Home > Commodities, Gold, Good One, Investing, Theory > Commodities Re-Thunk

Commodities Re-Thunk


I want to talk about commodities today.

To be sure, I have talked a lot about commodities over the last year. Below I reprise one of the charts I have run in the past (source: Bloomberg), which shows that commodities are incredibly cheap compared to the GDP-adjusted quantity of money. It was a great deal, near all-time lows this last summer…until it started creating new lows.

gdpadjcommod

Such an analysis makes sense. The relative prices of two items are at least somewhat related to their relative scarcities. We will trade a lot of sand for one diamond, because there’s a lot of sand and very few diamonds. But if diamonds suddenly rained down from the sky for some reason, the price of diamonds relative to sand would plummet. We would see this as a decline in the dollar price of diamonds relative to the dollar price of sand, which would presumably be stable, but the dollar in such a case plays only the role of a “unit of account” to compare these two assets. The price of diamonds falls, in dollars, because there are lots more diamonds and no change in the amount of dollars. But if the positions were reversed, and there were lots more dollars, then the price of dollars should fall relative to the price of diamonds. We call that inflation. And that’s the reasoning behind this chart: over a long period of time, nominal commodities prices should grow with as the number of dollars increases.

Obviously, this has sent a poor signal for a while, and I have been looking for some other reasonable way to compute the expected return on commodities.[1] Some time ago, I ran across an article by Erb and Harvey called The Golden Dilemma (I first mentioned it in this article). In it was a terrific chart (their Exhibit 5) which showed that the current real price of gold – simply, gold divided by the CPI price index – is a terrific predictor of the subsequent 10-year real return to gold. That chart is approximately reproduced, albeit updated, below. The data in my case spans 1975-present.

realgoldproj

The vertical line indicates the current price of gold (I’ve normalized the whole series so that the x-axis is in 2015 dollars). And the chart indicates that over the next ten years, you can expect something like a -6% annualized real return to a long-only position in gold. Now, that might happen as a result of heavy inflation that gold doesn’t keep up with, so that the nominal return to gold might still beat other asset classes. But it would seem to indicate that it isn’t a great time to buy gold for the long-term.

This chart was so magnificent and made so much sense – essentially, this is a way to think about the “P/E ratio” for a commodity” that I wondered if it generalized to other commodities. The answer is that it does quite well, although in the case of many commodities we don’t have enough history to fill out a clean curve. No commodities work as well as does gold; I attribute this to the role that gold has historically played in investors’ minds as an inflation hedge. But for example, look at Wheat (I am using data 1970-present).

realwheatproj

There is lots of data on agricultural commodities, because we’ve been trading them lots longer. By contrast, Comex Copper only goes back to 1988 or so:

realcopperproj

Copper arguably is still somewhat expensive, although over the next ten years we will probably see the lower-right portion of this chart fill in (since we have traded higher prices, but only within the last ten years so we can’t plot the subsequent return).

Now the one I know you’re waiting for: Crude oil. It’s much sloppier (this is 1983-present, by the way), but encouraging in that it suggests from these prices crude oil ought to at least keep up with inflation over the next decade. But do you know anyone who is playing oil for the next decade?

realcrudeproj

For the sake of space, here is a table of 27 tradable commodities and the best-fit projection for their next 10 years of real returns. Note that most of these fit a logarithmic curve pretty reasonably; Gold is rather the exception in that the historical record is more convex (better expectation from these levels than a pure fit would indicate; see above).

tableofproj

I thought it was worth looking at in aggregate, so the chart below shows the average projected returns (calculated using only the data available at each point) versus the actual subsequent real returns of the S&P GSCI Excess Return index which measures only the return of the front futures contract.

realindexproj

The fit is probably better in reality, because the actual returns are the actual returns of the commodities which were in the index at the time, which kept changing. At the beginning of our series, for example, I am projecting returns for 20 commodities but the 10-year return compares an index that has 20 commodities in 1998 to one that has 26 in 2008. Also, I simply equal-weighted the index while the S&P GSCI is production-weighted. And so on. But the salient point is that investing in spot commodities has been basically not pretty for a while, with negative expected real returns for the spot commodities (again, note that investing in commodity indices adds a collateral return plus an estimate 3-4% rebalancing return over time to these spot returns).

Commodities are, no surprise, cheaper than they have been in a long while. But what is somewhat surprising is that, compared to the first chart in this article, commodities don’t look nearly as cheap. What does that mean?

The first chart in this article compares commodities to the quantity of money; the subsequent charts compare commodities to the price level. In short, the quantity of money is much higher than has historically been consistent with this price level. This makes commodities divided by M2 look much better than commodities divided by the price level. But it merely circles back to what we already knew – that monetary velocity is very low. If money velocity were to return to historical norms, then both of these sets of charts would show a similar story with respect to valuation. The price level would be higher, making the real price of commodities even lower unless they adjusted upwards as well. (This is, in fact, what I expect will eventually happen).

So which method would I tend to favor, to consider relative value in commodities? Probably the one I have detailed here. There is one less step involved. If it turns out that velocity reverts higher, then it is likely that commodities real returns will be better than projected by this method; but this approach ignores that question.

Even so, a projected real return now of -2% to spot commodities, plus a collateral return equal to about 1.9% (the 10-year note rate) and a rebalancing return of 3-4% produces an expected real return of 2.9%-3.9% over the next decade. This is low, and lower than I have been using as my assumption for a while, but it is far higher than the expected real returns available in equities of around 1.2% annualized, and it has upside risk if money velocity does in fact mean-revert.

I will add one final point. This column is never meant to be a “timing” column. I am a value guy, which means I am always seen to be wrong at the time (and often reviled, which goes with the territory of being a contrarian). This says absolutely nothing about what the returns to commodities will be over the next month and very little about returns over the next year. But this analysis is useful for comparing other asset classes on similar long-term horizons, and for using useful projections of expected real returns in asset allocation exercises.

[1] In what follows, I will focus on the expected return to individual spot commodities. But remember that an important part of the expected return to commodity indices is in rebalancing and collateral return. Physical commodities should have a zero (or less) real return over time, but commodity indices still have a significantly positive return.

Advertisements
  1. eric
    January 13, 2015 at 7:38 pm

    This is an utterly fascinating post. Thanks for writing it. Its certainly very sobering for any gold investor. Assuming, just for giggles, 2% inflation over the next year, it predicts a 35% _nominal_ decline in the price of gold. That would put the (nominal!) price very close to what it was in the late 70s! On the other hand it would still be 42% higher than it was in 2002 (in real terms). So there’s that.

    Here’s what I find most puzzling about the last chart. According to the chart, in 2005, the expected real return of commodities was about -2%. And the actual subsequent return was -8%. That, in itself, is a bit puzzling. But what’s even more puzzling is that despite that massive underperformance, we still have negative expected returns. Now, I take your point about changes in the constitution of the index, etc. But still: if you look, e.g., at Hussman’s charts of the same ilk w.r.t. equity valuations, you do sometimes get mismatches between expected returns and subsequent returns, but as he is keen to point out: when you do, that is always matched by corresponding high or low expected returns 10 years later. how do you explain the failure of the chart to “catch up” in that way?

    a more pedestrian question: are you saying that you would expect, for example, USCI to outperform spot commodity prices by 490-590 basis point, annualized, for the next ten years?

    • January 13, 2015 at 9:54 pm

      I will answer the second question first. Yes, I would actually expect USCI to outperform spot commodity prices by MORE than that, because USCI also outperforms a “dumb” equal-weighted index fairly consistently by avoiding the commodities with the worst momentum/worst contango and buying those with the best momentum/best backwardation.

      As for the other question – notice that you really only have 1.5 full cycles here. You have a 1998 cycle (ending in 2008) that produced better-than-projected outcomes of 8% or so real compared with 2% or so projected; and at the end of that period the projection was -9% real and I feel fairly confident that by 2018 we will have done better than that on a real basis. We have lots more history of equity valuations than of commodity valuations.

