on that commodity. In our opinion, direct ownership of a physical commodity is a more speculative investment than a continuously rolled over futures position. In other words, as a financial investment, we'd be more comfortable with an ETF tied to the gold futures contract that trades on the New York Mercantile Exchange.
Our second concern is with the treatment of gold as a separate asset class. We have included it as part of the broader commodities asset class. Our reasoning is as follows. Between 1976 and 2000, the total return on gold, in U.S. dollars, had a very low correlation to the total return on other asset classes, including (as measured by the Goldman Sachs Commodities Index, in which gold has a very low weighting). The specific correlations were as follows: U.S. Investment Grade Bonds (-.01); U.S. High Yield Bonds (.03); U.S. Commercial Real Estate Investment Trusts (.05); Goldman Sachs Commodities Index (.25); U.S. Equities (.04); Foreign Equities (EAFE) (.22). These low correlations suggest that a strong argument can be made for gold as a separate asset class.
On the other hand, over the same period, the average annual return on gold was much lower, and the standard deviation of returns was much higher, than it was for these other asset classes. On balance, this more than offset the advantages of gold's low correlations, and caused most asset allocation software programs (including ours) to reject an allocation to gold. However, this still leaves unanswered the question of whether there exists a set of circumstances under which an allocation to gold would make sense.
As we have written, we like to think of the economy as being in one of three states: normal (cyclically varying real growth with low to moderate inflation), high inflation, and deflation. Traditionally, people looked at gold as a hedge against inflation. However, in recent years the total returns on gold have not been closely correlated with inflation. Broadly speaking, this has weakened the argument for investing in gold, and led people to look to commodities (more broadly defined) and real return bonds as hedges against inflation risk. The remaining question is therefore how gold would perform under a period of extended deflation. The traditional asset of choice for hedging against this risk is investment grade bonds. Moreover, as a commodity, one would generally expect to see the price of gold (and the returns on holding it) decline during a period of deflation.
However, this argument neglects gold's other historical role as a store of value and unit of exchange (note that this only applies to physical, monetary gold -- i.e., coins). One could therefore envision a scenario in which prolonged deflation (and expectations of an eventual sharp reflation) led people to lose faith in the long-term value of a currency (and/or a domestic debt market). Under these circumstances, in its role as a monetary unit, gold's attractiveness (and the returns earned by holding it) might sharply increase. Unfortunately, the world's recent experience with deflation has, thankfully, been so limited that very little data is available to support or contradict this scenario. Given this, we will continue to view gold as a potential tilt within the larger commodities asset class, rather than a separate asset class in itself. Moreover, if one intends to take such a tilt, the most logical implementation strategies seem to be gold futures contracts or gold coins, rather than the current gold ETF.
Last But Not Least: Don’t Forget About Timber
Unfortunately, neither the Goldman Sachs Commodities Index (tracked by the Oppenheimer Real Assets Fund) nor the Dow Jones - AIG Commodities Index (tracked by the Pimco Commodities Real Return Fund) currently includes timber in its mix of commodities. This leaves an investor with a number of alternatives for including this in his or her overall position in the commodities asset class. First, he or she could continuously role over a position in lumber futures contracts. As we have noted, this is our preferred means of investing in commodities (e.g., it is the way the GSCI and DJ/AIG Indexes are contructed). Unfortunately, the operational details involved put this approach beyond the practical reach of most investors. A second alternative would be to buy a mutual fund that only invests in the common stocks of companies involved in the forest products industry. An example of this is the Fidelity Select paper and Forest Products Fund (FSPFX). However, a fund like FSPFX contains exposure not only to timber prices, but also to the overall equity market. As such, during different periods, one or the other factor may dominate in determining the fund's return.
A third alternative for investing in timber would be to purchase one or more of the growing number of large timber holdings that have been structured as real estate investment trusts (REITS), and/or master limited partnerships (MLPs). Plum Creek Timber (PCL) is one of the largest of these, with eight million acres of holdings divided between northern and southern forests. Rayonier (RYN) is another, with two million acres in holdings. How have these direct timber investments performed over time? Between 1989 and 2003, Plum Creek delivered more than twice the return of the S&P500, though with about half again as much risk. Over a longer period (1957 to 2003) one index of raw timber prices (maintained by the state of Indiana) has delivered a real compound annual return of 1.2%.
Based on these considerations, we can understand why an investor might want to include timber as part of his or her allocation to the commodities asset class. Until timber is included in the main futures-based commodities indexes, we believe that the best way to do this is via direct investments in vehicles like PCL and RYN.
