My issue isn’t that they claim the increased congestion comes from additional personal trips. My issue is that they claim this is due to increased capacity which fundamentally seems untrue, as evidenced by the fact that traffic doubled both with and without increased road capacity.

In fairness, I have not read the 37 page paper yet. Maybe the paper makes a more compelling case than the summary article.

▲

NomNew 12 hours ago | parent [-]

I'm not sure I see the nuance in the two issues you point to. The paper looks at how adding new lanes changed congestion on these roads. The results found no change in congestion, as shown by VKT growing the same amount as the increase in road capacity.

A driver, besides the cost of the car and fuel, faces a time cost. An extra lane will reduce the time-cost, assuming no new vehicles enter the road. But if time costs fall, it's in effect 'cheaper' to drive. So previous car trips that were not happening, because the time-cost was too high, are now occurring. The amount of extra car-trips is such that we are back to the same time-cost as before the road expansion. That's why VKT ends up growing one for one with road expansion.

▲

dpark 10 hours ago | parent [-]

I did not read the paper yet, but the article says that KVT doubled in 20 years for both highways (which did not add new lanes) and urban roadways (which did). So we see KVT double regardless of lane expansion.

To me this says that lane expansion is not the driver of KVT (at at least not the primary one).

	▲	NomNew 9 hours ago \| parent \| next [-]
		Ah, I see. The key question the authors look at is what is the impact of adding a lane to VKT - this they found one-for-one growth. They control for many other factors that may have led to the growth over the observed time period (population being the big control). The broad question is what would reduce congestion - currently, on the margin, adding a lane will not reduce it. I imagine that hypothetically adding many lanes may reduce congestion, although there may be city driven bottlenecks that don't make this feasible at all.
	▲	NomNew 9 hours ago \| parent \| prev [-]
		And also - thanks for the comments and questions!