On the other hand, taking existing empirical data and extrapolating linearly from it seems like a bad idea in this case. There are obvious limits to how much latent demand is there. IIRC, we never did a proper experiment of suddenly e.g. quintupling the capacity of some major congested road.
That said, I think the source of this paradox is ultimately not in the road one is thinking of expanding, but in all the connections between it and other roads. It buys you little to turn one lane into five if all the exits are still the same capacity.
While there obviously has to be an upper limit somewhere, this isn't (only) about latent demand, but induced demand: People actually drive more because they have to. More lanes have more consequences than wider streets alone, they enable sprawl, funnel funding away from other transport options etc. That has consequences for the housing market, average distances between work and home etc.
Your second point is very important though, and IIRC is the main reason why so many city planners are so critical of the boring company idea of underground car tunnels: It tries to solve an aspect of traffic that isn't the problem.
> People actually drive more because they have to. More lanes have more consequences than wider streets alone, they enable sprawl, funnel funding away from other transport options etc. That has consequences for the housing market, average distances between work and home etc.
You're really just saying the same thing as I did -- congestion suppresses demand.
Suppose you have an uncongested two lane road. Making that a four lane road is not going to induce anything, because the widened road has more capacity than the original one, but you weren't even using the full capacity of the original one. It doesn't encourage anyone to do anything they hadn't already been able to.
The reason it "enables sprawl" is that if the original road was congested, people wouldn't want to travel it to commute, so buying a house in the suburbs and commuting via that road is suppressed. Congestion suppresses demand. If you reduce the congestion by widening the road, some of the demand comes back. And notice that the same thing happens if you alleviate congestion in some other way -- if you somehow make it more attractive to take the subway and then more people do, there is less traffic on the roads and it becomes more attractive to drive.
But you don't actually want the congestion. Traffic congestion sucks. You want to get rid of it somehow.
Part of the solution could be expanding the road, but you could have to expand the road a lot to satisfy all of the demand that would exist with uncongested roads, if you want to use that as the only solution. Which is why you shouldn't. There are other things that reduce congestion.
Like building higher density housing, so that more people can live closer to where they work and have a shorter commute, which in turn makes mass transit more viable. Then that in combination with widening some roads relieves the congestion.
The real problem here is the politics of it. Increasing the housing stock sufficiently to satisfy the demand would tend to reduce housing costs, which existing property owners don't want. So they fight high density construction with restrictive zoning rules and then the high cost and low supply of housing in the city pushes people into the suburbs, which increases the demand for roads.
That said, I think the source of this paradox is ultimately not in the road one is thinking of expanding, but in all the connections between it and other roads. It buys you little to turn one lane into five if all the exits are still the same capacity.