In a separate posting, I addressed the thoughtful criticisms of commodity futures by Bill Bernstein. In this one, I want to address a claim that is being made by some that the case for commodities rests almost entirely on their usefulness during the 1973-4 bear market in stocks, during which commodities soared anywhere from 77% to well over 100%, depending on which index you used (I actually used the one that had the SMALLEST increase during that time).
This is not so. Even ignoring those years, commodity futures have proven to be a useful diversifier. Of course, ignoring those years is a hard position to defend: one can make the case that home insurance is a total waste of money if we ignore the day your house burned down. But let's go along with this assumption.
Returns - The case for commodity futures rests principally on its negative correlation to stocks (and bonds and REITs), but I've cited its equity-like returns in my discussions. Those of us presenting historical returns have been accused of biasing the return data on commodity futures by starting in 1972. Although the primary reason is the absence of indexes that strongly resemble the Dow Jones AIG for periods before the 1970s, I probably should plead guilty to that charge and admit that, if we go back as far as the longest data series currently available, the annualized returns on commodity futures are lower than stocks. To give credit where credit is due, the longest data series available at no charge on the Internet can be found at:
raddr-pages.com/research/...utures.htm
It goes back to 1957, and is accompanied by a very good presentation on the subject (which goes along with my preference for an equal weighted index, since it shows that the diversification benefit is greater when US stocks are represented by smaller value stocks). It also notes that the return on commodity futures averaged 1.6% per year less than stocks from 1957 to 2003 (they've been about the same overall for the last 3 years, so the comparison doesn't change if we extend the data to 2006). I could argue that starting at 1957 is just as arbitrary, as this starts just after the end of the massive war and post-war inflation (and the data on the Commodity Research Board Index during that time suggests returns vastly exceeding stocks during the 1940s and early 1950s, although a formal futures index is yet to be constructed, as far as I know). But let's leave it at this starting point, which still has a large return margin over bonds: around 2% per year, depending on which bond index is used. Of course, PIMCO uses TIPS as collateral, and the historical data I included was based on T-Bills, so the advantage for commodity futures actually would have been greater.
What critics have then suggested is that this entire margin disappears if we remove 1973-1974. Now, besides wondering whether I'm now allowed to remove the two best years from bonds as well (fair is fair), let's consider what the effect would be of this removal. The big runup, which resulted from the OPEC oil embargo, and which was directly related to the recession that brought stocks down (this, of course, represents the very negative correlation between commodities and stocks I've been discussing), caused the commodity index I use in calculations to rise 77%. If that hadn't happened, the price index today would be ... the same as it is today: there would have been no effect on total returns.
Does that strike you as insane? Think about it: let's say someone computing stock market returns claimed that the tech bubble of 1998 to 2000 was so unusual that it should be ignored in calculating long-run stock returns. Would it then make sense to continue including the tech crash of 2000 to 2002? Of course not. If the artificial runup in prices during the 1970s, resulting primarily from Nixon's price controls that were finally lifted at the end of the Carter administration, hadn't occurred, the decontrol-related decline in the 1980s wouldn't have occurred either. If oil prices hadn't tripled in the early 1970s, does that mean oil would be priced at $24 a barrel around the time of this writing instead of $72? Of course not: the price today reflects conditions today. Those who want to remove the runup in commodities in 1973-4 forgot that, absent that runup, prices would have been lower, so subsequent corrections of price wouldn't have occurred.
Let me put it another way, mathematically: to calculate the total return on a commodity price index from the beginning of 1957 until the end of 2003, we only need to know the value of the index at January 1, 1957 and December 31, 2003. We don't care how much it fluctuated during those 47 years because of short-term conditions in various limited time periods: the values at the start and finish reflect the conditions at the start and finish, respectively. Some people believe commodities are overpriced today: the same argument can be made about the conditions in early 1957, because of the WWII and post-war inflation, so perhaps we have a representative half century as far as the normal returns on commodity futures. In any event, the total return advantage over bonds doesn't disappear by removing 1973-4.
(By the way, some have suggested that, since commodities were considered money for nearly all of human history prior to the abandonment of the gold standard in the last century, they don't need to earn more than other cash equivalents: if a receipt for gold was considered to be not only as safe as cash, but actually WAS cash, in 1907, why should it not be considered a safe claim in 2007? In fact, a claim on a basket of commodities is more stable than a receipt only redeemable in gold. I AM prepared for the possibility of eating crow someday on commodity futures long-run returns, should the evidence justify it. That won't, however, remove the case for commodities based on correlation, so let's continue.)
Correlation - Even if commodity futures earned no more than bonds, they would have a place in asset allocation theory. Investments with a positive expected return and negative correlation to other investments are the holy grail of Modern Portfolio Theory.
For those who have taken a look at the link in this message, you will notice a -.72 correlation of the 5-year returns of commodities and small cap value stocks (as I stressed in my original article, the short-term returns are positively correlated, and the diversification case is based on longer time periods). This is about as close to a perfectly negative correlation as any two investments that each have a positive expected return can be expected to have. Even though the 5-year negative correlation is highest, I'd like to use the 3-year, which is high enough. Although raddr didn't compute it, the 3-year correlation of the data in his piece came to -.61, and my computation of the 3-year correlation of the Equal Weight S&P 500 Index (a less specialized index than smallcap value, but less concentrated in a few large stocks than the traditional, market-weighted S&P 500 index) came to -.54 (the equal weight index returns have been computed by Leuthold Research, which at the present time has chosen not to allow a posting of annual data without fee). Using the traditional S&P index, the negative correlation is -.30 (at many times, the largest companies dominating the index are commodity producers, such as oil companies, making this a less useful combination, but still significantly negative. I used -.40 in my original article, which was a rounded average of the traditional and equal weight S&P 500 index correlations with futures, because I know many people won't follow my advice to use the latter to represent domestic stocks rather than the former).
Now, here is the issue: just as some people have wanted to artificially remove 1973-1974 in calculating returns, others have claimed that the negative correlation wouldn't exist if those years were removed.
Well, although I once again object to such an arbitrary action, unless I'm also permitted to remove the highest correlation years, let me pretend this is fair. I removed all 3-year comparisons that included either 1973 or 1974 (that would be the three years ended 1973, 1974, 1975, and 1976). The negative correlation drops, of course, if you remove the most extreme opposite performances, but it only drops from -.61 to -.41 for the data in the link, from -.54 to -.37 for the equal weight S&P vs commodity futures, and from -.30 to -.12 if we use the traditional S&P 500 index. In all 3 cases, even after this arbitrary adjustment, the correlations remain negative. The fact is that commodity futures would be a useful diversifier even if they had a moderately positive correlation: anything less than a perfect +1.00 means there is some diversification benefit. But even the most biased attempt to destroy the case for negative commodity futures correlation fails.
(BTW, although it was not a part of my original discussion, note that the data in the provided link shows the standard deviation of the commodity futures was lower than for stocks, meaning a portfolio with both would have had lower overall volatility even in the absence of the diversification effect.)
Let me address a couple of other objections I have heard. You'll notice that the 3-year and 5-year correlations are extremely negative, but it stops there. Probably due to the historic mean reversion of returns from one 3-5 year period to the next (good 3-5 year periods had a slight, though not absolute, tendency to be followed by weaker such periods, and vice versa), correlations over longer periods of time start to get closer again. So a 10-year comparison won't find the correlation as extreme, and over 30-year time periods, most good investments in a growing economy ought to do well, and diversification into commodity futures might be less needed, especially for someone who is dollar cost averaging new money each year, and might actually benefit from a more volatile portfolio. Thus, I'm not as concerned about younger investors ignoring commodity futures as I am with older investors, or wealthy investors who can handle more volatility than middle class investors. But I work with many clients at or near retirement, and not all of them are filthy rich: they are not as able to sustain, either objectively or emotionally, extremely negative 3-year and 5-year periods, and they still need good growth to be able to withdraw the sums they'll need to live comfortably for the rest of their lives.
(Of course, even the young and the wealthy might want to consider protection from a "Black Swan" event: a massive increase in inflation that might well seriously damage the economy but cause commodity prices to soar. For example, should we ever find ourselves someday with a president who spends like there's no tomorrow while simultaneously cutting taxes, we might want to consider the outside possibility of high inflation with economic stagnation).
Some people claim correlations between commodity futures and stocks have risen in the last 10 years. Maybe they have (actually, they've fallen for the traditional S&P 500 index and risen for the equal weight index and smallcap value index), but there are two problems with this argument:
(1) 10 data points is too tiny a sample to draw any conclusions.
(2) Even if we were to assume the correlation between commodity futures and stocks was now mildly positive rather than extremely negative, it would still provide substantial diversification benefits. International stocks and REITs both have much higher correlations to domestic stocks, yet are considered useful diversifiers.
A few people have pointed out that the return on commodity futures includes the yield on the collateral, which was higher over the past 35 years than is reasonable to expect in the next several, due to relatively high interest rates that prevailed over that time period. True enough, but that was also true of bonds. We are comparing alternative investments, and a scenario that reduces all the returns in the future doesn't change the comparison.
Also remember that the historical data was based on T-Bills, and PIMCO uses TIPS, which should provide a higher yield even after accounting for fund fees. I agree with those who complain that the expense ratios are too high: right now, the extra expected return on TIPS more than covers the expense ratio, but I would expect the expenses to come down due to competitive pressures once commodity futures really are as popular as some critics seem to think they already are today.
One last point, and I promise to stop. I never advised anyone to put 30% of their portfolio into commodity futures: I put in enough disclaimers in the report on my personal portfolio to make that clear. I do so for my own investments because (1) my entire investment portfolio is tax-sheltered and (2) I have known my advisor for more than 50 years and trust him completely: he's brilliant, has a great sense of humor, and is incredibly modest.
(Actually, my mean-variance optimizer says that the minimum variance all-equity portfolio should allocate 39% to commodity futures: for the cynical, starting from 1975 to exclude the 1973-1974 commodity boom only reduces the allocation to 37%, so the real question is why I have so little in commodity futures!)
Okay, seriously. My client portfolios typically have between 10% and 25% in PCRIX or PCRDX, depending on various circumstances such as age, tax bracket, percentage of sheltered assets, and the limitations of employer-based plan assets. For those who wonder what the exclusion of the early 1970s commodity boom does to the recommended allocation for those trying to maximize their risk-adjusted return, the appropriate allocation appears to be around 15%. Ibbotson Associates recommends 10% in its Moderate Allocation Portfolio. A neutral world portfolio that included commodities to the extent they were represented in total wealth would place approximately 13% into commodities. And as has been pointed out by many, the greatest benefit of adding a new asset class comes with the first few percent allocated to it: going from 0% to 10% captures most of the diversification benefit, and is far more helpful than going from 10% to 30%, even if it turns out that 30% is optimal.
By no means do I consider someone to be foolish if they prefer to limit their investment in commodity futures to a much lower level than historical data suggests. But for those who think the appropriate allocation is zero, I would suggest that they think again.
This is not so. Even ignoring those years, commodity futures have proven to be a useful diversifier. Of course, ignoring those years is a hard position to defend: one can make the case that home insurance is a total waste of money if we ignore the day your house burned down. But let's go along with this assumption.
Returns - The case for commodity futures rests principally on its negative correlation to stocks (and bonds and REITs), but I've cited its equity-like returns in my discussions. Those of us presenting historical returns have been accused of biasing the return data on commodity futures by starting in 1972. Although the primary reason is the absence of indexes that strongly resemble the Dow Jones AIG for periods before the 1970s, I probably should plead guilty to that charge and admit that, if we go back as far as the longest data series currently available, the annualized returns on commodity futures are lower than stocks. To give credit where credit is due, the longest data series available at no charge on the Internet can be found at:
raddr-pages.com/research/...utures.htm
It goes back to 1957, and is accompanied by a very good presentation on the subject (which goes along with my preference for an equal weighted index, since it shows that the diversification benefit is greater when US stocks are represented by smaller value stocks). It also notes that the return on commodity futures averaged 1.6% per year less than stocks from 1957 to 2003 (they've been about the same overall for the last 3 years, so the comparison doesn't change if we extend the data to 2006). I could argue that starting at 1957 is just as arbitrary, as this starts just after the end of the massive war and post-war inflation (and the data on the Commodity Research Board Index during that time suggests returns vastly exceeding stocks during the 1940s and early 1950s, although a formal futures index is yet to be constructed, as far as I know). But let's leave it at this starting point, which still has a large return margin over bonds: around 2% per year, depending on which bond index is used. Of course, PIMCO uses TIPS as collateral, and the historical data I included was based on T-Bills, so the advantage for commodity futures actually would have been greater.
What critics have then suggested is that this entire margin disappears if we remove 1973-1974. Now, besides wondering whether I'm now allowed to remove the two best years from bonds as well (fair is fair), let's consider what the effect would be of this removal. The big runup, which resulted from the OPEC oil embargo, and which was directly related to the recession that brought stocks down (this, of course, represents the very negative correlation between commodities and stocks I've been discussing), caused the commodity index I use in calculations to rise 77%. If that hadn't happened, the price index today would be ... the same as it is today: there would have been no effect on total returns.
Does that strike you as insane? Think about it: let's say someone computing stock market returns claimed that the tech bubble of 1998 to 2000 was so unusual that it should be ignored in calculating long-run stock returns. Would it then make sense to continue including the tech crash of 2000 to 2002? Of course not. If the artificial runup in prices during the 1970s, resulting primarily from Nixon's price controls that were finally lifted at the end of the Carter administration, hadn't occurred, the decontrol-related decline in the 1980s wouldn't have occurred either. If oil prices hadn't tripled in the early 1970s, does that mean oil would be priced at $24 a barrel around the time of this writing instead of $72? Of course not: the price today reflects conditions today. Those who want to remove the runup in commodities in 1973-4 forgot that, absent that runup, prices would have been lower, so subsequent corrections of price wouldn't have occurred.
Let me put it another way, mathematically: to calculate the total return on a commodity price index from the beginning of 1957 until the end of 2003, we only need to know the value of the index at January 1, 1957 and December 31, 2003. We don't care how much it fluctuated during those 47 years because of short-term conditions in various limited time periods: the values at the start and finish reflect the conditions at the start and finish, respectively. Some people believe commodities are overpriced today: the same argument can be made about the conditions in early 1957, because of the WWII and post-war inflation, so perhaps we have a representative half century as far as the normal returns on commodity futures. In any event, the total return advantage over bonds doesn't disappear by removing 1973-4.
(By the way, some have suggested that, since commodities were considered money for nearly all of human history prior to the abandonment of the gold standard in the last century, they don't need to earn more than other cash equivalents: if a receipt for gold was considered to be not only as safe as cash, but actually WAS cash, in 1907, why should it not be considered a safe claim in 2007? In fact, a claim on a basket of commodities is more stable than a receipt only redeemable in gold. I AM prepared for the possibility of eating crow someday on commodity futures long-run returns, should the evidence justify it. That won't, however, remove the case for commodities based on correlation, so let's continue.)
Correlation - Even if commodity futures earned no more than bonds, they would have a place in asset allocation theory. Investments with a positive expected return and negative correlation to other investments are the holy grail of Modern Portfolio Theory.
For those who have taken a look at the link in this message, you will notice a -.72 correlation of the 5-year returns of commodities and small cap value stocks (as I stressed in my original article, the short-term returns are positively correlated, and the diversification case is based on longer time periods). This is about as close to a perfectly negative correlation as any two investments that each have a positive expected return can be expected to have. Even though the 5-year negative correlation is highest, I'd like to use the 3-year, which is high enough. Although raddr didn't compute it, the 3-year correlation of the data in his piece came to -.61, and my computation of the 3-year correlation of the Equal Weight S&P 500 Index (a less specialized index than smallcap value, but less concentrated in a few large stocks than the traditional, market-weighted S&P 500 index) came to -.54 (the equal weight index returns have been computed by Leuthold Research, which at the present time has chosen not to allow a posting of annual data without fee). Using the traditional S&P index, the negative correlation is -.30 (at many times, the largest companies dominating the index are commodity producers, such as oil companies, making this a less useful combination, but still significantly negative. I used -.40 in my original article, which was a rounded average of the traditional and equal weight S&P 500 index correlations with futures, because I know many people won't follow my advice to use the latter to represent domestic stocks rather than the former).
Now, here is the issue: just as some people have wanted to artificially remove 1973-1974 in calculating returns, others have claimed that the negative correlation wouldn't exist if those years were removed.
Well, although I once again object to such an arbitrary action, unless I'm also permitted to remove the highest correlation years, let me pretend this is fair. I removed all 3-year comparisons that included either 1973 or 1974 (that would be the three years ended 1973, 1974, 1975, and 1976). The negative correlation drops, of course, if you remove the most extreme opposite performances, but it only drops from -.61 to -.41 for the data in the link, from -.54 to -.37 for the equal weight S&P vs commodity futures, and from -.30 to -.12 if we use the traditional S&P 500 index. In all 3 cases, even after this arbitrary adjustment, the correlations remain negative. The fact is that commodity futures would be a useful diversifier even if they had a moderately positive correlation: anything less than a perfect +1.00 means there is some diversification benefit. But even the most biased attempt to destroy the case for negative commodity futures correlation fails.
(BTW, although it was not a part of my original discussion, note that the data in the provided link shows the standard deviation of the commodity futures was lower than for stocks, meaning a portfolio with both would have had lower overall volatility even in the absence of the diversification effect.)
Let me address a couple of other objections I have heard. You'll notice that the 3-year and 5-year correlations are extremely negative, but it stops there. Probably due to the historic mean reversion of returns from one 3-5 year period to the next (good 3-5 year periods had a slight, though not absolute, tendency to be followed by weaker such periods, and vice versa), correlations over longer periods of time start to get closer again. So a 10-year comparison won't find the correlation as extreme, and over 30-year time periods, most good investments in a growing economy ought to do well, and diversification into commodity futures might be less needed, especially for someone who is dollar cost averaging new money each year, and might actually benefit from a more volatile portfolio. Thus, I'm not as concerned about younger investors ignoring commodity futures as I am with older investors, or wealthy investors who can handle more volatility than middle class investors. But I work with many clients at or near retirement, and not all of them are filthy rich: they are not as able to sustain, either objectively or emotionally, extremely negative 3-year and 5-year periods, and they still need good growth to be able to withdraw the sums they'll need to live comfortably for the rest of their lives.
(Of course, even the young and the wealthy might want to consider protection from a "Black Swan" event: a massive increase in inflation that might well seriously damage the economy but cause commodity prices to soar. For example, should we ever find ourselves someday with a president who spends like there's no tomorrow while simultaneously cutting taxes, we might want to consider the outside possibility of high inflation with economic stagnation).
Some people claim correlations between commodity futures and stocks have risen in the last 10 years. Maybe they have (actually, they've fallen for the traditional S&P 500 index and risen for the equal weight index and smallcap value index), but there are two problems with this argument:
(1) 10 data points is too tiny a sample to draw any conclusions.
(2) Even if we were to assume the correlation between commodity futures and stocks was now mildly positive rather than extremely negative, it would still provide substantial diversification benefits. International stocks and REITs both have much higher correlations to domestic stocks, yet are considered useful diversifiers.
A few people have pointed out that the return on commodity futures includes the yield on the collateral, which was higher over the past 35 years than is reasonable to expect in the next several, due to relatively high interest rates that prevailed over that time period. True enough, but that was also true of bonds. We are comparing alternative investments, and a scenario that reduces all the returns in the future doesn't change the comparison.
Also remember that the historical data was based on T-Bills, and PIMCO uses TIPS, which should provide a higher yield even after accounting for fund fees. I agree with those who complain that the expense ratios are too high: right now, the extra expected return on TIPS more than covers the expense ratio, but I would expect the expenses to come down due to competitive pressures once commodity futures really are as popular as some critics seem to think they already are today.
One last point, and I promise to stop. I never advised anyone to put 30% of their portfolio into commodity futures: I put in enough disclaimers in the report on my personal portfolio to make that clear. I do so for my own investments because (1) my entire investment portfolio is tax-sheltered and (2) I have known my advisor for more than 50 years and trust him completely: he's brilliant, has a great sense of humor, and is incredibly modest.
(Actually, my mean-variance optimizer says that the minimum variance all-equity portfolio should allocate 39% to commodity futures: for the cynical, starting from 1975 to exclude the 1973-1974 commodity boom only reduces the allocation to 37%, so the real question is why I have so little in commodity futures!)
Okay, seriously. My client portfolios typically have between 10% and 25% in PCRIX or PCRDX, depending on various circumstances such as age, tax bracket, percentage of sheltered assets, and the limitations of employer-based plan assets. For those who wonder what the exclusion of the early 1970s commodity boom does to the recommended allocation for those trying to maximize their risk-adjusted return, the appropriate allocation appears to be around 15%. Ibbotson Associates recommends 10% in its Moderate Allocation Portfolio. A neutral world portfolio that included commodities to the extent they were represented in total wealth would place approximately 13% into commodities. And as has been pointed out by many, the greatest benefit of adding a new asset class comes with the first few percent allocated to it: going from 0% to 10% captures most of the diversification benefit, and is far more helpful than going from 10% to 30%, even if it turns out that 30% is optimal.
By no means do I consider someone to be foolish if they prefer to limit their investment in commodity futures to a much lower level than historical data suggests. But for those who think the appropriate allocation is zero, I would suggest that they think again.
