Saturday, May 26, 2012

Density and Transit: Some Numbers

Note: Updated 5/29 with an expanded list of cities.  This has slightly weakened each of the correlations.

I've posted on density-related issues several times, including the relationship between height and density, building footprints and density, and street width and density, but I haven't said much about density and transit use.  Cap'n Transit has, though, recently writing a series of posts skeptical of the notion that certain areas "don't have the population density to support transit."

I think the Cap'n is right that this statement, as it's phrased, isn't all that helpful in clarifying the issue.  Instead, in all or virtually all cities, the presence of public transit is a matter of political will – in low-density Arlington, TX, the largest American city without public transportation, only political opposition has stood in the way of the establishment of a bus network. Whether the system would be profitable is presumably only one of several factors in the debate.  Below some point of total population or density, we might be confident in saying the economics would forbid any political consideration of public transportation, but that line is not a bright one.  Private mass transit is more directly tied to profitability, but this in turn is affected by political choices, both toward the private operators themselves and to competing forms of transit (which, as Cap'n has pointed out, includes cars).

A more interesting question, to me, is whether we can say anything more broadly about the relationship between density, both of population and jobs, and transit modal share. The chart I've put together below plots the residential density of a group of the largest American cities, measured by urban area, against commuting transit modal share, using data from the 2009 ACS (raw data is further down):

While there are a few transit overachievers, notably Portland, Seattle, Boston and Washington D.C., the correlation is overall very strong.  No city with an overall density of less than 4,000 per square mile, and there are many, has broken a 10 percent commuting modal share.  The most notable outlier, Miami, may possibly be explained by examining its CBD employment density and job concentration in the charts below.
  
Here is the same exercise, but using the employment density of each city's central business district (in thousands per square mile) in place of residential density:


A similar correlation is there, but it is not quite as strong.  Cities can build equally dense downtown employment districts with almost any level of transit share – cf. Dallas (4), Los Angeles (11), Philadelphia (25) and San Francisco (32).  Again, however, those cities with lower employment densities have difficulty attaining higher mode shares (the cutoff point appears to be around 100,000 jobs per square mi.).


Finally, a comparison substituting employment centralization (proportion of all jobs in the metro area located in the CBD) for employment density:


The correlation is by far the weakest here – a dense, polycentric city can achieve reasonably transit high modal share – but highly monocentric cities appear to have an easier time boosting transit use. Low density monocentric cities fare poorly. (Interestingly, Washington DC, much criticized for the economic effects of its height limit, which has allegedly driven jobs outward to edge cities such as Tyson's Corner, is the second most job-centralized of large American cities. High-rise Dallas and Atlanta, by contrast, are heavily decentralized).

The lesson here seems to be that, when it comes to increasing the share of city residents using transit residential density is a key, if not the key, factor.  It may also increase the financial viability of the transit service, although it is possible that mode share and transit expenses are also closely correlated.  That, and some other possible comparisons, will have to wait for another post. 


Employment data from 2000 Demographia studymode share data from 2009 ACS via The Transport Politic.

26 comments:

  1. Seems to me that the problems of "average density" might make the number less useful (ie it depends a lot on where you draw the borders, the presence of parks, etc). I'd like to see you do this with "perceived density" instead, which I think is a much better metric for this purpose.

    ReplyDelete
    Replies
    1. Alai -- I used the urban areas of each city rather than the entire metro area in calculating population density, so the figures I present here are probably closer to Chris' weighted/perceived density figures (at his link below) than the "standard" density figures. Since the political boundaries vary, my method is definitely open to critique, but nonetheless Chris and I appear to have reached basically the same conclusions (which is reassuring).

      Delete
    2. Are you sure those are the urban area densities? I'm pretty sure LA's would be the densest...

      Delete
    3. No, I misspoke. The second post uses urbanized areas. This one uses population density within city boundaries.