Update: From our March, 2005 Issue
A Deeper Look at Commodities Returns
Historically, commodities have delivered real returns roughly equal those on domestic equity. While their volatility has typically been higher, their correlation with most other asset classes has been very low (with the exception of real return bonds). As a result, including commodities has historically provided substantial diversification benefits (see, for example, "Strategic and Tactical Allocation to Commodities for Retirement Savings Schemes" by Nijman and Swinkels).
With the benefits of commodities as an asset class being recognized by more investors, they are also receiving increasing attention from academic researchers. A number of interesting papers have recently been published, which we will review in this note. To understand the arguments, made in them, we need to review some rather arcane technical aspects of commodity futures pricing. Please bear with us!
A futures contract is a promise entered into with a commodities exchange (e.g., the New York Mercantile Exchange) to deliver (when you sell a futures contract) or receive (when you buy a futures contract) a specified quantity of a specified commodity at a specified date in the future -- for example, 1,000 barrels of oil, at a price of $40 per barrel, 12 months from today. However, rather than delivering or receiving the actual physical commodity, most sellers and buyers of futures contracts buy or sell an offsetting contract at the specified maturity date.
Commodity indexes, such as the Goldman Sachs Commodities Index, or the Dow Jones-AIG Commodities Index, are weighted baskets of different commodity futures. Funds that track these indexes (e.g., the Oppenheimer Real Assets Fund -- QRAAX or the PIMCO Commodities Real Real Return Fund -- PCRDX) are net buyers of commodities futures contracts (this is also known as being "long futures").
The return generating process within a commodity index fund is complex, and has three parts. The first is the return they earn on their excess cash. A futures contracts is purchased on margin, which means that you pay less than 100 percent of its face value when you initially buy it. As long as this futures contract's price is lower than the spot price, the margin requirement is usually quite small (if the spot price drops below the futures contract's price, the commodities exchange may demand more cash -- a so-called "margin call" -- to limit its credit risk exposure). For example, let's say an investor buys a share of a commodity index fund for $100. Let's further assume that the fund manager has to spend only $10 to purchase $100 of futures contracts on the commodities that make up the index being tracked. That leaves $90 in excess cash that can be invested to earn a return that is unrelated to the earnings on the futures contracts. Theoretically, there is no constraint on where that $90 might be invested. It could, for example, be invested in emerging market equity shares. However, this might create marketing problems for the commodity index fund, since its returns would therefore be a mix of two volatile asset classes – commodities and emerging markets equities. To avoid this problem, commodity index funds typically invest their excess cash in low volatility asset classes, such as nominal or real return government bonds.
The second source of the return on a commodity index fund is the diversification benefit that results from investing in a mix of different commodities whose returns have very low correlations with each other.
The third source of return for a commodities index fund is the insurance premium it earns from being a buyer of commodity futures contracts. Let's look more closely at the theoretical source of this return.
The classical argument for the use of futures contracts starts with a commodity producer, who faces relatively high fixed costs (e.g., as would be the case for a farmer, mine owner, or oil production company). This producer is assumed to sell the commodity to an intermediate customer who faces relatively variable demand from final customers for his goods. In addition, the intermediate customer is assumed to have lower fixed costs than the commodity producer. Swings in demand by final customers cause the intermediate customer to cut back his purchases from the commodity producer. Given that the producer's production is relatively fixed, this causes big swings in prices for the commodity. Unfortunately, because the commodity producer has high fixed costs, these price swings can force it into bankruptcy. Commodity producers therefore have a strong interest in hedging the risk they face from price swings.
Obviously, one way to limit these swings would be to insert a storage operator into this simple system, who buys from the producer when demand falls, and sells this accumulated inventory to the intermediate customer when demand rises. The operations of this storage provider balance swings in demand, and in so doing, keep the price for the commodity relatively stable. However, two obstacles may prevent a storage operator from entering the system. First, it may be technically impossible (or very expensive) to store the commodity in question for anything other than a short period. For example, this is the case with oil. While it is feasible to build tank farms to maintain small oil inventories, the amount of tanks that would be necessary for large inventories is prohibitively expensive. This means that most of the physical response to swings in oil demand comes not from inventory in tanks, but from changes in the amount of oil that is pumped from the ground. Apart from physical storage challenges, there may also be financial ones. If the price swings over time in a given commodity are large, so too must be the amount of equity in a storage operator's capital structure. If the capital costs of building the storage facility are large, this can create financial barriers to entering the storage business.
The financial futures markets provide an efficient alternative to physical storage as a means of managing the price risk facing commodity producers. When futures markets exist, producers can sell their production forward (i.e., sell a futures contract) to lock-in a price (and hopefully a profit) for their company. When the futures contract matures, the producer simply buys an offsetting contract, and delivers the physical commodity to a physical buyer. Theoretically, any loss or gain on the physical commodity should be almost completely offset by the loss or gain on the futures contract. Why almost? Because the buyer of the futures contract is, in effect, providing price insurance to the commodity producer, and needs to be compensated for taking this risk. However, futures contracts to not include an explicit risk premium; instead, this is created by having the futures contract trade at a lower price than the "spot price" (i.e., the price at which the physical commodity can be purchased for immediate delivery). In the arcane language of the commodities markets, the fact that futures contracts trade at lower price than the spot price is known as "normal backwardation", or simply "backwardation." The buyer of the futures contract therefore earns her risk premium in the form of the difference between the lower price at which she buys it, and the higher price at which she sells it (which assumes that, as the futures contract nears maturity, its value rises to match the level of the spot market price).
Simple enough, right? However, as is true in most areas of life, sometimes that futures markets don't behave according to this theory. Specifically, there are times when the spot price is actually lower than the futures price. In the language of commodities, this is known as "contango", which has nothing to do with a dance from Argentina.
Why might a commodity be in contango? First, a market may experience an unexpected increase in supply (e.g., think of oil producing nations breaking their OPEC production quotas), or an unexpected fall in demand (e.g., the impact of the discovery of mad cow disease in Canadian beef). In this case, one would expect adjustments in supply and demand that returned the market to a state of "normal backwardation." An alternative theory has also been advanced that proposes the existence of more consistently "contangoed" markets. For example, consider a breakfast cereal producer that, through effective marketing, has created a steady demand for its products. One of its most important input costs is grains -- e.g., wheat and corn. To limit variation in its reported profits, it may purchase futures contracts to lock in its grain costs. Assuming there are also financial investors bidding for a finite supply of futures contracts being sold by farmers, the price of the futures contract may rise above the spot price for the grains. In this situation, the net provider of price insurance would not be the party buying the futures contracts, but rather the party selling them, since at maturity (when the futures price moves to the level of the spot price) the seller will be able to purchase an offsetting futures contract at a lower price than he received for the one he originally sold.
Careful readers will, like us, probably be shaking their heads at this argument. Why? Because it presumes that, having just received a negative risk premium on the futures contracts they bought, the financial investors would repeat their mistake. We think this is unlikely to happen, as such investors would, over time, abandon a market with such unattractive structural features.
One of the most interesting of the recent papers on commodity investing is ""The Tactical and Strategic Value of Commodity Futures" by Erb and Harvey. In their paper, the authors analyze the returns earned on long positions in different commodity futures between December, 1982 and May, 2004. They assume that the presence of positive average returns implies a market that was, on average, backwardated, while negative average returns would imply a market that was, on average, contangoed. Specifically, Erb and Harvey measure the "excess return" on the commodity futures above the return on three month treasury bills. Assuming the latter is approximately equivalent to inflation, Erb and Harvey's measure of "excess returns" is roughly equal to real returns.
The following table summarizes their findings:
| Commodity | Average Annual Excess Return over 3 mos. T-Bill | Standard Deviation of Excess Returns |
| Heating Oil | 10.51% | 32.55% |
| Live Cattle | 5.94% | 13.98% |
| Live Hogs | 0.17% | 24.21% |
| Wheat | -3.32% | 21.05% |
| Corn | -3.32% | 22.65% |
| Soybeans | 1.92% | 21.49% |
| Sugar | 3.69% | 38.65% |
| Coffee | 0.85% | 39.69% |
| Cotton | 2.60% | 22.64% |
| Gold | -4.81% | 14.36% |
| Silver | -5.30% | 25.03% |
| Copper | 9.15% | 25.69% |
| GSCI Index | 5.81% | 16.97% |
This table makes a number of interesting points. Erb and Harvey point to the much lower standard deviation of returns on the GSCI compared to those on the twelve individual commodities as evidence of the substantial diversification return provided by the index.
They also note that support for "normal backwardation" (which would be indicated by a positive average excess return) appears to be uneven, with some commodities apparently more often in contango. This leads them to conclude that "long-only" commodities futures indexes like the Goldman Sachs Commodities Index or the Dow Jones AIG Commodities Index are inefficient ways to invest in commodities futures. This leads them to two proposals. The first is for a "long/short" commodity fund that would permanently have long positions in commodities that, historically, have on average been backwardated, and short positions in those commodities that have, on average, been contangoed. In both cases, the commodity fund would be earning a premium for providing price insurance.
They also show how an active approach could be implemented at the index level, using GSCI futures. They note that "since the inception of GSCI futures trading in 1982, the GSCI has been backwardated as often as it has been in contango. The annualized payoff from buying the GSCI when the term structure is backwardated is 11.2%. However, when the term structure is contangoed, the annualized excess return is negative (5.0%)." This leads them to propose an active management strategy that buys GSCI futures when the index is backwardated (as evidenced by a positive return over the previous year), and sells them when it is contangoed (as evidenced by a negative lagged return). On a backtested basis, they find that this strategy would have generated an excess return of 8.2% per annum over the period they studied.
Besides Erb and Harvey, other researchers have also identified commodities active management strategies that would have delivered higher returns than the index had they been used in the past. For example, in "Dynamic Commodity Timing Strategies" by Vrugt, Bauer, Molenaar, and Steenkamp the authors investigate various timing strategies, models and indicators that could be used in an active management strategy. They conclude that "variation in commodity future returns is sufficiently predictable to be exploited by a realistic timing strategy." Similar papers include "An Anatomy of Futures Returns: Risk Premiums and Trading Strategies" by de Roon, van den Goorbergh, and Nijman and "Conditional Means, Volatilities, and Correlations in Commodity Futures Markets" by Chong and Miffre (which also finds that long-only commodities strategies provide excellent diversification benefits during recessions).
However, another paper provides an important cautionary tale about the dangers of depending too heavily on the accuracy of a relatively short period of backetested results as a guide to a commodities active management strategy. In "Backwardation in Energy Futures Markets: Metallgesellschaft Revisited" Carupat and Deaves describe how, in early 1990s, MG lost an enormous amount of money using futures to hedge its long term gasoline and heating oil physical delivery contracts (the actual loss was caused by the company selling at a loss futures contracts on which it had had to post rapidly increasing amounts of cash collateral). MG's strategy was premised on the continuation of backwardation in crude oil futures markets. The authors find that, using data available to MG (1984 to 1992), risk of contango appeared acceptable to MG at the time the strategy was undertaken -- the probability of negative returns at least as large as the ones MG actually experienced was "only" 3.77%, based on backtesting the available data. However, when the authors backtested a longer data set that extended to 2000, they found that the estimated probability of a disaster scenario rose to 7.32%.
So where does this leave us? As we have repeatedly noted, the underlying process that generates asset class returns is very complex, and broadly composed of two subparts: a fundamental process (e.g., that generates company investments, cash flows, and reported earnings) and a behavioral process (e.g., the reaction of investors to changes in the fundamentals, plus their anticipation of how other investors will react). Moreover, the way this process operates changes over time (i.e., one or both sub-parts is "non-stationary"). It is therefore non-linear, and extremely difficult to forecast accurately for any length of time. This is no less true of commodities than it is of any other asset class. We are therefore suspicious of all claims that make the potential gains from active management in commodities seem too easy. After all, if academics have discovered them, why wouldn't hedge funds use them and arbitrage away their expected excess returns?
However, this still leaves us with the question of whether the GSCI or DJAIG are appropriately designed indexes for investing in commodity futures. We believe they are, for two reasons. First, as Erb and Harvey note, both are heavily weighted towards commodities that, in the past, have typically been backwardated.
Second, think about the implications for commodity markets of the changes now underway in the world economy. The most important of these has been the rapid growth in demand for commodities by fast growing developing countries, and especially China. While this demand growth has driven up the price of many commodities, it has also caused an expansion of supply. This means that if you own a commodity producing company today, you are looking at a world with higher supply, in which the marginal demand is provided by developing countries whose economic growth rates (and hence commodity demand) are significantly more volatile than those of developed countries. While some developed country commodity markets may still have demand that is relatively more stable than supply (as in our cereal producer example), this is much less likely than before to be the case for the global market as a whole for that commodity. In sum, our view of the global economy leads us to conclude that backwardation in commodity markets -- and hence, positive insurance premiums for long-only commodity index funds -- should be more common than in the past. For both of the reasons we have cited, we continue to believe that diversified commodity index funds are an efficient and prudent way for most investors to gain exposure to the commodities asset class.