      And, I really don’t want to short sell the idea of how coarse this projection is. I have these equal weight but the GSCI is 75-80% energy and 1% precious metals!

    • January 13, 2015 at 11:27 pm

      I ran the time series again, with 2008 GSCI weights (roughly)…78% energy, 6% industrial metals, 10% agriculture, 3% meats, 3% precious metals. And the chart looks MUCH better: https://mikeashton.files.wordpress.com/2015/01/gsciindexproj.gif

  2. Eric
    January 13, 2015 at 10:10 pm

    meanwhile, it looks like copper traders read your post! wowza.

    • January 13, 2015 at 10:13 pm

      I hope this isn’t a 10-year trade for them. 🙂

  3. Philippe
    January 13, 2015 at 10:46 pm

    Hi. I have started reading your blog a short while ago. And I simply wanted to say that I very much like reading your insights. Being now more of a long term investor than a trader, I particularly appreciate the “long view”. Please keep it going. And for someone who did not want to predict short term moves, well, that was pretty spot on! Cheers.

    • January 13, 2015 at 11:28 pm

      Well, THANK YOU Philippe! Always nice to hear from new readers. Please feel free to chime in any time!

  4. January 14, 2015 at 2:30 pm

    This is the best explanation I have come across, of why commodities are so low in price, that doesn’t draw upon “super-cycle” and such 🙂

    By the way, you masterfully bring the reader to this very question that you don’t answer yet: has velocity of money bottomed out? Next post, perhaps?

    • January 14, 2015 at 4:10 pm

      Ha – I’ve long thought velocity was poised to spring higher. All of our models say it should already have begun rising. I am just not sure what will force it to move (other than when the Fed starts to tighten. Then velocity will definitely rise). Frankly I think it’s already happening but it’s hard to write a comment about it – I just don’t have the analytics.

  5. Dan
    January 14, 2015 at 4:12 pm

    Great post!

    For the first chart, it seems like the velocity of money hasn’t been very cooperative.

    As compelling as the starting real price vs. subsequent performance charts are, I feel like something is missing — I suppose it’s the fact that commodities lack an “E” or cash flow. Presumably, marginal production cost must come into play, but those numbers have got to be all over the place. (Sorry, I’m just thinking out loud.)

    A couple questions:

    1. As the denominator in your first chart, could you use a broader definition of money that allows you to neutralize the vagaries of velocity? (Or perhaps just use GDP (world GDP?) in the denominator — a la Buffett for equities?)
    2. Are there any economic data points, such as changes in lending, that correlate strongly with velocity? (Unfortunately, in recent years, it seems that M2 correlates strongly — negatively, that is — with velocity!)

    Keep up the good stuff!

    • January 14, 2015 at 4:40 pm

      Commodities are basically zero-coupon instruments. You have a pile of copper today, and in 5 years it pays … a pile of copper. So the real return over time has GOT to be near zero because by definition your return in terms of stuff is zero. That said, the fact that marginal production costs decline over time is why actual real returns average slightly less than zero.

      As for your questions. 1) The broader the definition gets, the more velocity ends up mattering. If you take “base money,” a very popular choice, you also have a multiplier, for example. I think the “right” M to use over time probably changes and is related to how people actually spend money. When people paid cash for everything, M1 probably worked better, for example. But I think the bigger problem is that velocity has been SO unstable. Now, we know something about why it’s unstable: velocity tracks interest rates pretty darn well since it’s the inverse of the demand of real cash balances. So we can do a better job of prediction if we can predict interest rates better. Whoops. 🙂
      2. I guess I just answered #2. Yes – search this site for “verizon and velocity” and you’ll see a good chart. It’s not really news…Friedman said this about fifty years ago but it has been forgotten except among monetarists. When I stumbled on it I thought I’d discovered something new, and then I found Milton had anticipated me by a couple of generations!

  1. February 15, 2015 at 6:32 am

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: