The Urbanist Musings of Pete Saunders

The Density/Vacancy Relationship

One view of the gentrification process.  Source:

About this time last year I wrote a blog post that attempted to develop a gentrification typology for large U.S. cities.  I did this because of three things I found in my research and writing — 1) the older, denser, less auto-dependent and mixed-use development forms of some neighborhoods in older cities were becoming more attractive to a larger segment of the American people; 2) the presence of African-American neighborhoods was having some impact on the revitalization of these neighborhoods, and 3) gentrification activity is not the same in all cities and should not be reported on as such.  I think I’ve found a way to arrive at similar yet clearer conclusions by relying on just a couple of data points and avoiding any previous contrivances.

First, a quick review of the earlier piece.  By making estimates of a city’s relative development age by looking at its population growth profile, and looking at the numbers of black residents each city had in 1970 (a critical year because it roughly coincides with the end of the Great Migration that brought many blacks to major Northeastern and Midwestern cities), I drew some conclusions about gentrification activity in the 64 U.S. cities that had populations of more than 250,000 in 2010.  Here’s the chart I did:

And the general conclusion I drew from it:

There are more than a few cities that are exceptions, largely because recent consolidations or large-scale annexations have boosted them into more unfamiliar boxes. But some patterns are evident, and if you think of these in terms of gentrification, you might be able to make the following general assumptions:

Old Form + Low Black Population = Expansive Gentrification (OFLB)

Old Form + High Black Population = Concentrated Gentrification (OFHB)

New Form + Low Black Population = Limited Gentrification (NFLB)

New Form + High Black Population = Nascent Gentrification (NFHB)

This construction worked reasonably well, but there were some outliers that made things questionable, and frankly, I could see how some would quibble with how I arrived at the old/new development forms or low/high black population formulations.  So I went back to the drawing board.

This time, I examined the 25 largest U.S. cities in terms of population, and looked at two data points — density (as in gross people per square mile), and housing vacancy (as in the percentage of unoccupied housing units).  Density was used as a relative proxy for old or new development form.  It’s not always true, but generally higher city densities correlate pretty strongly with older cities.  Cities that boomed in the 19th century, for example, are more dense than those that boomed in the early 20th century, which are more dense than those that boomed in the latter half of the 20th century (think Philadelphia > Los Angeles > Houston).  Housing vacancy was used as an indicator of housing demand in each city.

I pulled the data points for the 25 cities for 2000 and 2010, arrived at an average for them, and placed the cities in one of four quadrants as a result.  Here’s how the quadrants look:


In 2000, the average person per square mile density for the 25 cities reviewed was 6,153, and the average housing vacancy rate was 6.7%.  Here’s how that appears on a graph:

In 2010, the average density was up slightly to 6,269, while the average housing vacancy rate shot up to 11.2% (thank you, Great Recession).  Nonetheless, the relationships remained pretty stable:

Here are my takeaways on this:

  • City density does seem to be increasing.  Overall, average population density in the 25 cities increased by 1.9%, from 6,153 in 2000 to 6,269 in 2010.  However, the seventeen cities on the low density end of the spectrum increased density from an average of 2,742 people per square mile in 2000 to 3,018 in 2010 — an increase of 10.1%.  Much of this could be cities sprawling on recently annexed and previously undeveloped land, but it’s conceivable some cities may be trying a new development approach.
  • But the Great Recession was hell on cities.  The U.S. Census may not be the best indicator of a fluid data point like housing vacancy rate, but going up from 6.7% vacant to 11.2% vacant is astounding.
  • Detroit’s increase in housing vacancy from 2000 to 2010 flung it into an entirely different realm.  In 2000 Detroit’s density/vacancy profile put it in a class with Chicago, Philadelphia and Washington, DC.  By 2010 it was decidedly not, and in a class by itself.
  • Washington, DC will soon join the select group of high-density/low-vacancy cities.  DC appears to be undergoing this transition not through the construction of new units, which would further drive up overall population and density, but through reducing vacancies in heretofore unknown neighborhoods.
  • The cities in the high-density/low-vacancy quadrant are the ones best known as driving the national narrative or discussion on gentrification.  Conversations about general housing affordability, rapidly changing neighborhoods, low-income resident displacement, rapidly growing income and wealth inequality all define the cities in this quadrant — New York, San Francisco, Boston, Los Angeles and (slightly surprising to me) Seattle.
  • The low-density/low-vacancy cities here tend to be the “nouveau riche” of American cities.  In addition to Austin and San Jose, shown on the graphs, cities here include Charlotte, Denver, Nashville and San Diego.  All are highly desirable cities whose explosive growth has been rather recent (i.e., post mid-20th century).
  • The low-density/low-vacancy cities may face the same challenges as the high-density/low-vacancy cities, but at a lesser scale.  As the demand for living in these cities increases, they will have to decide what is the most suitable development path to address future affordability, displacement and inequality issues.
  • Continued sprawl may mask larger problems in the Sun Belt version of the low-density/high-vacancy cities, but deeper problems likely exist with the rest.  Cities like Phoenix, Houston, Dallas and Jacksonville may still be able to focus on outward development, but cities like Indianapolis, Memphis, Detroit and Columbus may have land abundancy issues (either through abandonment or annexation) that make revitalization more difficult.
  • The high-density/high-vacancy cities may align themselves with the high-density/low-vacancy cities, but whatever economic challenges they have that prevent them from higher housing demand causes problems.  Cities like Chicago and Philadelphia do indeed see themselves in the same light as New York and San Francisco, as does Washington.  But the economic renaissance that has remade the latter cities hasn’t fully come to the former, and that means they travel on a different trajectory.
So what does this mean with respect to gentrification activity in American cities?
The data points I used for this could lead to this general conclusion:
I think this passes the smell test.  As I said in the earlier piece, gentrification looks different in different types of cities.  In cities that are already dense yet also in high demand the gentrification process is widespread, touching neighborhoods in all areas, and dominates public discourse about the future of the city.  In cities that are dense but lack the overall housing demand, the gentrification process is mature but concentrated in select areas; in newer cities that grew in low density environments but are experiencing high housing demand, gentrification is in its early stages and not a major feature of city discussions.  And there are yet other cities where outward expansion still rules, largely because cheap land is still available.
There should be different policy prescriptions for each city type.  I’ll get to those in a future post.

2 Responses to “The Density/Vacancy Relationship”

  1. DevinComments

    Pretty interesting stuff. I would love to see you graph the changes as vectors, as it would allow for some interesting \”eyeball\” tests. I think in a very loose way, that the length of the vector might correlate with rate of gentrification, albeit averaged over the whole city. The slope of the vector could also be general evidence of the influence of added supply on gentrification – you would expect that horizontal movements could mitigate vertical movements. Of course these would be pretty unempirical, and wouldn't capture the complexity of the situation, but would still be interesting as broad observations.


  2. greg

    Reading this, I do wonder about how the vacancy rate relates to churn. I've been comparing ACS vacancy rates with USPS rates lately ( ACS is consistently higher, I believe because USPS waits three months before declaring a property vacant, while ACS is more like a snapshot. (I believe — I've contacted HUD to clarify this. One nice thing about the USPS data is that it has categories for length of time vacant.) If that's the case, ACS captures churn, whereas USPS would do a better job of capturing slack demand. I doubt it would change your overall charts, but it's an interesting wrinkle.



Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

Basic HTML is allowed. Your email address will not be published.

Subscribe to this comment feed via RSS

%d bloggers like this: