Saturday, April 28, 2012

Density per net acre

After this discussion about 'The need for concentration' I found myself wondering about Jacobs method of measuring density in districts. She uses a metric that almost nobody else uses: dwelling or housing units per net residential acre, which counts only land designated for housing. Typical densities are expressed in terms of persons per square mile, counting all types of land. But this can include land that's really devoted to highways, parking lots, institutions, parks, or other places where people do not live. This has a tendency to skew the numbers because most boundaries are drawn to include arbitrary amounts of these non-residential uses. For example, Boston is said to have 12,752 people/sq mile, but residential neighborhoods such as Allston have a density of 18,505 people/sq mile and the North End is listed as nearly 30,000 people/sq mile, or more depending on how you draw the borders.

Thanks to a footnote in her book, we know that in 1960 the North End had a dwelling unit density of 275 per net acre of residential land, and Roxbury ranged between 21-40, according to planning commission measurements. But how does that compare to today, and to other neighborhoods? And what does it mean, anyway?

What are the proper densities for city dwellings? The answer to this is something like the answer Lincoln gave to the question, "How long should a man's legs be?" Long enough to reach the ground, Lincoln said. Just so, proper city dwelling densities are a matter of performance. They cannot be based on abstractions about the quantities of land that ideally should be allotted for so-and-so many people (living in some docile, imaginary society).
Still though, and perhaps somewhat uncharacteristically, she does attempt to give some hard numbers for what she feels is a reasonable categorization of different densities. Below ten dwelling units per net residential acre is typical for a "suburban" neighborhood. Densities from 10-20 indicate a kind of "semi-suburb" or a really dense suburb, but it cannot function well as part of a city. Then there is a range of "in-between" densities which are too dense to be a suburb, but not quite dense enough to generate necessary city vitality for safety and interest. According to her, this range seems to cover dwelling unit densities from above 20 to somewhere in the vicinity of 100 units per net acre. The exact numbers have to be determined on a case-by-case basis, asking yourself, "does this level of density produce true city life?" for each neighborhood.

With that in mind, I obtained both the Census 2010 data and the MassGIS Land Use 2005 survey. Using a PostgreSQL+PostGIS database I put together a series of queries to produce a map of housing unit densities per net residential acre for Boston. The query includes all land within a 10km radius of Park Street station. Then I uploaded it into a Google Fusion table and colored the areas using a scheme based upon her categories. I've adjusted the numbers somewhat though, for present-day Boston. In fifty years, I think that the way people live and their personal consumption patterns have changed considerably. In the interest of not spending a whole lot of time on this, I've settled on some choices which probably won't make everyone happy, but hopefully will be somewhat useful.

     60-80Borderline urban
     80-100Low density urban
     300+Extremely high density

The biggest caveat with this data is that one set is from 2005 and the other from 2010. I've already spotted one development on Morton Street that was grassland five years ago. Another problem is that the land use data was semi-automatically obtained using imagery from a surveying company, so it is not precise. It appears there are some spaces marked residential which are not, and vice versa. There are some missing zones, too. These problems seem to be relatively rare, but it is always a good idea to take a closer look at any numbers that seem off.

By this scale, the following neighborhoods include a piece of land that is settled at "urban densities:"

  • Allston
  • Back Bay
  • Bay Village
  • Beacon Hill
  • Brookline Village
  • Bunker Hill, Charlestown
  • Central Square, Cambridge
  • Chinatown/Leather District
  • Cleveland Circle
  • East Cambridge
  • Fenway
  • Kenmore Square
  • Longwood
  • Neighborhood Nine, Cambridge
  • North End
  • South End
  • South Boston near W. Broadway

The North End seems to have dwelling unit densities ranging from 90-180 on this map, which is down from the estimate of 275 in 1960. On the other hand, Roxbury seems to fall pretty neatly into the category of 21-40 units per net acre even after all these years. Looking further, nearly all of the formerly separate towns that were annexed onto Boston still retain their suburban or semi-suburban density levels, the main exception being Brighton.

