The Road That Made Everyone Slower
A fable, with the arithmetic left in. After Pigou (1920), Wardrop (1952), Braess (1968), and Roughgarden–Tardos (2002). Every number below was rebuilt from a seed; nothing is taken on anyone’s word.
There was a town whose people all worked at the same market, and all lived on the same far side of a low green valley, so that every morning four thousand of them made the same crossing from home to market and every evening made it back.
There were two ways across, and they were mirror images. You could climb the Old Hill on the north side — a fixed, honest slog of forty-five minutes no matter how many climbed it with you, because a hill does not get steeper when it is crowded — and then drop down onto the North River Road, which was the opposite kind of thing: fast when empty, slow when full, its travel time simply the number of wagons on it divided by a hundred. Or you could take the South River Road first, mud that thickened with every wheel, and then finish over the South Hill, another flat forty-five minutes of climbing. North wagons and south wagons never met; the two routes shared nothing.
So the town sorted itself, as towns do, until neither route was worth switching off of. Two thousand went north, two thousand went south. On the North River Road, two thousand wagons made it a twenty-minute road (two thousand over a hundred), and with the forty-five-minute hill that was sixty-five minutes to market. The southerners, by their mirror, spent sixty-five as well. Nobody could do better by changing sides — a lone wagon defecting north would only make the north road the tiniest bit muddier for itself and everyone already on it — and so sixty-five minutes was the settled price of the crossing, the thing the whole town had quietly agreed on without a meeting.
This settling has a name. In 1952 a British engineer named John Glen Wardrop wrote down the principle every commuter obeys without being told: at rest, every route actually used costs the same, and none unused costs less. No one can gain by switching. It is not fairness and it is not efficiency; it is only the flatness a pile of self-interested choices reaches when it stops moving. Economists would later borrow a grander word for the same flatness — a Nash equilibrium — but the wagons had been finding it every morning for a century before it had a name.
The gift
Now a benefactor came to the town — a clever engineer, genuinely trying to help — and looked at the valley and saw waste. Look, he said: a northbound wagon climbs the Old Hill, forty-five honest minutes, when down below, just across a narrow neck of the valley, the South River Road is right there. And a southbound wagon slogs the South Hill when the North River Road sits beside it. Why should anyone climb a hill at all, when I could lay a little level causeway across the neck — a few planks, costing so near to nothing to cross that we may as well call it free — and let a wagon take the North River Road down, skip across my causeway, and finish on the South River Road? Two fast roads and no hill. The town thanked him and built it.
Here is what happened, and it is the whole point of the fable, so watch the wagons and not the intentions.
A northern wagon does the arithmetic every wagon does. It is on the North River Road. Ahead lie two ways to the market: the Old Hill’s honest forty-five, or the free causeway onto the South River Road, whose time is however many wagons are on it, over a hundred. On the first morning the South River Road is nearly empty, so the causeway route is much faster, so the wagon takes it. Of course it does — it is right for that wagon. And so does every northern wagon. And, by the perfect mirror, so does every southern wagon, pouring onto the North River Road. And now both river roads carry all four thousand wagons, and each is a forty-minute road, and the crossing that used to be a fast road and an honest hill is now a fast road and another fast road, both ruined — forty minutes, plus the free causeway, plus forty minutes: eighty minutes to market.
Nobody was foolish. Every wagon, at every moment, took the route that was fastest for it given what the others were doing — that is exactly Wardrop’s flatness, and you can check that it holds: a wagon on the eighty-minute causeway route that tried to peel off onto the Old Hill would find the hill still costs forty-five, and the muddy road up to it still costs forty, for eighty-five — worse. So no one defects. The town has settled again, honestly, into an equilibrium every member is individually powerless to leave, and it costs each of them fifteen minutes more than the morning before the gift. A road that was free to cross made every single journey longer. The benefactor built exactly what he promised — two fast roads and no hill — and the promise was the trap.
This is Braess’s paradox, named for Dietrich Braess, who found it in the mathematics of traffic planning in 1968, though the wagons, again, had been suffering it silently forever. I rebuilt the whole crossing from a seed to be sure I wasn’t being told a story: sixty-five minutes before the causeway, eighty after, by two independent methods that agree — the direct best-response reasoning above, and a second route entirely, minimizing a certain energy-like quantity (the Beckmann potential, the sum over each road of the area under its congestion curve) whose lowest point is provably where the self-interested wagons come to rest. Both roads land on eighty. The paradox is not an artifact of how you count. It is real.