      Delete
  2. Here you go: http://austinzoning.typepad.com/austincontrarian/2008/09/the-association-between-density-and-mode-of-commuting.html

    Nice work, Charlie.

    ReplyDelete
  3. I wonder if excluding DC would change the trend line (especially the latter two graphs) and thus the analysis. DC's high transit use has to be at least partially explained by having a major employer (U.S. government) exclusively subsidize that mode of transit for its workers. I cannot imagine that most of those professionals would travel by train/bus if it was not basically free, which would be the case in other cities/professions.

    ReplyDelete
    Replies
    1. I don't buy that, Tim. Would you exclude all of the other cities where free parking for employees is the predominant perk?

      DC has high usage because riding the Metro is an advantage over driving, usually both in terms of cost and time.

      Delete
    2. My hunch is that DC's transit share is so high largely because it built so few freeways into the urban core, combined with the presence of a Metro system designed with the needs of suburban commuters in mind. This makes transit a more time and cost-efficient option, as Alex B. points out. Another test might therefore be checking mode share against freeway and transit miles per capita -- I'll see what I can dig up on that.

      Keep in mind, though, that this is only commuting share. Looking at transit share as a total of all trips might show a considerably different result.

      Delete
    3. Alex B., no I wouldn't exclude areas where free parking is the predominant perk because for this analysis that implies the point I was trying to make that DC's story may not fit the same type of density or location of a CBD as other U.S. cities. I was suggesting that perhaps transit share exceeds what is expected because of the subsidized commute and not the density of other cities. Is this true? I don't know but I think it is interesting question to ponder on trend-lines if done without DC.

      Having lived in Alexandria, VA and worked for the federal government, I know of people that rode Metro only because it was free and many offices did not have available parking (and, yes, I know anecdotes don't make trends, hence a question, not a definitive statement). Don't get me wrong, I loved living in the DC-metro area and the city has done a ton right (CaBi for one), but I am just not convinced the full story is that Metro is cheaper or less time exhaustive than driving. Or at least not convinced that this calculus would be made for many federal employees if it was not for the total subsidy from the federal government.


      Charlie, interesting additional data in the follow up post. Thanks for doing! An additional question (not asking for even more work on your part, just curious) from these charts is what this data says about cities with commuter rail. How are VRE/MARC trips treated in DC, Metro-North trips in New York, or Metra trips in Chicago? I see that they are not included in light rail, but would those trips equate to transit trips above? If not, seems like NYC may be undercounted but I am not sure if there trips are a substantial enough number to change the analysis at all.

      Delete
    4. @Tim: I excluded commuter rail from the sample in the second set of charts. Thus MARC and Metro-North are out. The modal share figures are for residents, rather than for employees, so the match is a fairly good one. As you say I doubt there are very many NYC residents who commute to work via Metro North or LIRR -- in fact Metro North has a policy against use as intracity transit, and this is reflected in their boarding and schedule policies. In Grand Central Terminal, for instance, the list of stops for a given train never even shows Harlem/125th, since it's intended as an "accepting passengers only" stop -- I suspect they also don't want to encourage people to plan GCT-Harlem trips.

      Delete
  4. DC also has congested highways and river crossings. That makes mass transit time competitive and cost competitive. Dense TOD around WMATA stations in northern Va also drive ridership.

    ReplyDelete
    Replies
    1. Congested highways are certainly a major factor. The extremely high rail ridership of Calgary and Edmonton, relative to North American cities of comparable size, is largely due to the fact that no freeways reach downtown in those cities.

      Delete
  5. This is useful data, but to get the whole story you need to get data from other countries as well. Canada has better transit ridership than the US, and Europe and Asia do much better. From what I've read, some cities in the Netherlands have 15 to 20% transit mode share despite being lower density. I think there is still a correlation between transit ridership and density of employment, residences and destinations, but it would be interesting to see how much it changes due to the priority given transit in other places.

    ReplyDelete
  6. Where's Miami? I ask because Miami is one of the most notable underachievers- 10,000 people per square mile (in the old city of Miami) and a transit market share like LA's

    ReplyDelete
    Replies
    1. Miami was inadvertently left off this list since going by city population alone, it is not even on the list of the top 30. It really does belong on there, though, as do Tampa, Orlando, Pittsburgh and a few others -- I will redo these numbers using the longer list of cities in the second post.

      Delete
    2. As a sneak peek, adding Miami plus the other cities decreases the correlation to .80 from .88, with almost all of that change due to Miami alone -- it is an extreme outlier. I'll add the revised numbers soon.

      Delete
  7. Could you control for city age and see what that does to the correlation?

    Most of the cities with reasonable transit penetration were, by and large, built before WWI and the advent of the cheap car. Moreover, they were built at a time when mass transit cost less to build.

    So being old might be as big a deal as being dense (or might not, obviously).

    I realize this would be a tricky thing to do because the advent of the cheap car correlated with (caused?) zoning that basically forbid dense construction, so I think you'd be hard pressed to find any large dense American area that was built after the cheap car.

    ReplyDelete
    Replies
    1. Well, high density certainly correlates with age, with the striking exception of Miami -- which may be strong evidence in favor of the argument you are suggesting (the one very dense car-era city has very low transit modal share).

      On the other hand, many of the major transit systems in old cities were not legacies of the pre-car era. The systems in Baltimore, DC and San Francisco are all quite recent, roughly the same vintage as Miami's own Metrorail. If density was what permitted them, why did Miami's not lead to higher levels of ridership? Is the form of the density, rather than simply the level of the density, what is most important?

      There's also the issue of how to measure age -- cities are built out slowly over time, so how do we measure how "old" Cleveland is? One possibility would be using 1910 population (just post Model T) as a proportion of peak population, to get a sense of what amount of the city consists of pre-car infrastructure. Or 1880 population, if you want to look at pre-electric streetcar. That could certainly be done. So many possible charts, so little time...

      Delete
    2. I wonder if you could also regress against parking, both/either existing available and minimum parking requirements. Perhaps age of city is somewhat of a proxy for this.

      Cities such as Los Angeles, Houston and Atlanta achieve fairly high densities but undermine transit use by requiring excessive amounts of parking. There was an infamous article in the LA Times a couple of years ago (sorry don't have link) about how transit wasn't working in LA. It provided an example of a doctor who lived in a TOD (literally across the street from a metro station) who commuted by car to the hospital he worked at that was above a metro station. The article failed to mention that the TOD provided two covered parking spaces for each condo (as required by the city) and the hospital had loads of parking. Obviously providing reduced or no parking would have self-selected for transit users, rather than rich doctors who just thought the location was cool. I think this exemplifies LA's half-hearted embrace of transit.

      Delete
  8. There's an interesting article in the recent University of California Transportation Center's Access Magazine on the relationship between transit and density. Here's the link: http://www.uctc.net/access/40/access40_transitanddensity.shtml

    ReplyDelete
  9. I'm still understanding from you, but I'm making an attempt to achieve my ambitions. I undoubtedly enjoy studying every thing that is posted on your internet site.Keep the stories coming. I preferred it!.
    airport transfer service in atlanta
    airport service

    ReplyDelete
  10. Hello, I've been looking through the weblog a pair of times and basically needed to say that the way with phrases is best notch. All the best.
    Reliable Car Service in Putnam County,NY
    Car Service In Putnam County,NY

    ReplyDelete
  11. If you maintain publishing well prepared posts just like this then I will always preserve returning back to your website. Truly great content.

    Arlington Cabs Companies
    Taxi Services in Arlington TX
    Transportation Services Arlington TX

    ReplyDelete
  12. Great Article it its really informative and innovative keep us posted with new updates. its was really valuable. thanks a lot. movers of ottawa

    ReplyDelete