How much of a commodity index's return is a design decision?
Far more than most people assume. Two products can both claim to track the same commodity, both do exactly what they say, and diverge by tens of percentage points over a decade — entirely because of choices made in a rulebook.
This is a description of design axes, not an evaluation of any product.
Axis 1 — which contract to hold
Take a curve: M1 $80.00 · M2 $81.50 · M3 $82.60 · M6 $84.50 · M12 $86.00. It slopes up and, as curves usually do, it flattens with tenor.
- A front-month design rolls M1 into M2 and pays the widest spread on the curve: $1.50 on $81.50 = 1.84% a month ≈ −20% a year if the shape persists.
- A twelve-month deferred design always holds the contract a year out and rolls M11 into M12. If those settle at $85.80 and $86.00, the spread is $0.20 on $86.00 = 0.23% a month ≈ −2.8% a year.
A seventeen-point annual difference from one line in a rulebook. But there is no free lunch in it, and the reason is the same flattening that produced the advantage:
The deferred contract also responds far less when spot moves. If spot rallies 20% and the far end of the curve rises only 8%, the deferred design captures a fraction of the move. You have traded roll drag for spot sensitivity, one for one. Neither end of that trade is "better" — they are different instruments answering different questions.
Axis 2 — when to roll
- A published fixed window. Roll a fifth of the position on each of five specified business days each month. Transparent, replicable, auditable — and completely predictable to everyone else in the market.
- Continuous rolling. Move a small slice every trading day, so the position is always spread across two or more months. Less predictable, more operationally complex.
- Discretion within a band. Roll somewhere inside a stated window, at the manager's choosing. Less replicable; harder to benchmark.
Axis 3 — rule or optimisation
Some designs choose, from among all listed months, the contract whose implied roll is least negative — the "cheapest" point on the curve to stand. These are described as dynamic or optimised roll methodologies.
What they buy is a better expected roll return. What they cost is that the position's maturity — and therefore its sensitivity to spot — now varies over time, so the product's behaviour is no longer a fixed thing you can reason about from the front of the curve.
Roll congestion, and why the windows keep moving
Here is the second-order effect. A very large roll, on a published schedule, in a fixed direction, is a known quantity of one-sided flow arriving at a known time. Other participants can position ahead of it. The calendar spread tends to widen against the roller during the window, which raises the cost of the roll for the very fund whose rules announced it.
This is a well-studied effect, and it is why methodologies have drifted over the years toward longer, staggered and less mechanically predictable windows. Transparency and execution quality are genuinely in tension here, and every design picks a point on that trade-off.
The conclusion worth carrying
"The commodity returned X" is an incomplete sentence. Complete it: which contracts, rolled when, against what collateral. Until those are specified, two honest people quoting two honest numbers for "oil last year" can differ by twenty points and both be right.
Nothing in this lesson recommends any methodology or any product. It describes how the same underlying commodity produces different measured returns depending on rules that are published in advance and that anyone can read.
Try it now
- Find the methodology sections of two commodity products on the same commodity. Write down, for each: which months, what roll window, rule or optimisation, and what the collateral earns. Four rows, eight cells — that table is the whole of the difference between them.
- One of those design decisions is visible below: the commodity itself against a front-month, fixed-window fund on it. Measure both over the same span.
- Now predict where a deferred-month or optimised fund would sit between those two lines, and say which of the three design axes puts it there. If your explanation needs anything beyond months, window and collateral, you have missed one of the rules.