In Chad Newton’s provocative stop spacing post last week, I decided to omit his graph of ridership per mile vs. center city density.  The comments made it clear that that was a mistake.  So here it is:

I thought it was pretty self-evident that dense cities support transit use more, but share whatever you learn from this in the comments.

41 Replies to “Chart of the Day”

  1. The problem with this is that it just uses Central City population density, even though PATH mostly serves people from New Jersey and Staten Island Railway just serves Staten Island. BART almost completely serves the suburbs of San Francisco, and Muni is a lower-capacity street railway.

    1. Alex,

      There is no perfect way to characterize population and population density for this kind of exercise, and certainly none that can be done without dissertation-grade levels of research.

  2. Canadian cities do so much better than the US cities, except for Vancouver. I guess the skytrain’s ridership doesn’t refelct the opening of the new line (Canada line) though. Our light rail has only about 1,200 per mile ridership now, so it is still not in a good shape. Just one more comment. I think the graph should combine different modes of rails within one city. For instance, BART and SF light rail is interconnected and a lot of riders use both, although they are operated by different agencies.

  3. this refers to residential population density when the ridership is largely driven by employment density in center cities, right?

  4. The relationship may be easier to see as a scatter plot (population density vs weekly ridership). Could you post the data, so that readers could play around with how to visualize it?

  5. I’ve learned that Seattle has done one of the worst jobs on the continent with urban rail, even with advantages in density.

    The three least dense US cities on the chart are Phoenix, Charlotte, and Salt Lake City. Salt Lake City and Charlotte have about TWICE the ridership per mile, and Phoenix has about 50% more ridership per mile than Seattle.

    We are building one new line and extending our current line in the next 15-20 years. SLC, for example, is building 4 new light rail lines in the next 5 years.

    1. I am pretty sure this has been said on here many times, but a single metric, even ridership per mile, is not the end-all-be-all of a light rail transit line. Building 4 additional lines in 5 years is probably a more telling fact about the relative success of the Salt Lake City line than ridership per mile.

      1. Yes _i’m more interested in the huge variation around the trend line here than in the general rule that more density equals more ridership. It does, sort of, except… that some places with ten or even twenty time the density on this metric have almost the same ridership…

      2. Not to mention that there is NO tunneling involved in SLC’s extensions, as has been mentioned here countless times. Topography plays a role in how fast systems can be built out.

    2. Is SLC’s system projected to have 280,000 riders in 2030? Considering the line hasn’t even been open a year yet I find it just a tad premature to try and draw conclusions on its success already.

  6. Wow, it pays to be Canadian. Do Canadian policies favor mass transit more than American policies do?

  7. Canadian cities are a little less sprawled than their US counterparts, plus we have less wealth disparity. Conversely we fund our transit systems less systematically. No idea if that accounts for the differences.

    However, what would be more interesting would be population density around stations. vs ridership.

    1. I’m not sure about those explanations. New York has a huge rich-poor gap. Funding in Canadian is spotty, that should hurt mass transit This graph should account for sprawl. What I’m always struck by in Canadian cities is the lack of freeways. They just really didn’t design their freeways in Canada as an intra-city transportation system. I’m willing to bet that that has made a really big difference in urban development between the two countries.

      1. Are you basing this statement entirely on Vancouver? Because Toronto is sprawling – very much so. And there are several freeways in Toronto.

      2. Even metro Vancouver sprawls like mad. It has dozens of dense clusters, mostly downtown, near Broadway, near the SeaBus, or adjacent to SkyTrain stations, but the Fraser Valley as a whole is still sprawltown. Even in Vancouver proper, I’d bet that 70% is single-family homes (or single-family homes that have been subdivided into apartments).

        That said, it is quite nice that the Trans-Canada Highway stays east of town and Hwy 99 abruptly changes from freeway to four-lane arterial as soon as you cross into the city limits.

    2. Vancouver is more willing to build satellite cities connected by rapid transit. Note the highrise buildings in New Westminster, and the new downtown in Whalley. The closest equivalent here is Bellevue, which got tall office buildings in the 90s, highrise residences in 2005, and still doesn’t have rapid transit. The other suburban centers in Pugetopolis have none of these, though many have made smaller advances toward dense cores.

      1. I agree; R makes some nice old-school graphs,, and mathematics or Apple Numbers make good flashy graphs. Excel 2003 mostly just makes crappy graphs that look like they were made in excel. I’m in grad school at nyu and there are profs around that would squeal out loud if I presented my work on a graph like that.

      2. Oh. If the issue is the program, I’m going to use what’s convenient and handy, which is what’s here on my work computer and my computer at home.

        I imagine the same goes for Chad, who made this graph.

      3. I imagine part of the complaint is that that type of graphic shouldn’t be save as a jpeg — jpeg compression doesn’t suit graphics that are line art/text/color blocks. Should use png or gif for that — you will get smaller size with better quality.

        To get the equivalent small size for that sort of graphic with jpeg, you would have to compress it down enough to make the text all blotchy. To get the equivalent quality with jpeg, the file size is much bigger. This one’s not that bad, though, I think.

        It’s the other way around for photos. For those, jpeg is better.

      4. Posting the excel file would be useful so those who want to look at the data in different ways can do so.

  8. To relate this to upcoming questions for Seattle transit planning: Does rail format (grade-separated light rail, shared right of way light rail vs subway vs elevated) have a positive or negative correlation to ridership per mile? In other words, seems like population density is the primary factor in ridership per mile, how much can choices in design and planning affect that?

    My intuition, backed by no data, is that subway has positive correlation, while shared right of way and elevated have negative correlation.

    Also, where does Tacoma Link fall in the graph, or is it included in Sound Transit numbers?

  9. Whoops, still working on my posting skills for this blog.
    The book is:

    The Visual Display of Quantitative Information
    by Edward R. Tufte

  10. To Misha’s comments above about Phoenix having very high ridership numbers but a sprawling population and how to make sense of that; this is a question that is often on the minds of Phoenicians and those at ASU’s Global School of Sustainability.

    In a recent class I took there for grad work I learned something that even I missed as a resident of Phoenix for many years. Of the 517 sq miles that account for the City of Phoenix land mass, about 200 sq miles are comprised of park and preserve land that has no development whatsoever and limited paved roads.

    The city is on track to purchase more land and within Phoenix there is over 300 sq miles of land that has no development and is prime for preservation. In reality, the “urban footprint” of Phoenix has a population density similar to Seattle’s (7,313ppsm)and is remarkably different compared to the total “land mass” density (3,200ppsm).

    This might help explain the discrepancy when comparing Valley Metro LINK and where much of the population lives in both cities. With the extension of both the cities’ lines (Seattle and Phoenix) in the future, both will reach even more densely populated regions and ridership for both will grow astronomically (especially for Seattle). That means that in the future I believe Seattle will have similarly high ridership and be comparable to Phoenix.

    1. The big difference… Phoenix serves the ASU campus. When Link reaches the UW it will have no problem matching or exceeding Valley Metro.

      1. Only problem with that assessment is that even when ASU was out of session and numbers for Valley Metro in July were expected to drop WAY below the 26,000 weekday riders because of ASU and the heat (above 107° on average), this did not occur…The ridership then started to rebound weeks before school started in August and even in Nov/Dec with no school have exceeded 40,000 daily boardings. When VM extends the Metro Center and NE lines, ridership is expected to multiply.

        I expect similar things to happen for LINK, but not to the same extent.

      2. But the University is not just students. In fact students account for a tiny proportion of the ridership because most live within walking distance to classes. Football. Research. The restaurants and nightlife that surround a University. Just the demographic of people that tend to live close to such an institution. That’s the core ridership. I expect the effect to be even greater for Link because the UW Medical Center is such a big draw and the neighborhoods around UW are so much older and more established. UW to downtown is a LONG established transit market and the geography in Seattle makes the light rail line even more compelling. Instead of trying to be a driver of development UW to downtown serves an existing need. Oh, and a stop on the very fashionable Capital Hill won’t hurt either.

      3. Very true…but then I don’t understand you first comparison since you discredited it yourself in your second comment.

        Also, I wonder about the “riders per mile” of this graph when most national graphs, including Misha’s wiki link, show that Seattle LINK has an average of 1,068 riders/mile and the Seattle Street car averages 1,308 riders/mile but the Phoenix system shows 1,675 riders/mile. I’ve never been a good interpreter of this specific type of chart so that is my folly if I am missing something.

        Bernie, I am interested in knowing my the medical center would be such a “huge draw?” Is this the case in many urban settings with a medical center? Will ridership skyrocket by tens of thousands because of this connection? If so, is it possible that the new ASU/UofA Medical & Research Center downtown will have the same effect in a couple of years? How does that work for transit in Seattle?

      4. I’m confused by the first assertion, “you discredited it yourself in your second comment.” I thought I was explaining why the ridership didn’t plummet during winter break and summer.

        I don’t know how much research money ASU brings in. I’m sure it is substantial and a new facility will indeed help Phoenix and the ridership figures will go up. UW brings in about a billion dollars a year in grants and 1/2 of that is medical research. It’s one of the leading medical institutions in the country (consistantly top 10, often compared with Stanford, Johns Hopkins, U Cal, U Mich, etc.). That hasn’t been an area where ASU has focused.

      5. With the UW a fairly high percentage of the students live more than 1/2 mile from campus. Furthermore many who live in the University District aren’t students and don’t work for the UW.

