Induced traffic

Why does a new lane added to a congested road end up just as congested?

The reasoning behind a road widening is almost impossible to argue with. Here is a road carrying more cars than it was built for. Add a third lane and it carries half as many again, so the queue should shorten. Yet the same road, a few years later, queues much as before — often after every stage of a decades-long programme of widenings. Something is wrong with an argument that seems watertight. What?

The flaw is a hidden assumption. The widening argument treats the number of cars wanting to use the road as fixed — a given quantity of demand, to be met by more or less supply. Is it fixed?

Ask what a driver actually faces. Not a road, but a choice: this route or another, this hour or an hour later, driving or the train, going or not going at all. What governs those choices is not distance but the cost of the trip, most of which is time. Congestion is a price, paid in minutes. The assumption now looks strange, because we have described a quantity demanded that depends on a price — and the widening lowers the price.

Follow what a lower price does, in order. Immediately, drivers who had been detouring onto side streets come back. Slightly later, people who left at six to beat the traffic drift back toward the hour they preferred, and the peak re-forms. Then some who had taken the train or bus, now the slower option, drive instead. And underneath all three, trips never made at all — the meeting worth attending only if the journey is thirty minutes, the job across town not worth the commute — become worth making. That last group is latent demand: not new people, but journeys suppressed by the old cost and released by the new one.

Each refills part of the added capacity, at different speeds — route switching within weeks, timing and mode shifts within months, new trips over a year or two. That staged filling is what a naive "the road is bigger now" model cannot produce, and it is what people observe.

Now the slow channel, easiest to miss and mattering most. Cheap travel changes where things are built. A widened corridor makes land at its far end usable: houses go up, a distribution park opens, retail follows, and people rationally accept a longer commute for cheaper space. Ten years on, the same road serves a population whose journeys are longer. That feedback runs on the timescale of planning permission and mortgages, and does not reverse.

How large is all this? The clearest estimate comes from Gilles Duranton and Matthew Turner, who compared American metropolitan areas over roughly two decades and found the elasticity of vehicle-kilometres travelled with respect to lane-kilometres to be close to one: increase the road stock by ten per cent and driving increases by about ten per cent. They named it the fundamental law of road congestion — and found, too, that expanding public transit did not reduce road traffic either, because space freed by anyone who switches is taken by someone waiting for it.

Two things are easily confused. Induced traffic is not the claim that roads cause driving out of nothing, nor that drivers are irrational. It is the claim that trip-making responds to its cost, that congestion is the cost, and that the system settles at whatever delay travellers are just willing to tolerate. Capacity moves the quantity of travel, not the equilibrium delay.

That has a consequence both sides of the argument tend to skip. If the new lane fills, the road is carrying more trips than before, and those trips have value — someone is reaching a job, a hospital, a customer previously out of reach. A widening ending in the same queue is not worthless; it bought access rather than speed. The honest conclusion is not "roads are never worth building" but "congestion relief is the wrong justification", and a scheme appraised on time savings that will be competed away is appraised on a benefit it will not deliver.

Caveats belong on the finding itself. The unit elasticity was estimated for major roads in American metropolitan areas over a particular period and is not a universal constant — the response depends on how much suppressed demand exists, small in a shrinking region and large in a growing one. Short-run relief is real and can persist for years. And a genuinely new link — a bypass taking lorries out of a village, a crossing where none existed — is doing something other than adding capacity to an already saturated corridor.

What holds congestion down, if capacity does not? Only something that prices or rations the scarce thing directly. Charging for road use at the peak, as Singapore, London and Stockholm do in various forms, has the clearest record, precisely because it acts on the price rather than the supply — the same move the restaurant makes when it starts taking reservations.

• How congestion charging in Singapore, London and Stockholm changed traffic volumes.
• Whether removing road capacity produces the mirror image, traffic evaporation, and how well that is evidenced.

Every term the collection defines is gathered in the glossary.