What Does a Price Know That No Planner Can?

On New Year's Eve 2014–15 in New York, a technical fault switched off Uber's surge pricing for a stretch of the night. Nothing else changed: the same city, the same crowds, the same drivers with the same phones. Waiting times rose sharply and the share of ride requests that were actually fulfilled fell steeply. Removing one number from the screen did not make rides cheaper in any useful sense; it made them harder to get.

The episode cuts against a common intuition. Most people treat a price as a label that somebody chooses, a figure set by a seller's greed or a cost-plus sum, and therefore something that could be chosen differently with no further consequence. On that view a lower price is simply a transfer from seller to buyer. The New York night suggests something else: the price had been doing work, and when it stopped, the work went undone.

The stronger claim is that a price is not a label but a signal produced by scarcity, choices and competition. It forms where limited supply meets people who want the thing, and it compresses what thousands of participants know into a single figure. That is why capping, fixing or ignoring prices produces side-effects predictable enough to be named in advance. It is also, less comfortably, why prices sometimes fail, and why the failures follow a pattern of their own.

A number that carries knowledge

Friedrich Hayek made the case in 1945 with an example about tin. Suppose a new use for tin appears somewhere, or a mine closes. The price rises, and users everywhere economise, switching to substitutes or wasting less, without anyone needing to know why.

This is price as information. The knowledge that matters for using tin well is scattered across miners, manufacturers and households who never meet, and no single mind could gather it in time. The price does not transmit the knowledge itself; it transmits the one conclusion each user needs, which is that tin has become scarcer relative to the demands on it. Leonard Read's 1958 essay written in the voice of a pencil made the companion point that no one person knows how to make even a pencil, yet pencils are made.

The experimental evidence is striking. Vernon Smith (1962) handed students private cards, each showing either a value as a buyer or a cost as a seller, and let them trade by calling out bids and offers. None of them could see the supply or demand curves, and there were only a handful of traders, yet prices converged quickly on the predicted equilibrium. The market found a figure that none of its members knew.

The contrast case is administrative pricing. Soviet planners set prices for vast numbers of goods, and chronic shortages, queues and hoarding followed. The planners were not short of intelligence; they were short of the information that only moving prices generate.

Two jobs at once

A price does two things at once, as the New Year's Eve fault shows: a higher fare rations demand, so that rides go to the people who value them most at that moment. It also draws out supply: work by M. Keith Chen and colleagues (2019) on Uber data shows drivers adjusting their hours flexibly to expected earnings, so a surge pulls more cars onto the road. Switch the signal off and both mechanisms stall together.

The supply side works on longer horizons too. While oil traded at around a hundred dollars a barrel in the early 2010s, drilling and hydraulic fracturing in Texas and North Dakota expanded rapidly and US output rose steeply. When the price collapsed in 2014–16, rigs were idled in large numbers and output growth stalled. A drought or an emergency raises need, but need alone produces nothing; the price is what tells producers that extra effort will be paid for.

Surge pricing remains unpopular, and that reaction deserves to be taken seriously rather than dismissed. Efficiency and perceived fairness judge prices by different standards. A price can coordinate a market well while still striking people as unjust, and the two verdicts are answers to different questions.

What suppressing the signal produces

If a price carries information and does two jobs, then holding it below where the market would settle should produce predictable results. It does. A price ceiling does not make a good more plentiful; it makes it harder to find, and rationing shifts from price to queues, connections or lower quality.

The US petrol queues of the 1970s are the standard case: price controls combined with supply disruptions in 1973–74 and 1979 produced long lines and odd–even rationing by number plate. Rent control shows the same logic unfolding slowly. Rebecca Diamond, Tim McQuade and Franklin Qian (2019) studied San Francisco's 1994 extension of rent control to small multi-family buildings built before 1980. Covered tenants were more likely to stay in their homes, but landlords of covered buildings cut rental supply by about 15%, converting units or redeveloping, which pushed up rents across the city.

The pattern is precise enough to state as a rule. Interventions move a problem rather than erase it: sitting tenants gain, and those still searching pay, often without knowing why the market is so tight. Subsidies can misfire by the mirror route; when supply cannot expand, help to buyers is largely captured by sellers, as studies of UK Housing Benefit and of US student-loan limits have suggested, though the size of the effect is disputed.

Not every case is this clean, and the minimum wage is where the evidence is most contested. The competitive model predicts that a floor above equilibrium raises pay for those who keep jobs but reduces employment. David Card and Alan Krueger (1994) compared fast-food restaurants in New Jersey, which raised its minimum from $4.25 to $5.05 in 1992, with those across the border in eastern Pennsylvania, and found no relative fall in employment; David Neumark and William Wascher, using payroll records, reported the opposite.

Monopsony offers one reconciliation, since where employers set wages a moderate floor can raise both pay and employment, and UK evaluations have generally found small effects while Seattle's large increases produced conflicting results. The fair reading is that size matters and the argument is not settled.

Where the signal lies

The same reasoning that defends prices also explains when they mislead. A price summarises the costs and benefits felt by the parties to a trade, and nothing else. When a flight's fare covers fuel, crew and airport fees but not the carbon dioxide it adds to the atmosphere, the signal is too low and output too high. These externalities are the classic failure, and the classic remedy is to put the missing cost back into the price, as with Sweden's carbon tax from 1991 or London's congestion charge from 2003, after which traffic in the zone fell noticeably.

Market power distorts the signal in a different way. A seller facing the whole demand curve, as De Beers largely did for much of the twentieth century, can hold supply back to support the price, so the number reflects restricted output rather than genuine scarcity. Hidden information does further damage. George Akerlof (1970) showed that when sellers of used cars know their quality and buyers do not, buyers offer only an average price, good cars are withdrawn, and in the extreme only lemons remain.

Each failure has the same anatomy: the price is still carrying information, but the information is incomplete or deliberately shaped. That diagnosis points to remedies that repair the signal (a tax on the external cost, competition policy, warranties and disclosure rules) rather than remedies that replace it.

Reading the number

The practical consequence is a change in the first question asked of any price. Instead of whether it is too high or too low, the better question is what it is reporting, and whether the report is complete. A high rent in a city where planning restrictions limit building is an accurate message about scarcity; a cheap flight that ignores its emissions is an inaccurate one.

That distinction separates two kinds of intervention that are often confused. Capping an accurate signal suppresses information and invites queues, shortages and quality cuts. Correcting an inaccurate one adds information the market was missing. Much of the disagreement about price policy turns on a factual question: which kind of signal is being dealt with.