More on weighted density for the crackpot blogging stats geeks

I've hit on this in the comments to other posts, but one of the nice features of weighted density is that it permits us to compare apples to apples (even though the comparison may be royal gala to red delicious).  Standard density is extremely sensitive to geographic boundaries.  This is why most debates over density quickly turn into squabbles over the proper geographic unit.  CMSA?  MSA?  Urbanized area?  Central city?  And then it's just a matter of time before someone brings up City  X's moutains or parks or other uninhabitable land that drag down its density. 

Because weighted density is determined by the density at which most people live, adding or subtracting land at the margin -- even a lot of land -- does not matter much unless the land contains a large chunk of the population.  Weighted density does not depend on precisely how the geographic area is defined.

Example:  In 2000, Houston's urbanized area had a population of 3,822,000 and a standard density of 2,951/square mile.  Houston's Primary Metropolitan Statistical Area* had nearly four times the land area but only about 355,000 (10%) more people.  (The PMSA has a lot of cows.)  As a result, Houston's PMSA had a measley 706 persons per square mile, just a little over one person/acre.

The PMSA's weighted density, however, was a more respectable 4,296.  That's only 5% off the urbanized area's weighted density of 4,514.  Adding a bunch of cow pastures does not change the fact that the vast majority of the PMSA population was bunched at a much higher density.

The same holds for Portland.  In 2000, its PMSA had a standard density of just 381/square mile.  But the PMSA's weighted density was 3,943/square mile, down just 10% or so from the urbanized area's weighted density of 4,383. 

Note that a lot of the vacant land around Houston is vacant simply because no one has gotten around to developing it yet.  A lot of the land around Portland is vacant because of policies designed to preserve open space and slow sprawl.  This difference is often a flash point for disagreements over geographic boundaries. Weighted density allows us to sidestep these arguments by focusing on the density at which the average person lives.

*A PMSA consists of one or more counties within an MSA that have substantial commuting interchange.

Comments

[john]

I've really enjoyed your series of density posts and your thorough analysis.
I think that your Portland and Houston examples bring up the issue of urban form. From my experience, the most successful and pleasing metro areas have a certain level of density that supports a lot of walkable neighborhoods and urban transit; however, these cities also have "tendrils" of nature.
I think your analysis misses the benefits of such tendrils and wild spaces that can be found in cities such as Portland, Seattle, San Francisco, Pittsburgh and maybe others that include very large park systems or wild spaces within the MSA.
For me, the biggest advantage of a place like San Francisco over New York City is that you have this intensely urbane environment a short distance from some very sacred natural places that offer an obsolute 180-degree alternative to the intensity of the city. I think Phoenix even scores better on this than a place like Houston; Phoenix has the very large South Mountain Park and more than 24,000 acres of desert and mountain parks within the city's boundary.

[john]

Oh, I guess it would be me that brings up the mountains and parks issue.

[AC]

Weighted density certainly doesn't measure the quality of the density. It can't distinguish between a walkable nieghborhood of low-rise apartments with street-fronting shops and a clump of gated apartment complexes on a busy highway. It's just a better measure than standard density.

As for parkland, open green space, etc., I should have been clear why I mentioned it: It doesn't affect the weighted density calculation, unless it is distributed in small parcels among a bunch of different census tracts. Again, weighted density doesn't tell you whether there's a lot of green space, but it's not dragged down by a lot of green space.

[M1EK]

john, what AC said is true - the density QUALITY problem is one that's impossible to solve - for instance, you can have medium-to-high-density unwalkable sprawl - like the townhouse complexes that infest DC suburbs.

[The Overhead Wire]

Well it might be possible. If you have network analyst in ArcGIS you can take Tiger files and measure block size and street connectivity through nodes. I imagine if you can match that with the weighted density you could tell how good the urbanism is in a given neighborhood. That however would take a lot of work.

[The Overhead Wire]

I reread and thought you mean to measure M1ek. You're right, creating new good urbanism is a hard problem to solve for sure.

[AC]

Is there a standard measure of street connectivity? Something like "intersections per square mile"? I'm trying to think of something that approximates how easy it is to get around (by car or foot) in a neighborhood. I think one neighborhood would be superior to another in connectivity if the average distance (by car or foot) between two points in one one neighborhood is shorter than in the other neighborhood.

[M1EK]

Nobody has good enough data on sidewalks - Austin (mis)spent hundreds of thousands generating an inventory of such a few years back rather than just getting on with it and starting to build in spots they knew didn't have them.

What you'd really need, in other words, is a google pedestrian map - where only sidewalks are used (no, Texans, walking in the street isn't acceptable). Right now, the ped component of transit trips just assumes "as the crow flies" because they don't have any better data.