    2. Does Phoenix have any census tracts (Besides the ASU campus(es) over 40,000 people per square mile? Seattle has two besides the U-District. I would assume the density of Phoenix is much more consistent than Seattle, where you go from sprawling single family type dwellings to the much much denser parts around downtown.

      1. Yes, this is generally the case; there are census tracts in the 85016, 85013, etc and more midtown zip codes with densities similar to Tempe, 48,000+/sq mile but only a few more. Downtown Phoenix is 1.5 square miles but only has about 14,000-16,000 residents as of 2009 estimates. By 2020 they expect 20,000 more and by 2030 40,000-60,000+ in the 1.5 sq miles downtown.

        I think that the medical research center thing for Seattle is a cool aspect. ASU does not have medical research in terms of hospital services, but has one of the nation’s largest Bio-design Institutes of Research. The UofA is the state’s largest medical research facility and they are opening up a downtown Phoenix campus in association with ASU instead of stretching the Tucson facility.

        So is it fair to say that a popular and well ranked medical facility lures the crowds? I think more cities would benefit from more connections via street car rail similar to Seattle’s. We have two top ten nationally ranked medical centers but one is 7 blocks from a station (not too far in Midtown) and the other over 12 blocks (north of Downtown) and only with a slow bus connection.

        When the Seattle extension to the UW is completed, I will be there for opening day! This is long overdue…

  11. I have no link to back this up, since I only remember reading it in a history of the NYC Subways.

    What we are now discussing is the same issue that confronted the early builders of that system, and that was the nature of the commuter. Short haul vs. Long haul, or Local vs. Express.

    On various routes, the ridership increased when they built express tracks alongside the local tracks.

    For smaller cities, it’s an issue of when. When the local ridership saturates the system, and drives away the longer distance rider, when do you build the infrastructure to handle those riders, along with making the system enticing for future long distance riders?

    On 2 track systems there is a way to accomplish this: Skip-Stop service.
    What scares planners is the passengers. Most people generally assume all the trains will stop at a station, and would likely walk in front of a train assuming it was stopping at the station they are heading towards.

    Granted, you might only save the 20 second dwell time, but the perception of a fast ‘express’ service is enticing to those longer distance commuters.

    Rail systems are both about building for the current conditions, but knowing how you build the service has much to do with how it (and the community) grows.


    1. Local-express service has not really been tried anywhere except NYC, so it’s premature to say smaller cities wouldn’t have increased ridership. People in NYC routinely walk across the platform from an express train to a local and vice-versa, often without waiting. So they can take the express except for the very end portions of their trip.

      1. Sometimes it’s an upstairs/downstairs configuration.

        However, my point was that it’s a greater infrastructure investment to put greater than 2 track systems in place, so that’s why skip/stop service would be the other alternative.

        It would be necessary to figure out how to mitigate the safety aspects.


  12. Density is not the only success factor for transit ridership, nor might it be the most important one, as is evidenced by the serious anomalies shown in this graph for most Canadian cities, especially Toronto, Montreal, and Calgary.

    Quality of ridership is probably a more important determining factor, as measured by door-to-door travel time, passenger safety, proximity of the start and end of the trip to transit stops, frequency, convenience, reliability, seat availability, capacity, relative immunity to congestion, waits for transfers, exposure to weather, and station spacing, among other factors.

    One of the best transit systems in the nation was omitted from this graph, which would have showed another serious anomaly in the graph’s otherwise smooth and predictable curve: Morgantown, West Virginia.

    Morgantown has a population density of just 3025 people per square mile according to wikipedia, placing it near the bottom of the cities listed in terms of density. However, its rapid transit system, the Morgantown PRT with its 16,000 riders per day and a length of 3.6 miles between first and last stations would give it one of the highest per-mile riderships in the nation, behind only Boston and Houston if it were classified as light rail.

    The Morgantown PRT, of course, features extremely short wait times, high average speeds, and an unexcelled safety record for its 30+ years of operation, so its high ridership should not come as a surprise.,_West_Virginia

  13. Wow, wow, wow.

    There are some wild differences there.

    (1) NYC Transit obviously does great. There are so many reasons for this it’s hardly worth listing them.
    (2) Toronto and Montreal do *spectacularly*, especially considering that Toronto has only the two subway lines. The general Canadian outperformance vs. the US is really, really interesting. Maybe they will get high-speed rail before us. ;-)
    (3) The main lesson I get from these are, *send your transit lines to the right places* and make them *work*. San Francisco is infamous for missing the densest corridor and many of the densest parts of town and it does poorly. San Jose is famous for poor design. Cleveland and Pittsburgh are scraping by with leftover lines from 1900s-era property development schemes.
    (4) As others have noted, the cities with severe overperformance relative to the curve seem to be places where car usage is particularly hard (as well as having transit which gets where they want to go).

Comments are closed.