Commodities and (a Lack of) Diversification
Over recent years, commodity indices have become increasingly (positively) correlated with the U.S. stock market, and today, provide little in the way of diversification.
To illustrate, below is the 3-year monthly correlation between the Goldman Sachs commodity index (ETF GSG) and the S&P 500, from 1970. Note right side of graph.
The Goldman Sachs index is the least diversified of the major commodity indices, but this observation extends to all the majors.
Below I’ve compared the GS index (red) to the CRB, Dow Jones-UBS (ETF DJP), and Deutsche Bank (ETF DBC) indices in blue. Note the similar results.
Of the major asset classes, only corporate bonds, Treasuries and gold (and currencies if you view that as an asset class) provide any level of real diversification to equities in this market.
Asset classes which have in the past not exhibited strong positive correlation, including: commodities, international stock indices (developed and emerging markets), and real estate, all now see strong positive correlation to U.S. equities.
This hurts strategies that use diversification as a way to reduce portfolio volatility, whether it’s traditional MPT or something more active like Tactical Asset Allocation.
None of this is new information and I’m not the first to highlight it – just something that we as investors need to keep on the radar. Reducing portfolio volatility through diversification continues to become more and more difficult.
Happy Trading,
ms
. . . . .
To stay up to date with what’s happening at the MarketSci Blog, we recommend subscribing to our RSS Feed or Email Feed.
Filed under: Random Stuff | 25 Comments
one thing interesting is the correlation seems to be regime dependent, always turns from significantly negative to significantly positive every a few years. Any thought? thanks, Michael.
Hello abiao – no immediate thoughts – this rise in correlation came during both bear and bull markets – but it’s something to look at. More to follow. michael
Always surprised me people don’t talk of corporate bonds more (Unfortunately, don’t know much about them!).
Interestingly I can’t imagine a reversal of this correlation any time soon:
I don’t think it’s likely we’ll see a strong bull market in stocks coupled with falling commodity prices (because if they fall, it’s due to decreasing industrial demand, probably partly due to a hard landing in china), nor can I imagine that commodities increase in case of a deleveraging recession that causes falling stock prices (other than maybe gold).
Only after a major recession it’s likely that commodities stay low but stocks start rising again, but that might not happen before 2012 (some think not even before 2014).
Hello Alec – good comment – further supporting your point is the fact that all of the commodity indices are heavily weighted towards energy products which I think are even more likely to follow the scenarios you laid out above.
Having said that, I didn’t show it in this post for brevity’s sake, but more recent data (ex. rolling 1-year instead of 3-year correlation) show correlation coming down a bit from its highs, possibly (again, possibly) showing this relationship is losing some steam.
michael
Hi there,
On bonds and other asset classes, I’m surprised to hear that they’re uncorrelated in recent years. I would’ve thought that bonds (corporate and sovereign) would have become much more correlated (even if negative in direction) with equities over the last 3-4 years. Do you have a chart showing that number?
Hello Chon – I should clarify my post – when I say “high correlction” in this post I’m referring to high positive correlation. Some of the asset classes I mention in the post are strongly negative correlated to equities. Sorry for the confusion – bad choice of words on my part. michael
“Of the major asset classes, only corporate bonds, Treasuries and gold (and currencies if you view that as an asset class) provide any level of real diversification to equities in this market.”
Umm, I don’t know about Treasuries. Last time I looked (9/30) they were running about -.62 correlation over the preceding 6 months. That’s pretty highly correlated. Of course, I realize we’re talking different time periods here (and slightly different products, I was looking at ES and US futures), but I think T-bonds are relatively highly (negatively) correlated with stocks these days.
Hello Carlos – see my last comment above – when I said “high correlation” in this post I’m referring to high positive correlation. That’s a bad choice of words on my part – you are correct, that some of the asset classes I mention may have a strong negative correlation. michael
P.S. changed the wording of the post to make that distinction between positive/negative correlation more clear. michael
I think the change helps, but I believe one of your main points is when you state “Of the major asset classes, only corporate bonds, Treasuries and gold (and currencies if you view that as an asset class) provide any level of real diversification to equities in this market.”
I would argue that because of their high (albeit negative) correlation, Treasuries shouldn’t be on that list.
