Price signals

How does a failed harvest on one continent change what a shopkeeper charges on another?

A drought ruins the coffee harvest in one country, and some months later a shopkeeper eight thousand kilometres away raises the price of a bag. She has never seen the farms, does not know which ones failed, and could not describe the weather that did it. Yet her price is right — roughly right, in the direction the world's actual scarcity moved. Who told her? And why did nobody have to?

Start with what would be needed if we tried to do this deliberately. To set the correct price of coffee you would want to know how much was lost, how much is in store, what buyers are willing to give up, what substitutes exist, what it would cost to plant more and where, and what everyone expects next year to look like. Notice that this knowledge is not merely large — it is scattered. It exists in pieces, in the heads of thousands of growers, traders, roasters and drinkers, and much of it is local, tacit, and changing. No office could collect it, because a great deal of it is not written down anywhere and some of it is not even conscious until someone has to decide.

So how does anything coherent come out of that? Consider what a single participant does. A grower with a small crop will not sell at last year's price. A buyer who needs beans will offer a little more rather than go without. Each is acting on the tiny fragment of the situation they can see, and each is asking only their own question: what is this worth to me? Yet each of those private answers becomes a public act — an offer, a bid, a sale at a number.

And the number is the point. What information does a rising price actually carry? Not a cause. The shopkeeper learns nothing about drought, and does not need to. What she learns is that coffee has become more valuable relative to everything else — which is precisely the fact she needs in order to act well. Friedrich Hayek's 1945 essay "The Use of Knowledge in Society" is the classic statement of this: the price system works as a mechanism for communicating information, and its economy is that only the most essential fact is transmitted, and only to those who need it.

Now watch the number do two jobs at once, which is the part most worth slowing down for. The higher price reaches buyers, and some of them buy less, switch to tea, or reduce waste — so the shrunken supply is rationed toward whoever values it most, without any rule about who deserves it. The same higher price reaches producers, and some of them harvest marginal trees, release stock they were holding, or plant more for next year — so the shortage is also being repaired. One number, doing rationing and incentive simultaneously, in opposite directions, to different people. And it does not need to be believed or understood, only responded to.

Which suggests a check: does it work if nobody knows why the price moved? It does. Hayek's own example was tin becoming scarce — users economise and seek substitutes whether the cause was a mine collapse or a new use elsewhere, and the adjustment ripples outward without the cause ever travelling with it.

Two refinements first, then the honest limits.

The refinement is that a price is not a measure of importance or of need. It is a measure of scarcity relative to willingness and ability to pay. Water is essential and usually cheap; a rare stamp is useless and dear. The signal answers "how hard is this to obtain, relative to what people can offer for it" — and it weights a rich buyer's mild preference above a poor buyer's urgent need, because money is what it counts. That is not a malfunction of the mechanism; it is what the mechanism measures, and it is why "the market decided" is never a complete moral argument.

The second refinement is that the signal is a compression, and compression loses things. Because only the number travels, information about why is discarded — which is efficient, and also means the system cannot distinguish a temporary shock from a permanent shift. Producers who read a spike as permanent overplant, and the correction arrives as a crash. Slow-to-adjust supply, as in agriculture and mining, makes such cycles routine.

Now the limits, which are real and specific. Where a cost falls on someone outside the transaction — pollution, depletion, congestion — the price omits it, so the signal is accurate about private cost and wrong about total cost, and directs effort confidently in the wrong direction. Where one party has market power, the price reports that party's discretion as well as the world's scarcity. And where one side knows something the other cannot verify, the price can stop sorting quality at all: George Akerlof's argument about used cars showed that buyers who cannot tell good from bad will only pay an average price, which drives good sellers out and degrades what remains. In each case the mechanism is not being violated — it is faithfully aggregating the information it is given, which is incomplete.

• How carbon pricing tries to force an unpriced cost back into the signal, and where it fails.
• Why the socialist calculation debate turned on this question, and what modern computation does and does not change about it.

Every term the collection defines is gathered in the glossary.