A lot of this list comes as no surprise. Everyone knows Beacon Hill has lots of homes. Actually, I expected to see more neighborhoods on it, based on "feel" which apparently turned out to be misleading. I thought there would be more density in East Boston, South Boston, Mission Hill, Somerville, parts of Roxbury, and more of Brighton. Especially since Somerville is the most densely populated city in New England (persons/sq mile). But on this map, it is almost entirely semi-suburban or "in-between."

So does this mean anything? Not necessarily. It is just one factor to be considered, among many. But it is potentially useful to get a feel for which parts of Boston have true density, and which parts might lack city diversity and vitality.


  1. Somerville is the densest city in New England in the same manner that Hoboken and Union City are denser than New York. Its residential density is high by the standard of everything other than Boston and Cambridge, its boundaries include less outlying territory like the airport, and it doesn't have big job centers displacing residential land.

    Incidentally, Jacobs' density numbers are one of the weakest parts of her book. There are so many functional urban neighborhoods with density in the in-between range that one wonders how many of her other observations were only really applicable to 1960-era America. Tel Aviv's Old North has 48-64 apartments per residential acre excluding streets (buildings have 4 floors and 2 apartments per floor, sometimes with the first floor not inhabited, and there are 8 lots to an acre), and realistically about 35 or 40 including. It's sufficiently nice and walkable that local urbanists mention it as an exception to Jacobs' general criticism of garden city neighborhoods.

    1. That's true about Somerville, but my general impression from walking about the place was that it is built up like Cambridge is. Or so I thought.

      I agree that her numbers from 1960 were probably too generalized even for that time. 100 d.u. per net acre is too neat, I find it hard to believe. But I think the idea is right. You need a certain number of people to be in a particular area for it to be statistically likely that businesses can survive there. The nice thing about measuring dwelling units is that it largely distinguishes dense from overcrowded areas. If lots of people can afford to have their own place, then it's likely they have enough disposable income to drive nearby retail too. That's not guaranteed by any means, though.

      I've noticed a lot of neighborhoods in Boston and surroundings that can be described as quite residential in character, but only a few that really have an interesting and vibrant street life. I've been curious for a long time why that is the case, and I was hoping that this data might be a partial guide.

  2. This is awesome! I am always so jealous of American access to data like this — I would have loved to have produced similar maps like that for Toronto for my post, but the data just isn't available.

    1. Well, besides the Census, the data seems to be specific to Massachusetts and only because it seems that MassGIS commissioned this land-use study back in 2005. The more I look at it, the less satisfied I become with it, as I notice additional errors. But I don't think I have any alternatives at this time. The Zoning map data is even worse.

      Maybe Toronto or one of the local universities has something similar?

  3. Hi Matthew - this post is really helpful! I am working on some development density documentation for a LEED building, and there is surprisingly little info available, short of actually visiting City Hall and going through the Assessor's maps...
    Any chance this map of yours is available publically on Google? I would love to be able to draw a radius on it around my project site, etc.
    If you don't mind sharing, it would sure be appreciated!

    1. The original map/fusion table, which mashes up census blockgroups with land-use data, used by the article was:

      but since then I've decided to forgo the land-use data in favor of just examining census blocks, which are fine-grained enough to get a good feel, I believe:

      all this is based on the publicly accessible census and American Community Survey data.

  4. Hi Matthew, this is super cool. I was just reading Jacobs and started wondering the same thing. One question: how did you get number of dwellings per census block? Was that in the MassGIS survey?

    I was trying to calculate the same values for Pittsburgh and kept getting very low numbers, I think because I'm just using dwellings per neighborhood / acres per neighborhood, so the streets/etc are diluting the density.

    1. The simplest thing to do is to look at the Census Block data, the smallest unit available. There should be numbers for dwelling units as well as area. It's not as good as having land use, but the blocks might be small enough to make it all come out in the wash.

    2. Yeah, gotcha. I've got tract-level from the census, but can't seem to find block-level. I'll keep digging. Anyway, thanks for making this!

    3. Nevermind, as soon as I posted that I figured out FactFinder. Right, thanks again :)