Rick and Larry battle it out with their opposing views on using commodities in a well diversified portfolio:
- How useful are commodities in a long-term asset allocation strategy?
- Is negative correlation alone reason enough for an allocation to commodities?
HardAssetsInvestor (HAI): Welcome, gentlemen. There's an ongoing debate among financial advisors about the usefulness of commodities in a long-term asset allocation. Most of the debate centers on returns. Can commodities, like equities, be reasonably expected to offer a return for a portfolio?
Rick Ferri: Physical commodities, of course, are uninvestable unless you own oil storage tanks and grain bins.
Research published by Erb and Harvey in "The Tactical and Strategic Value of Commodity Futures" indicates that commodities futures' real expected return-that is, the return net of inflation-is actually zero.
Erb and Harvey show that trading strategy is the sole source of returns in commodity futures indexes. In other words, it's the way an index provider constructs and rebalances an index that hypothetically created real returns in the past. Unlike stocks and bonds, simply buying and holding a diversified basket of futures contracts does not create real returns.
Larry Swedroe: It's true that commodities themselves have no expected real return. That's a pretty good reason to avoid investing in them. However, collateralized commodity futures, or CCFs, are another story.
The PIMCO Commodity Real Return Strategy fund, as an example, has total costs of about 1%. The portfolio managers use Treasury Inflation-Protected Securities (TIPS) as collateral to support the embedded commodity index derivatives. The real return on TIPS is about 2%, so the PIMCO fund provides a return, net of costs and inflation, of about 1%. That's above the real historic return of the benchmark one-month T-bill. Unless one forecasts persistent contango, where futures trade higher than the spot market, one should expect a real return from the PIMCO fund.
Rick Ferri: PIMCO's use of TIPS may boost returns, but that has nothing to do with commodities. PIMCO's capitalizing upon the spread between T-Bills and TIPS. Since the T-Bill yield is imbedded in commodity futures prices, PIMCO is essentially shorting T-bills and buying TIPS. That's interest rate arbitrage, not commodity investing.
HAI: Financial advisors often point to correlation - or more properly, noncorrelation - as a reason to include a commodities allocation in a portfolio. Where do you stand on the correlation issue? Would negative correlations alone be reason enough for an allocation?
Swedroe: Futures, or more specifically, CCFs, are one of the rare asset classes that have negative correlation to both stocks and bonds. That makes them excellent risk diversifiers. A negative correlating asset acts just like portfolio insurance because it tends to produce higher-than-average returns when the other asset is producing lower-than-average returns.
In each of the nine years since 1970 that long-term bonds produced negative returns, CCFs produced positive returns. The average is about 30%. And, in the eight years of negative stock returns, since then, CCFs produced positive returns six times, garnering an average gain of 23% for all eight years.
The negative correlation is easily explained and quite logical. CCFs are positively correlated to unexpected inflation while stocks and bonds are negatively correlated. Hence the very strong performance against bonds. But CCFs also provide a hedge against event risks. Some event risks such as the oil crisis of the ‘70s impact stocks and commodities differently. Others, notably 9/11, cause the markets to react similarly. Put simply, if an event creates inflationary-type recession, then it is good for CCFs but bad for stocks. If an event creates a deflationary environment, than you'll see the correlation of stocks and CCFs rise.
Negative correlation is not enough reason for making an investment. One should also consider the return. But the mistake so many people make is considering return in isolation. The question really centers on the impact of adding the commodity return on the overall portfolio. Harry Markowitz basically won a Nobel Prize dealing with this issue. He showed that adding risky assets can actually increase returns without increasing risk, depending on volatility and correlations.
With negative correlation we actually like higher volatility because, with rebalancing, the diversification return increases. You buy at lower lows and sell at higher highs. Economic theory states that assets with negative correlation should have less than the riskless return because they act like insurance.
Ferri: It's not correct to say that CCFs are negatively correlated. They've been both negatively and positively correlated. It really depends on the period measured. There are many periods, such as the last three years, when the correlation is positive with stocks and bonds. At best you can say correlation varies and is dynamic. That said, the risk reduction benefits of commodities are period-sensitive.
I think correlation analysis is misunderstood and overused in portfolio management. When a new asset class is being investigated for unique risk, correlation is a good test. If two separate asset classes have varying correlation, then there is probably unique risk in those asset classes. Thus, an investor can go to the next level, which is a return analysis. But commodities fail the expected real return test. That's why I don't include them in portfolios. The lower expected returns of commodity funds is not a fair trade for the proven real returns of assets such as stocks.
Low correlation is not, by itself, a good reason to do something. After all, stuffing money in a mattress has no correlation with stocks and bonds, but I don't recommend doing it.
Swedroe: Of course correlations move over time, Rick. What's key is the long-term correlation. Since correlations vary, one should look at when they tend-on average-to turn up and down. The historical data shows the correlations for CCFs tend to turn negative when needed most, especially with respect to bonds.
Ferri: Correlation measures the direction of movement, not the magnitude of the movement, Larry. If one index returned 0.1% and another returned 1000%, the correlation is perfectly positive. Although the direction of return any given year may be similar, actual risk and return can differ by many hundreds of basis points.
HAI: The Erb and Harvey paper argued that individual commodity futures may exhibit no real return, but it left the door open for futures portfolios. Is there, in fact, a diversification effect?
Swedroe: The Erb and Harvey research established that the most reliable part of the CCF return is the diversification return which they estimate at about 3% to 5% per annum. This figure is not, of course, included in the estimated return of CCFs themselves.
Here's what I mean. Using simulated data from 1970 through 2006, a 100% equity, globally diversified, "sliced and diced" portfolio--the kind I recommend in my books--produced a return of 15.9% with a standard deviation, or risk, of 15%. The S&P/Goldman Sachs Commodity Index (GSCI) returned 11.5% with a standard deviation of 18.8%.
Now, why would one want to add an asset with higher risk and a lower return to a portfolio?. Well, here's why: If you added a 5% GSCI position, the portfolio return would have increased to 16% and the standard deviation would have fallen to 14.3%. Keep in mind, this is based on GSCI; the use of the PIMCO fund would have almost certainly produced superior returns.
Ferri: As you say, Larry, the data in your example is simulated. The indexes didn't exist in 1970. Erb and Harvey had to develop their own commodities allocation and trading strategy to come up with that hypothetical return using back-testing. Returns from CCF indexes are all alpha- based.
Individual commodities and futures on their own have no expected return. It's the index provider's strategy that produces alpha in back-testing. So the question really is, "Is it a strategy or an asset class?"
You mentioned Harry Markowitz's work on portfolio diversification, Larry. I don't think Markowitz would agree that a strategy is an asset class.
I'm not refuting your claim of a portfolio effect, but hindsight is 20/20.
Swedroe: You can't say the Erb and Harvey findings are based on a trading strategy, Rick. All they talk about is rebalancing.
Erb and Harvey found that if you went long when in backwardation and short in contango, you could obtain incremental returns. Now that's a trading strategy. I prefer to not follow it because it's not insurance. In other words, you would be lifting your hedge when the costs go up, like eliminating hurricane insurance after Katrina. The evidence shows that this has may have been a more profitable strategy in the long run, but there are many periods when this is simply not so.
The main, if not only, purpose of including CCFs in a portfolio is for insurance. For that reason alone, I don't like the trading strategy idea. Even if I did, there's currently no investable way to capture the strategy. There's no retail product based on the strategy available now.
Let me make a distinction here. There is a big difference between data mining and intelligent fund design. Dimensional Fund Advisors (DFA), a favorite money manager of financial advisors, doesn't include all securities in their well-respected portfolios; they screen out stocks with penny prices, IPOs and the like. That's led to superior returns over simple market cap-weighted indexes. The same thing is true here. Instead of a simple market cap-weighting CCF index, the literature shows that if you create a more equal-weighted index, you will smooth returns by reducing the negative impact of volatility. That's simply intelligent design, not data mining.
In addition, if one limited the allocation to commodities that are persistently in high contango and increased the allocation to those that are more typically in backwardation, the evidence suggests you will get higher returns. That would also be intelligent design, not data mining. And that is exactly what some of the indexes, such as the AIG benchmark, has done.
Ferri: Larry, you say that the low correlation of commodity futures to stocks and bonds helps the overall portfolio. That's only theory, based on hypothetical data. The returns don't include fund fees, which are higher than those of funds built of stocks and bonds. Let's not forget the trading costs associated with a high turnover product built of forward contracts.
More important, no one talks about the opportunity cost of taking money from stocks to invest in commodity index products. Just what is the cost?
There's simply no guarantee that the index you choose will generate an alpha from the investment strateg
The bottom line for me is this: I'm not willing to give up the return from stocks and bonds to invest in an expensive active strategy that has no expected real return except a hypothetical and inconsistent alpha that the index provider says it can generate from an investment strategy.
From the standpoint of portfolio risk reduction, commodity indexes have been shown to reduce risk in a portfolio. The portfolio risk reduction occurs if there is continued low correlation between commodities, stocks and bonds. In that sense, T-bills also offer portfolio risk reduction because the return of T-bills have low correlation with the return of stocks and bonds.
But risk reduction is only one side of the equation. The other side is the return element. One must look at how adding a new asset class will change the expected return of the portfolio before placing that new asset class in the portfolio. Ideally, an asset will have low correlation and high return relative to stocks and bonds. REITs are an excellent example. REITs have high expected returns and low correlations. Unfortunately, commodity indexes offer no such benefit. Aside from a successful trading strategy, commodity indexes do not have a real expected return.