MAUP-ing it up.

Geographers are well aware of the Modifiable Areal Unit Problem (MAUP). Basically, MAUP causes the results to change when you change the unit of aggregation. This is especially true of things like rates. 



As I've written elsewhere, what unit of geography you use for urban analysis matters. Anytime anyone uses city boundaries, be suspicious. By cherry-picking cities (either within the nation, or within a metropolitan area), you can generate all sorts of spurious correlations, and really bias what your results are.

Urban Boundaries

Using cities for urban analysis is a joke. Travel behavior (commuting, shopping, traffic) doesn't stop at a city boundary. Cities are surrounded by suburbs, and suburbs are tied together by commuting flows. That's been recognized since the '70s, with the advent of Metropolitan Planning Organizations. Not to say that MPO boundaries are perfect.  Metropolitan Areas are made up of county boundaries, and inclusion within an MPO is based on the urban commuter flows.  Counties are large, and contain rural and urban parts--many analyses (notably of density) naively use county area totals. The best unit of analysis I know is the Federal Highway Urbanized Area boundary files. Sadly, each state prepares its own, and there is no central repository. 

Randal O'Toole likes to cherrypick his examples.

"By comparison, the average speed of auto travel in most American cities is more than 30 mph. The slowest city is New York, at 17.6 mph, which helps explain why New York also has the highest rate of transit usage. The only others under 20 mph are San Francisco and Washington. At the other extreme, average speeds in Kansas City and Tulsa are more than 40 mph, probably because those cities, unlike so many others, haven’t actively tried to discourage driving in a doomed effort to get people to ride transit." - AntiPlanner

Comparing Tulsa to NYC is a joke. NYC has a population of 8.5 million; Tulsa has a population of barely 400,000. Which means NYC is 20 times larger. 

O'Toole didn't present from the Texas Transportation Institute, which has data for the top 50 metropolitan areas. And guess what? The top 10 are all really congested!