Federal government requirement to locate near Metro.

Federal employee Metro subsidy.

Is the data for the incorporated city only? Looks like it. I don't think Metro Washington region is 10,000 persons per sq mi.

City figures do strange things when you draw the boundaries differently. For example, LA metro area is denser than NY metro area.

by Michael P on Sep 17, 2008 6:12 pm • link • report

Yay DC!

But to follow up on Michael P a bit, it does strike me as slightly strange to compare transit use of the metro region with density of the central city. I get that the denser a city is, the easier it is for its residents to use transit rather than automobiles. But does the density of DC have much of a relationship to whether a commuter from Loudon hops on the metro?

What does the chart look like when you compare regional densities with regional transit use?

by RyanA on Sep 17, 2008 6:18 pm • link • report

How are the cities on this graph chosen? It is not the top x densest cities. Minneapolis (just one example) is absent. It has a density of 6200residents per square mile. I can show whatever relationship between two variables I want if I choose the right data points.

by Ross on Sep 17, 2008 7:10 pm • link • report

Maybe the simple explanation is the D.C. region's average commute time?

http://www.washingtonpost.com/wp-dyn/content/article/2006/08/29/AR2006082901651.html

As car travel gets slower and slower, the advantages of public transit begin to seem more and more appealing.

by Lindemann on Sep 17, 2008 7:12 pm • link • report

To echo Ross, Honolulu isn't on here.

Density of the center city is probably the biggest determinant (more so than density of the metro). You need enough ridership to support a train, and you need trains to get more than 11% of people to commute by mass transit. If the jobs are densly located you can have millions of people have a quarter mile or shorter walk to work.

If you like LA's "core" density ... its just like tyson's corner

by Kiran on Sep 17, 2008 7:36 pm • link • report

It is important to look beyond simple population density. DC has the third largest commercial core in North America --ranked in terms of millions of square feet of office space (it follows NY and is neck and neck w/ Chicago). This clearly leads to a much larger "daytime" population which in my view accounts for the higher transit use.

by Alan on Sep 17, 2008 8:27 pm • link • report

Another point: many sunbelt and midwestern cities ostensibly annex their suburbs. Indianapolis and Louisville, for example, have annexed the entire county into their municipalities. I'm sure this distorts the numbers slightly.

However, I wouldn't be surprised if the DC suburbs had higher ridership than the less dense areas/suburbs of places like Kansas City, Houston, and Phoenix. I'd be interested in seeing that study.

by Dave Murphy on Sep 17, 2008 9:37 pm • link • report

I'm surprised no one has mentioned the obvious reason why DC bucks the trend - DC actually has a good rail transit system. Every city above DC on that list of density has a strong legacy system of pre-war rail transit (though SF does have BART, the Muni system dates back a long time).

Imagine if Baltimore, Milwaukee, and Detroit had received the same kind of mass transit investment that DC did. Transit ridership would be much higher (and parts of the cities would be much denser, too).

by Alex B. on Sep 17, 2008 10:40 pm • link • report

Off-topic, but what struck me about the graph was: what a poor choice in graph design! Better to put density on the x-axis and transit usage on the y-axis and make each city a data point. As it is, the x-axis is made to look like a continuous variable, but it isn't. The lines connecting the data points make it appear as if the data are discrete samples of a continuous data set--not even close. Belmont ought to work through Edward Tufte's books.

by thm on Sep 18, 2008 9:26 am • link • report

Dave -

As someone who recently lived in the suburbs of Kansas City I can assure you, no one there takes public transit. It's almost completely nonexistent. It's really sad.

by RyanA on Sep 18, 2008 9:32 am • link • report

Michael P: "Federal government requirement to locate near Metro."? I'd love for this to be, but if so, how did we get these massive relocations to Ft. Belvoir under BRAC?

by Lou DC on Sep 18, 2008 10:35 am • link • report

Lou, it's not a requirement that all Federal locations be near metro, but there is a GSA rule (covered in Zach Schrag's history of the Metro book) that any Federal agency leasing private office space must locate near Metro. That obviously doesn't apply to Federally owned buildings, and I'm not sure GSA rules would apply to military installations anyway (for the examples you mention related to BRAC).

by Alex B. on Sep 18, 2008 11:03 am • link • report

Good to see that you're reading Belmont. As you know, I think it's the best book on urbanism since _Death and Life_. And to Steve P., I'll give you your copy back at the alley party at the end of the month. (I finally bought my own copy.)

by Richard Layman on Sep 18, 2008 12:44 pm • link • report

thm -- not off topic at all. That was bugging me too. I guess the city labels were too difficult to see if they were included next to each point on an x-y scatterplot.

In fact, I went ahead and eyeballed the numbers and typed them in to make the more appropriate graph myself in Excel. The story is the same. DC is an outlier and everyone else lines up on a trendline. Incidentally, NYC and SF are slightly below the trendline. The city with the next largest positive residual (higher than predicted transit usage) is Boston.

by Ward 1 Guy on Sep 18, 2008 1:05 pm • link • report

Ward 1 Guy: Want to send me the Excel? I'd love to post the graph.

by David Alpert on Sep 18, 2008 1:07 pm • link • report

An important thing to note is that we're in the realm of exploratory data analysis and not modeling, especially because of the outliers. Therefore, we cannot attach much importance to the fit (line). The line is just an empirical descriptive device. Understanding the data more fully requires many more variables, many of which may not be quantifiable. How exactly do you quantify a convenient mass transit system? Assuming you can do this, you can create a model to explain mass transit use and evaluate the effects of different policies.

by Chuck Coleman on Sep 18, 2008 9:55 pm • link • report

Anybody want to compare these numbers to European cities?

by John on Sep 21, 2008 9:57 am • link • report