Where the cost hides
Why does the gift curse? Because the causeway lets each wagon do a thing that is good for it and bad for the sum: dodge its own honest, uncongestible hill by piling onto a road that everybody shares. The forty-five-minute hill was a cost that fell only on the wagon that climbed it. The muddy river road is a cost that each wagon spreads onto all the others: when I roll onto it, I add a hair of mud, a hair of delay, to every wagon behind me — and I do not feel that hair, so I do not count it. The causeway’s whole function is to route everyone off the costs they bear alone and onto the costs they inflict on everyone else.
This — a cost you impose but do not pay — is the oldest idea in the fable, older than Braess by half a century. Arthur Cecil Pigou described it in 1920: the externality, the part of the reckoning that lands on strangers. A wagon choosing the muddy road weighs only its own forty minutes; it never weighs the delay its presence adds to the four thousand. Selfish equilibrium is inefficient for exactly this reason and no other — every agent optimizes its private ledger while the shared ledger goes unread.
You can see how much goes unread by asking what a wise planner would do — one who could simply order each wagon where to go, to minimize the town’s total time on the road rather than any one wagon’s. Here the fable turned on me, and I want to tell it straight, because the truth is better than the story I expected. I had assumed the planner would simply forbid the causeway — tear out the gift, send two thousand north and two thousand south as before, total time four thousand times sixty-five. I was wrong, and the arithmetic said so. The planner uses the causeway — but barely. He sends only five hundred wagons across it, and splits the rest evenly over the hills, and the town’s total falls below where forbidding it would land. The causeway is not evil. It is a good road used in the wrong amount. Selfishness sends four thousand wagons across it; wisdom sends five hundred. The paradox was never that the new road is bad. It is that unpriced self-interest overuses it, eightfold, and drowns.
How bad can the overuse get? This is the part that should comfort you a little, after all the alarm. In 2002 Tim Roughgarden and Éva Tardos proved a ceiling. For roads whose congestion grows in a straight line — as the muddy roads do, time rising as flow over a hundred — selfish routing can never be worse than four-thirds as costly as the wise planner’s optimum. Never mind the town, never mind the causeway; over every network of such roads, anarchy costs at most thirty-three percent more than perfect coordination. The tight case, the worst there is, is almost insultingly simple — a single fork, one road that always takes an hour and one that takes an hour times how crowded it is; the crowd floods the second, pays one full hour, when splitting evenly would have cost three-quarters. I rebuilt that fork too, and out came four-thirds exactly. Our town’s own waste is smaller — two hundred fifty-six over two hundred seven, about twenty-four percent — comfortably under the ceiling, because the town is not the worst case, only a famous one.
The price, and the string
So what heals it? Not, note carefully, tearing out roads. The wise planner kept the causeway. The disease was never the road; it was the unpriced cost. Pigou’s own cure follows straight from his diagnosis: make each wagon pay the delay it inflicts. Put a tollkeeper on each muddy road who charges you exactly the extra time your presence costs everyone behind you — a marginal-cost toll — and now the private ledger and the shared ledger are the same ledger. I added those tolls to the model and re-solved, and the selfish wagons, each still doing nothing but minimizing its own bill, walked themselves onto the planner’s optimum: five hundred across the causeway, the rest over the hills, the town’s travel time back down to the best it can be. The toll does not ban the gift. It prices it, and the pricing is the whole repair. The road you build must carry, written on it, what it costs everyone else — or it becomes a trap made entirely of good decisions.
I will end where the fable stops being about wagons, because the same shape runs through things with no drivers in them at all. In 1991 Joel Cohen and Paul Horowitz hung a weight from the ceiling by a chain of two springs joined in the middle by a short string, with two slack safety-strings hanging to either side, and they cut the middle string — the one thing holding the whole assembly together — and the weight did not fall. It rose. With the linking string intact the two springs hung in series, each bearing the full weight, each stretched long; cut it, and the load rerouted through the taut safety-strings so the springs now hung in parallel, each bearing half, each barely stretched — and the weight climbed. The middle string was the causeway. Present, it forced the whole load through one long series path and the weight hung low; remove it and the load split and the weight came up. I rebuilt the statics: the weight rises by an amount I can write in closed form, and the force balance holds in both configurations to the last digit. Adding the connection made the weight sink; the connection, freely taken, made the whole worse. It is Braess’s paradox in brass and steel, obeying no traffic law, only the same arithmetic of a load finding the path that is locally easy and globally wrong.
That is the fable, and here is the moral with the sugar left off. An extra choice, freely taken, can make the whole worse — and every step of the descent is somebody’s best response, nobody’s blunder, a gift accepted with thanks. The trap is built out of good individual decisions and one cost that no chooser is made to see. So when someone offers the network a new road, a new link, a new shortcut, and it is free and it can only help — ask what it lets everyone dodge, and onto whom the dodging falls, and whether anyone will ever be made to pay it. The cure is not fewer roads. It is the true price on each.
