How And Why I Am Breaking My Rules For Commodity Trading

LYagovy/iStock via Getty Images

When an asset class is moving 5-10% a day in a single direction and the terminal point is demand destruction – and “demand destruction” is likely a euphemism for “general market implosion” – what rating ought one give that class? Buy, sell, or hold?

I hope I am not telling any tales out of school by informing readers that Seeking Alpha has a ‘fast track’ process for writers to submit articles for which time is of the essence. But, there is no procedure for writing and submitting an article that could deal with these kinds of price moves. Nickel futures, for example, are 60% higher than they were when I started writing this morning. Who knows where they will be by the time I finish this sentence. 120% higher? 60% lower? Ah, I just checked again. Up 75% on the day. (This paragraph was written on Monday).

There is no choice but to turn to the principles I followed before this onset of Weimar conditions in commodity markets. So, what follows is a discussion about some of these first principles and what they say about what commodities will do. And, within those parameters, we can then consider how to incorporate current events.

There are five principles I follow in anticipating movements in commodity markets. The first two relate to long-term (decadal) moves and the last three to cyclical (one to three years) moves. These principles do not translate into simple buy and sell signals, but I think they point to where the preponderance of upside and downside risks may lie. In my opinion, the long-term (to the end of the decade) and short-term (over the next one to three years) risks are primarily to the downside.

So, let’s begin with the long-term principles.

Principle #1: Commodity prices track the earnings yield over the long term

Nine years ago, I wrote a five-part series of articles on this question in which I laid out the evidence going back to 1730 for the following conclusions about historical commodity prices:

1. Nominal goods prices, especially producer prices, were almost universally correlated with bond yields from as far back as 1730, but in the twentieth century, this correlation collapsed.

2. Nominal commodity prices were correlated with the earnings yield from 1871-1959.

3. Deflated [i.e., “real”] commodity prices have been highly correlated with equity yields since 1871.

4. Therefore, it seems more likely that the Gibson effect is primarily a relationship between commodity prices and equity yields.

I do not want to re-litigate these points here, but I want to give readers an impression of the depth and resilience of this relationship.

The “Gibson effect” in Point 4 is my term for a market relationship that lay at the root of what Keynes called “Gibson’s Paradox“. Economic theory at the time suggested that interest rates and goods prices ought to be negatively correlated, yet, as a banker named Gibson discovered in the 1920s, wholesale prices had always been h2ly positively correlated with interest rates.

Lots of economists have worked on why this might have been, and it is research on this by Lawrence Summers and Robert Barsky that led to the (erroneous but oft-repeated) principle that real gold prices are determined by real interest rates.

My conclusion is that what Gibson observed as a relationship between interest rates and producer prices was actually an artefact of a relationship between the earnings yield and primary commodity prices (although this relationship is itself probably an artefact of something more fundamental).

The following chart from that series illustrates the third point above. It shows annual deflated prices for the Grilli-Yang Commodity Price Index (the GYCPI, in dark blue) and for a producer price index (pink) spliced from Warren & Pearson wholesale index ( Series L.2) and the Fed’s industrial commodities index ( PPIIDC) and the dividend yield (yellow) and earnings yield (light blue) for the S&P Composite index going back to 1871.

Chart A. Commodity prices have always tracked earnings yields. (Stephan Pfaffenzeller, St Louis Fed, Warren & Pearson, Robert Shiller)

Incidentally, I did not find that any individual commodity or class of commodities within the GYCPI had an obviously h2er or more consistent relationship with equity yields than did others, but I did find that the relationship between commodities and the earnings yield did seem to be h2er than that between commodities and the dividend yield.

And, as another aside, when comparing British prices and yields, it appeared that negative correlations between goods and yields tended to involve goods that were not easily traded internationally (for example, potatoes).

The following chart uses monthly World Bank commodity data, deflated by CPI, to update this relationship.

Chart B. Short-run real commodity prices can deviate from the earnings yield but eventually realign. (World Bank, Shiller, St Louis Fed)

Over even multi-year spans, this correlation can be weak, but over the long term, through whatever sorcery exists in markets, commodity prices and the earnings yield track one another.

