Spread

Because a spread is defined by its operands, the useful taxonomy is by what the two prices differ in. Each kind isolates a different variable, and each is read against a different logic. Treating them as one concept is the source of most confusion about the term.

Calendar spread

Two delivery periods of the same contract. Isolates time: the relationship reflects the cost and value of carrying the commodity between the two periods, and is the structure that contango and backwardation describe.

Inter-commodity spread

Two different but economically linked commodities — substitutes in a feed ration, or a raw material and its product. Isolates the linkage between them, and is only meaningful where a real substitution or transformation relationship exists.

Inter-market spread

The same or a comparable commodity quoted in two markets or origins. Isolates location and origin, and is closely related to what an origin differential expresses.

Quality spread

Two qualities or grades of the same commodity. Isolates the market’s valuation of the quality difference, and is closely related to a quality premium.

Bid–offer spread

The gap between the best price a buyer will pay and the best a seller will accept, at one moment, for one thing. Isolates nothing about the commodity at all — it is a property of the market’s trading conditions, and belongs to a different discussion entirely.

The bid–offer spread deserves separating out, because it is the odd one. The other four compare prices for different objects. The bid–offer spread compares two prices for the same object, differing only in which side of the transaction they are on, and it is therefore a measure of the market’s trading condition rather than of any economic relationship between goods.

The reason to look at a spread rather than at its two components is that a spread removes a shared influence. If a broad shift moves an entire market, it moves both legs of a calendar spread together and largely cancels out — so the spread is a cleaner view of the time relationship than either leg alone. This is the analytical function of the concept, and it holds regardless of which kind of spread is being considered.

The cancellation is only as clean as the relatedness of the two legs. Two months of the same contract share almost everything, so the cancellation is nearly complete and the residual is genuinely about time. Two different commodities share much less, so an inter-commodity spread cancels only what they have in common and leaves a residual containing everything else that distinguishes them — which may be a great deal. The looser the relationship, the less a spread isolates and the more it simply subtracts two numbers that happened to be to hand.

A quoted spread is uninterpretable without its construction. The number carries no record of what was subtracted from what, and the same market relationship can be quoted as a positive or a negative figure depending purely on convention.

• Which two legs, exactly — which contracts, months, markets, qualities, or origins
• The direction of the subtraction, and the convention the quoting source uses for the sign
• Whether both legs are quoted in the same unit and currency; a spread between prices in different units is not a spread but an error
• Whether both legs are the same kind of price — subtracting an assessed value from a transacted price mixes two kinds of fact
• The moment both legs were observed; legs read at different times contain a time difference that the spread silently absorbs

The last of these is a persistent and quiet source of error. A spread is only about the difference between its legs if the legs are contemporaneous. Two prices observed hours apart in a moving market produce a spread that contains the market’s movement in that interval, attributed to whatever the spread was supposed to be isolating.