I generally agree, but there are better versions of the argument. One of them is the Braess network paradox[1]. In short, under a model with simple game-theore

I generally agree, but there are better versions of the argument. One of them is the Braess network paradox[1]. In short, under a model with simple game-theoretic assumptions about behaviour ("people find the shortcuts") and traffic/speed congestion functions an added road/connection can do more harm than good -- in the extreme case, slowing down every actor. The toy model may not actually be realistic, but I think reality might not be too different. Say the congestion delay function is smooth, and one added car doesn't change travel time much. How much does one of the new trips with the new infrastructure improve the life of the driver over the status quo? Very little: if travel times are the same, they only benefit to the extent that their trip would have slowed down traffic before. So generally you might say, - Adding the lanes but leaving behaviour the same would speed up traffic, improving global utility, - Letting traffic reach equilibrium (if speeds go back to what they were before -- big assumption) removes that benefit for existing travellers and doesn't much improve the life of new travellers, and is a net negative for global utility nearly equal in size to the earlier increase. - Maybe an argument for tolls? 1: https://en.m.wikipedia.org/wiki/Braess%27s_paradox https://en.m.wikipedia.org/wiki/Braess%27s_paradox