Based on this relationship, it would appear that a gap has opened up in the last few years between commodity prices and the earnings yield. There are multiple ways this gap can be closed, but a brief review of how those relationships work suggests that some are more likely than others.

One is that consumer inflation could rise faster than commodity inflation. Since commodity prices tend to have a high beta relative to consumer inflation, this seems somewhat unlikely. A further surge in the underlying level of inflation would likely boost real commodity prices even further. Real commodity prices are likely to fall primarily through a fall in nominal commodity prices.

Another possibility is that the earnings yield could rise. There are very few instances in history of a meaningful rise in the earnings yield occurring via a decline in stock prices. So, a rise in the earnings yield would almost certainly have to be generated from a rise in earnings.

So, the two most likely ways for this gap to close is through a fall in nominal commodity prices and/or a rise in earnings. Of course, the gap could widen or be sustained for a considerable amount of time, but on the question of commodities, this relationship leans more in the direction of deflation.

We will consider other implications later in the article, but one that has to be considered is what would it mean were this relationship to break down. No sooner had Gibson observed this relationship and Keynes wrote about it (that is, the specific relationship between nominal prices and interest rates) than it seemed to vanish. But for now, under the assumption that this relationship is stable, it points more to commodity deflation than inflation.

Principles #2: Bear markets in commodity prices last at least 20 years

This chart is from an article I wrote in 2020.

Chart C. Since the establishment of the Fed, commodity ‘supercycles’ occur every 30 years (Warren & Pearson, St Louis Fed, World Bank)

Since the establishment of the Federal Reserve in 1914, commodity inflation peaks every 30 years, and then commodity deflation (especially in real terms) lasts for 20 years. Before the Fed was established, this deflation could last longer (something like 30 years).

This pattern is less consistent, therefore, than the one described by Principle #1 above. But, Principle #1 suggests that it is not arbitrary, because since the establishment of the Fed, equity yields also peak every 30 years and tend to decline for 20 years.

The following chart shows the earnings yield and commodity price cycles (current price divided by a 36-month moving average). The earnings yield tended to peak in 1920, 1950, 1980, and 2010, although this was last one was significantly lower (as were real commodity prices per Chart B).

Chart D. Peaks in commodity cycles tend to coincide with peaks in the earnings yield. (Warren & Pearson, St Louis Fed, World Bank)

In the chart above, I call these long waves in yields and commodity prices “Neo-Kondratieff Waves”. N D Kondratieff, a Soviet economist, wrote in the 1920s that a host of economic phenomena, including goods prices, tended to move in 50-year “long waves” (from crest to crest). It appears to me that with the establishment of the Fed, these long waves effectively doubled in frequency. And, as I have outlined elsewhere, and as the Austrian economist Joseph Schumpeter hypothesized, these appear to be linked to technological supercycles. I have also written about the long-term swings between the relative performances of the energy and technology sectors that seem to align with this, and how technological booms in the real economy coincide with general disinflation but not necessarily booms in technology stocks.

History, in other words, shows that the length of these waves in inflation and yields is not consistent (again 50 years in the centuries before the Fed and 30 years over the last century), but it also suggests that there is some deeper fundamental driver tied to technology. In a series of articles that I wrote late last year and into January, I argued that the earnings yield has been depressed too far and for too long and that this was likely to revert, suggesting significantly higher levels of commodity inflation, but partly because of the persistence of deflationary pressure in these waves, this was more likely to occur in the 2030s than 2020s.

So, without knowing the underlying mechanism behind these waves, we cannot be certain, but history points more in the direction of low inflation and low yields in the 2020s.

Let’s turn now to the cyclical side of things.

Principle #3: Commodity prices tend to peak with cyclical growth rates

Commodity prices are cyclical, and for the purposes of this discussion, that means when the cyclical rate of growth peaks, prices peak.

This relationship can observed to some degree in Chart D. Peaks in the cyclical growth rate in commodities tend to coincide with peaks in the earnings yield.

The following chart shows the relationship between the cyclical rate of change in the World Bank’s iAGRICULTURE price index and that same index deflated by CPI.

Chart E. Peaks in commodity cycles coincide with peaks in real prices. (World Bank, St Louis Fed)

As I said in the introduction of this article, this is not a clear trading signal, because the cycle is itself volatile. Just in the last three months, for example, the agricultural commodity cycle began to re-accelerate after softening in late 2021.

The following chart shows the relationship between the agricultural cycle and the nominal price level.

Chart F. Peaks in cycles coincide with peaks in nominal prices as well. (World Bank)

I would estimate that there have been nine or ten agriculture cycles since 1970 (was the 2000-2008 run one or two cycles?). The cycles that concluded in 1974 and 2008 were the two most powerful. The current one is in a three-way tie for third place. This data only goes up to February of this year, so if the current levels can be maintained throughout this month, this one will likely take third place with little trouble, and perhaps it will take out first and second, as well.

It has to be remembered that these are monthly prices, so it conceals a lot of the short-term volatility that can occur from day-to-day, as evidenced by today’s moves in nickel prices (which settled at 67.22% on the day).

Trading cyclical changes in commodity prices is a momentum strategy and almost by definition, that implies that at the tail end of any long position, you are at risk of significant losses. Sometimes the cycle will take you out too early and other times a bit too late, especially if you got in late.

As long as the momentum is h2, things look bullish, so I have to say that this measure is generally bullish, but not unequivocally so. The fact that the current cycle is ranked third suggests that there is risk of a reversal.

Principle #4: Commodity cycles are correlated with earnings cycles

Earnings cycles (calculated as log changes over a 36-month moving average of S&P 500 trailing twelve-month earnings) and commodity cycles move together.

Chart G. Commodity cycles are correlated with the earnings cycle (World Bank, Shiller, S&P Global)

This is especially true for industrial metals prices, represented here by the World Bank’s iBASEMET index. There is not a clear correlation between the degrees of these moves, however. In fact, as I mentioned in my last article, the relationship between the earnings yield and real commodity prices suggests that real stock prices operate as if they were a function of the ratio of earnings to commodity prices.

And that is indeed the case, as the following chart shows.

Chart H. The ratio of earnings to commodity prices is highly correlated with real stock prices. (Shiller, World Bank, St Louis Fed)

What is more, the report of earnings data is always lagging, so although earnings may give important clues, there is a good chance that any intelligence provided by them will arrive too late.

If you plug in S&P 500 estimates for where earnings will stand in 1Q22, as I did in Chart G, it suggests that earnings growth is likely to decelerate, but I have never backtested how reliable earnings estimates are at anticipating changes in momentum.

If earnings data starts to look shaky, that might be a good signal that the gig is up. If you look at something like Ed Yardeni’s plush chartbook on S&P 500 Weekly Fundamentals, the earnings forecasts are clearly pointing to a slowdown in operating earnings and margins.

Chart I. Declines in earnings estimates seem to anticipate commodity crashes. (

In 2018, there was a similar such slowdown in estimates and this appears (I have to eyeball this) to have coincided with the peak in the industrial metals cycle in 2018 illustrated in the following chart.

Chart J. Industrial metals cycles peaked as forecasted earnings slowed (World Bank)

This, therefore, suggests that the commodity cycle is very close to its peak, but as alluded to at the outset of the article, it is not clear how meaningful that is when commodities can pack on 20 lbs. of muscle a day. Again, the peak of the earnings cycle gives us a vague clue about the timing of the commodity peak but little information about the magnitude of the cycle.

In any case, even though all the talk is about inflation and there are plenty of reasons to worry about that (related to monetary policy, supply chain woes, and geopolitical chaos), commodity investors should not at all be surprised if commodities were to begin to fall as dramatically as they rose in the coming months.

Principle #5: Precious metals lead the commodity cycle

Commodities have their own little ecosystem. Precious metals lead the cycle, industrial metals, textiles (cotton, rubber, etc.), and earnings are at the core of the cycle, and energy, fertilizers, and agricultural commodities lag.

That is a difficult thing to prove, but I have tried to illustrate the relationship below. This shows the precious metals, earnings, and oil cycles, along with a 16-month moving average of the precious metals cycle.

Chart K. Precious metals lead and energy lags the earnings cycle. (World Bank, Shiller, S&P Global)

It is hardly a surefire indicator, but the moving average of the precious metals cycle tends to correlate with the earnings and oil cycles.

Chart L. Monthly data to February continue to point to a slowing commodity cycle (World Bank)

Each of the three precious metals have exhibited a serious decline in momentum. Yet, if you look at these daily charts from, they paint a different picture.

Chart M. Gold prices appear to be breaking that downward trend. (

Each of these metals started putting in higher lows in December of last year, and now they are ripping, especially gold. Gold is a special animal, and a brief aside may be in order. (I wrote about gold in greater detail here).

First, look at the relationship between real silver prices and the earnings yield (the relationship described in Principle #1).

Chart N. Real silver prices respect the earnings yield. (Shiller, World Bank)

Silver is pretty loyal to the earnings yield. But, with gold, there is clearly something else at work.

Chart O. There is some other force at work in gold prices. (World Bank, Shiller)

The gap between real gold prices and the earnings yield keeps expanding. Even as real gold prices fell during the 1980s and ’90s, the gap between those prices and the earnings yield widened.

My working assumption is that gold still retains some of its monetary qualities and, although it has behaved by and large as other commodities do, rising when the earnings yield rises and falling when the yield falls, it also has this anticipatory quality, if you will.

For now, though, it is non-precious commodities that are in the ascendant, which suggests that this move in gold is less a warning of new inflation than it is a lagging reaction to the present burst in cyclical commodity inflation and/or increased defensiveness across markets.

In other words, the fact that gold and its cyclical rate of growth only bottomed late last year suggests that commodity inflation is likely to decline over the coming year.

Let me give an example that might be especially salient now. What follows is the phosphate/gold ratio and subsequent log changes in phosphate prices.

Chart P. The phosphate/gold ratio has historically anticipated phosphate prices (World Bank)

Fertilizers, like energy, as mentioned above, tend to lag the cycle, and this creates an effect whereby these commodities’ gold ratios have a slight tendency to predict changes in their prices. The peak in gold relative to phosphate prices during the pandemic suggests that we are likely to see deceleration in fertilizer prices over the next 16 months. And, as Principle #4 states, cyclical changes tend to correlate with absolute levels in commodities.

I do not want to exaggerate the strength of these correlations. For oil, the correlation between its gold ratio and its subsequent 16-month performance is -0.27. For natural gas, -0.23. Phosphate, -0.34. The fertilizer index -0.19.

Crude oil is shown below.

Chart Q. The oil/gold ratio anticipates oil prices. (World Bank)

You cannot set your watch by this, but commodity/gold ratios (especially those involving oil) tend to point to the subsequent direction of commodity markets.

What is more, it appears that some of these ratios and especially the oil/gold ratio has some hints about interest rate cycles, as well. The chart compares the oil/gold ratio and cyclical changes (log changes over a 36-month moving average) in the 10-year Treasury yield.

Chart R. The oil/gold ratio anticipates interest rates. (World Bank, St Louis Fed)

The oil/gold ratio also seems to anticipate the earnings cycle.

Chart S. The oil/gold ratio anticipates the earnings cycle (Shiller, S&P Global, World Bank)

So, to recap this point: gold and precious metals lead, earnings and industrial metals match, and ‘nondurable’ commodities like energy, fertilizers, and agriculture lag the cycle. In my opinion, the earnings cycle is the cycle, but that is a semantic point in this context.

My reading of Principle #5 is that both the slow growth in precious metals prices and their weakness relative to the more cyclical commodities suggests that a slowdown in growth and commodities is on the horizon.

What about now?

But when price moves that dwarf these cycles occur within days, what good is this? I have stopped checking nickel because after yesterday’s (Monday’s) move mentioned at the beginning of this article, nickel virtually doubled again, and trading was halted.

My solution is as follows. I do not pay attention to much fundamental analysis in the commodity space, because I think most of it is very difficult to backtest. But, there are a handful of analysts I listen to. One is Amrita Sen of Energy Aspects. If she says oil has the potential to go to $170 and $180, I pay attention. Especially within the context of the macro arguments I have been making for the last year.

In brief, those assertions (which can be found in links above) have been a) imminent extreme reversions in growth and stock prices (to the downside), b) low to negative inflation, c) high long-term cyclical volatility, d) underperformance in tech, and e) outperformance in cyclicals. Excessive growth, wildly excessive valuations, unsustainable pro-technology mania, unsustainable anti-energy mania, and, I am afraid, to say, repeated technocratic overreach has set us up for “The Death of Irrational Exuberance” and a return to history. Historically, the trigger that appears to unleash these reversions is a sharp cyclical spike in commodities, most notably energy.

