It genuinely protects a portfolio in some shocks and genuinely makes them worse in others, and the average of those two behaviours describes neither. That is the honest case, and it is a case for a small deliberate position.
10 min read · Level 3
The uncomfortable thing about this asset class is that the honest case for it is a case for insurance, and nobody likes buying insurance they can see the price of. An insurance policy has a premium you pay, a peril it covers, and a set of perils it does not - and this holding has all three: the carry is the premium, a physical supply shock is the covered peril, and a demand collapse is the excluded one. Presented that way the product is coherent and the decision is easy to reason about. Presented the way it usually is - as a hedge, a store of value, a diversifier with a favourable long-run correlation - it becomes something that is supposed to help in general, and a thing that is supposed to help in general will disappoint about half the time and be sold at exactly the wrong moment. What I would push against hardest is the long-run correlation figure, which I think does genuine damage. It is not wrong, it is a correct calculation, and it describes a behaviour the asset has never once exhibited: the average of helping strongly and hurting mildly is a mild relationship that no portfolio has ever experienced. The number that would actually inform a decision is the pair, held separately and named by the kind of shock that produces each. Anybody who can say which peril they are insuring, what the premium costs them per year, and what the excluded peril would cost them at their chosen size has understood this class completely, and they will have done it without forming a view about the price of anything.
The job this class does that nothing else in a portfolio does comes down to two things, and both need stating carefully.
Its price is set by physical supply and demand rather than by earnings, a contract or a rate. A share responds to what a company earns. A bond responds to what rates do. A building responds to local rents. A commodity responds to how much of a physical good exists and how much is being used, and those are genuinely different forces. In a shock that runs through the physical economy - a supply interruption, a harvest failure, a sudden industrial demand - this class can move in the opposite direction from everything else a portfolio holds, and no other holding will.
It is not anybody's promise. Lesson one. A bar owes nothing, depends on nobody, and is unaffected by the solvency of any institution. That property costs money every year and it is the thing being bought.
Now the half that gets overclaimed, which is the inflation argument.
Commodities are described as an inflation hedge, and over long periods there is a real mechanism behind it: the prices of physical goods are part of what a price index measures, and the cost of producing them rises with everything else. But the relationship is a tendency, not a rule, and the carry runs against it the whole time. Work an illustrative case - a nominal price rise of 30.00% over ten years against 3.0% inflation and 0.65% of carry:
Nominal effect 1.30 Carry 0.9935^10 = 0.936869 Inflation divided by 1.03^10 ------------------------------------------------------------------ Real outcome 1.30 x 0.936869 / 1.03^10 - 1 = -9.3746%
A 30.00% nominal rise over a decade of 3.0% inflation left the holder able to buy 9.3746% less than when they started. And running it the other way, the nominal rise required to simply hold real value over that decade is:
1.03^10 / 0.9935^10 - 1 = 43.4477%
That is the hurdle. It is not an argument that the hedge does not work; it is the size of the move the hedge has to produce before it has protected anything.
Because the diversification argument for this class is genuinely true and genuinely unreliable, and the usual way of presenting it - a single long-run correlation figure - averages the two into a number that describes neither.
Two illustrative stress episodes, six periods each, constructed to show a mechanism rather than measured from any market.
Episode A a physical supply shock A broad market -8 -12 +3 -9 -6 +4 cumulative -25.82% Metal M +9 +6 -4 +11 +5 +1 cumulative +30.57% Correlation -0.831
Episode B a demand collapse A broad market -7 -11 +5 -10 +6 -8 cumulative -23.72% Metal M -10 -16 +7 -8 -2 -13 cumulative -36.55% Correlation +0.874
Across both episodes together the correlation is +0.162, which is close to no relationship at all - and that figure is the least informative number on the page. It is the average of a strongly negative relationship and a strongly positive one, and the portfolio experienced each of them separately rather than experiencing their average.
Watch what the two do to a portfolio. Take a 95/5 split against holding the market alone:
Episode A market alone -25.82% with 5% in the metal -23.00% better by 2.82 points Episode B market alone -23.72% with 5% in the metal -24.36% worse by 0.64 points
Both of those are true, neither cancels the other, and which one a holder meets is not knowable in advance. That asymmetry is the honest case: the help in Episode A was more than four times the harm in Episode B, and the harm was real.
The reason the two episodes differ is structural rather than random, and it is worth naming because it is the only part of this that can be reasoned about. Commodities help when the shock comes through physical supply - something becomes scarce, its price rises, and the businesses that consume it suffer. They do not help when the shock comes through demand - a recession reduces how much of everything gets used, and a physical good with no income falls alongside the shares of the companies that would have bought it. A holder who understands which kind of shock they are living through will not be surprised by either.
Which leaves sizing, and here the carry does something unusual. In most classes an oversized position is a risk. Here it is also a running bill: 0.65% a year on a position is a permanent drag on the whole portfolio, computable in advance.
At 3% of a 200,000 portfolio 6,000 costing 39.00 a year At 5% 10,000 costing 65.00 a year At 10% 20,000 costing 130.00 a year
Over ten years those become a drag on the total portfolio of 0.1894%, 0.3157% and 0.6313% respectively, before any price move. Small, knowable, and paid whether or not either episode ever arrives.
Educational, not personalised financial advice. Nothing in the Investor Academy is a recommendation to buy, sell or hold anything — what is right for you depends on your own circumstances and the rules where you live.
Two portfolios, both holding 5% in Metal M, meeting the two episodes.
The portfolio that met Episode A. The market fell 25.82% cumulatively. The metal rose 30.57%. The 95/5 portfolio fell 23.00% rather than 25.82%, so the holding contributed 2.82 percentage points of protection at the moment protection was worth most.
The holder concluded that the metal had done its job, and it had. They will probably increase the position.
The portfolio that met Episode B. The market fell 23.72%. The metal fell 36.55%, which is further. The 95/5 portfolio fell 24.36% rather than 23.72%, so the holding made a bad period 0.64 points worse.
The holder concluded that the metal had failed, and it had. They will probably sell it.
Both conclusions are correct about what happened and both are wrong as general lessons, because each holder has met one episode and is generalising from it. The asset did not behave differently in the two cases - it did exactly the same thing both times, which is respond to physical supply and demand. What differed was the kind of shock.
That is why the case for this class cannot be made or broken by an experience. It has to be made structurally: this holding responds to a different set of forces from everything else you own, those forces sometimes move opposite to the others and sometimes with them, and the compensation for the times it helps is a running cost paid in all the times it does not.
Stated that way, two things follow immediately and neither requires a forecast. The position should be small enough that Episode B is an inconvenience rather than an event - at 5% of a 200,000 portfolio, Episode B cost 0.64 points, which is exactly that. And the position should be sized once, deliberately, rather than adjusted after each episode, because adjusting after an episode is the mechanism by which a holder ends up buying it after it has already worked and selling it after it has already failed.
Relying on a single long-run correlation figure. Across the two illustrative episodes here the combined correlation was +0.162, near zero, and it is the average of -0.831 and +0.874. A portfolio does not experience an average; it experiences one episode at a time.
Calling it an inflation hedge without the carry. A 30.00% nominal rise over a decade of 3.0% inflation, with 0.65% of carry, left the holder 9.3746% worse off in purchasing power. The nominal rise needed simply to hold real value over that decade was 43.4477%.
Generalising from one episode. Whichever episode a holder meets first will feel like the truth about the asset, and each is only the truth about one kind of shock: commodities help when the shock comes through supply and not when it comes through demand.
Adjusting the position after each episode. Increasing it after it has worked and selling it after it has failed is the most reliable way to hold the wrong amount at all times, and it feels like learning.
Sizing it as though it were free to hold. A 10% position on a 200,000 portfolio costs 130.00 a year and drags the whole portfolio by 0.6313% over a decade before any price move. Small, knowable, and paid whether or not the protection is ever needed.
Buying it for the not-a-promise property and then holding it through a futures product. Those two things are different: lesson one's distinction is what the physical route delivers, and a contract is a claim like any other.
It should name the forces: this responds to physical supply and demand, which is not what moves the rest of what I own. If the sentence turns out to be about an expected return, this class is a poor way to pursue one, and lesson two explains why.
A supply shock and a demand collapse are different events, and this class helps in one and not the other. Knowing which one you are buying protection from means you will not be surprised when the other arrives, and will not conclude the asset failed when it did exactly what it does.
At 5% of a 200,000 portfolio, Episode B made a bad period 0.64 points worse. Pick the size at which the version that does not help is something you can absorb without reconsidering, and then note that the helpful version was worth 2.82 points at the same size.
0.65% a year on the position is a drag on the whole portfolio of 0.1894% at 3%, 0.3157% at 5% and 0.6313% at 10% over ten years, before any price move. It is small, it is certain, and it is the standing price of the insurance - which makes it part of the sizing arithmetic rather than a detail.
When deciding whether to hold this class at all and in what size, after an episode in which it either helped or hurt and the temptation is to adjust, and whenever it is being recommended on the strength of a long-run correlation or inflation figure.
Do not use this lesson to choose a specific product - that is lesson four - and do not read the two episodes as measurements. They are constructed to show a mechanism, and it is the mechanism that transfers rather than the numbers.
Commodity allocations tend to be increased after periods in which they protected a portfolio and reduced after periods in which they did not, and both adjustments feel like responses to evidence. The difficulty is that each episode is evidence about one kind of shock rather than about the asset, and shocks of the two kinds do not alternate predictably. The result is a holder who is largest in the position after a supply shock has already passed and smallest after a demand collapse has already passed, which inverts the intended function: the protection is at its thinnest going into the conditions where it might have worked. Nothing about the reasoning at either moment is obviously faulty, which is what makes the pattern durable - it is composed entirely of individually sensible reactions to real outcomes.
Size this once, on the shock you are insuring against, and write down in advance that a single episode will not change it. The episodes are not evidence about the asset.
Create a free account to read the rest and apply it to your own investing.