But maybe I’m beating a dead horse here…. if so, forget I said anything. :)
Hello Carlos – I disagree. Taken to the extreme (correlation = -1.0), yes, a negatively correlated asset isn’t helpful, but in the real world where that isn’t the case AND the future is inherently unknowable, negatively (even strongly negatively) correlated assets combined together in a well-diversified portfolio are very effective at increasing returns relative to volatility. That’s the math of MPT. michael
Michael,
You’re incorrect. Negatively correlated assets reduce volatility, but also reduce returns. They do not improve risk adjusted returns as per MPT. A couple common sense way of thinking about this:
- You can short any asset (in theory). Going long a -0.8 correlated asset is equivalent to shorting a 0.8 correlated asset.
- A short SPY ETF could be -1.0 correlated to SPY, but owning both does not improve your risk adjusted returns.
You want zero correlation, not strongly negative.
Hello Chon – I disagree wholeheartedly.
Shorting a +0.8 asset is NOT the same as going long a -0.8 asset.
Correlation captures the relationship between two assets’ beta. It does not capture alpha.
That’s why you can have two assets with perfectly -1.0 correlation BOTH gain (or lose) together over a given time period. The correlation figure is not capturing the alpha that either asset exhibits.
When you short that +0.8 asset, you’ve also flipped the expected return (and the alpha).
If that didn’t compute, let me know and I’ll provide some simple data to illustrate.
michael
RE to Chon: to illustrate:
http://www.marketsci.com/supporting.docs/20111018.xlsx
Two fictitious assets. Asset A beta = 1.0, alpha = 5%. Asset B beta = -1.0, alpha = 10%.
CORRELATION = -1.0.
Yet, they both gain over the period. Had you shorted asset B, you would reached a totally different result because you would have been shorting that alpha as well.
michael
Michael,
Thanks for the example. I’m not sure I understand the application of the alpha value. Shouldn’t the alpha be multiplied by the previous asset value, instead of just added on as a percentage? I’m thinking that since alpha is the amount the asset outgains the market by, you have to apply that to its previous value to get the current one.
On a higher level, I think I see your point, but my gut is that the contributions to diversification from strongly negatively correlated assets will be minor in comparison to those from almost uncorrelated assets. I don’t have any calculations to prove that, but I’d be interested in your thoughts.
=Carlos=
Hello Carlos – alpha is unaffected by beta, so it’s just a flat value added on to the equation (note: we both know that in the real world assets don’t behave in such a logical way – this is a conceptual discussion)
IMHO: a big (biggest?) problem we face in producing diversified portfolios today is diversifying away equity-related risk because so many non-equity asset classes are now strongly positively correlated with equities. I think any assets we can add to the pot that are low/negatively correlated (but with positive expected return) is inherently good because there are too few to choose from in today’s market.
Is a non-correlated asset better or worse than a strongly negatively correlated one? You can’t answer that based on the description alone, there’s too many other considerations: expected return unrelated to beta, how the assets act during periods of market stress or crises (which correlation doesn’t necessarily capture), stability of the correlation measurements, how successfully you can time each of the assets, etc.
But the ultimate (albeit unachievable) goal would be consistently perfectly negatively-correlated assets generating copious alpha, minimizing beta and leaving you with pure alpha.
Just my $0.02.
michael
I have to agree with Michael. Generally speaking negative correlations are preferable over low correlations (assuming the assets are meaningful to own otherwise – alpha and other good factors like current income, tax efficiency etc as appropriate to the investor’s needs).
If you want to try this, you will find that long bonds make a better improvement to an equity portfolio then short bonds even though they come with more risk because of their generally negative correlation. Short bonds tend to be more or less uncorrelated. Do you recall all the press when there was shortage of long bonds issued by the Feds, that was a real unmet need for portfolio construction.
What drives the dramatic correlation shift around 2008? Looks odd.
Hello David – good question – the initial jump was a result of 2 big down months in the market that were matched by big down months in the commodity indices (09/2008 and 10/2008). Doesn’t take much to move the stat with only a 36 month lookback. michael
Makes sense. Thanks.
You said “Of the major asset classes, only corporate bonds, Treasuries and gold (and currencies if you view that as an asset class) provide any level of real diversification to equities in this market” and I was wondering what you considered minor asset classes that are available to retail investors that provide diversification. Additionally, do the normal long-term strategies you’ve covered on your blog in the past still apply to the major and minor asset classes used to diversify away from equity exposure (e.g., trend following and momentum)? I ask because I’d prefer to have a better option than sitting in cash during bear markets and I don’t particularly care to short any equity-like markets. I remember you saying that you haven’t had the best luck using trend following techniques with IEF, but as an average investor I’m simply looking for “respectable luck” versus sitting in cash.
Thanks for picking up the pace of posting again (and in another post you asked what your readers would like to see you post about: my answer is anything that can help me better understand the markets and the different strategies you can use to boost volatility-adjusted returns).