A simple but certainly incomplete illustration of this relationship between cyclical changes in energy prices at the conclusion of long-term energy regimes and bear markets in equities is shown below.

Chart T. Nearly every major stock market crash (marked by arrows) occurred at the transition of long-term energy prices. (Energy Information Administration, Shiller)

The chart below is another one I published in “The Death of Irrational Exuberance”. It shows one of the few scenarios for the S&P 500 price and earnings performances in the 2020s that conform to historical market patterns.

Chart U. Forecast for S&P 500 price and earnings in 2020s (Own calculations)

What might it take to achieve such negative, Depression-like outcomes? My guess is a dual implosion of commodity markets and the global economy combined with rising risks of hot regional – and potentially world – wars would just about do the trick.

So, within that context, when someone like Amrita Sen starts saying that oil prices will rise until we reach demand destruction and we do not know precisely where that lies, I have to take that into account.

In other words, very few of the principles I have laid out above suggests where the precise limits are, and the one that comes closest is Principle #1 that suggests commodity price levels track the earnings yield. But, as I pointed out in my discussion of that principle, that is a very long-term tendency. Short-term deviations are not at all irregular.

If the global economy is being uprooted (and I suspect it is) and the clearest manifestation of that is in surging commodity prices while everything else is beginning to tank, I think it is prudent to take small positions in commodities. If, by the time this article is published, oil is at $300/bl (a possibility I heard on one of the financial cable news channels), maybe not.

Commodity allocations

So, which commodities should we gain exposure to? In this context, this too is a real problem. The ones that have the most short-term potential-those most exposed to Russian- and Ukrainian-related disruptions-also happen to be the ones most likely to crash. Energy, fertilizers, and agricultural commodities are late-cycle and their gold ratios all point to downside. When you observe the way that this crisis is rippling through commodities that tend not to be late-cycle but are exposed to the present geopolitical crisis, like nickel, it is also prudent to gain exposure to them, as well. Finally, because this crisis is, I believe, ultimately a symptom of a general world-system crisis, even though precious metals (most notably gold) are likely to underperform, they likely provide some protection against certain extreme outcomes, plus they are likely to see some absolute gains during the general run-up in prices, and they are likely to have less downside when this commodity crisis finally peters out.

Therefore, I would want roughly equal exposure across commodity classes. And that means I would shun many of the broad commodity indexes, like the Invesco DB Commodity Index Tracking ETF (NYSEARCA:DBC). These types of ETFs are value-weighted, and because oil is the FANG of the commodity world, energy makes up nearly 50%-60% of their holdings. Either mix and match sector-focused commodity ETFs (like DBA, DBB, DBE, and DBP), being careful to make gold (GLD) a significant portion of one’s commodity exposure, or hold one that is relatively equally balanced.

Chart V. DBC is heavily weighted to energy (

The iShares Bloomberg Roll Select Commodity Strategy ETF (CMDY), for example, is more evenly balanced.

Chart W. CMDY has a more generous portion precious metals. (

CMDY tends to underperform DBC in these kinds of cyclical upswings but holds up slightly better when the commodity cycle falls, as evidenced in the following chart.

Chart X. DBC is effectively an energy index (

One chart I will be keeping my eye on is the gold/bond ratio.

Chart Y. The gold/bond ratio is one place I will look for signals about the commodity market. (

If the market starts smiling on Treasuries instead of gold, I think that may be a good time to start unloading commodities generally.

Ultimately, however, this kind of insanity is ultimately unsustainable, commodity inflation will burn through all of its oxygen (and everybody else’s), and markets, including commodities, will crash. This is not the beginning of a commodity supercycle, in my opinion. It is the death throes of a commodity cycle positioned at the end of an extreme equity and technological supercycle built atop an unsustainable global political-economic order.

Thus, my conclusion is that positions in commodities should be small, be designed for protection against inflation and crisis extremes, and contain healthy portions of gold; and investors should be prepared to take significant losses on those commodity positions.

Leave a Reply

Your email address will not be published. Required fields are marked *