A while back, Witold Rybczynski wrote an article in Slate about how urban sprawl was inevitable, had happened throughout history, and was impossible to stop. Naturally, it was pointed out that while that might be true, not all sprawl was created equal. Some forms are worse than others. A random bit of clicking around the World Bank's site dredged up this old study which can make some of the differences here a bit more palpable.
To see what we're dealing with here, take Boston (for some reason there's incomplete data on San Francisco). Boston's already a fairly spread-out city, but if its "population centrality" was as spread-out as Atlanta's, Bostonians would be driving about 9 percent more. Public transit also matters: If Boston had Atlanta's rail system, total driving would increase by about 5 percent. (If it had a transit system as shoddy as, say, Dallas', there would be even more driving.) The distribution of jobs to housing matters too. Boston has a very even mix in this regard, but if it was to become more unbalanced like, say, Washington D.C., driving would increase by roughly 9 percent. (I'm eyeballing the calculations here.)
That doesn't seem like such a big deal, but taken together, these changes start to have a real impactthe authors point out that if you could wave a magic wand and make Atlanta similar to Boston, then total driving per household would decrease by 25 percent. Obviously no one has a magic wand, and maybe people prefer living in Atlanta-type cities to Boston-type cities, although who can fathom why, but it's certainly something to consider.