It’s now pretty clear that transit ridership is falling in many US cities. Why?
I don’t know. (Don’t trust any pundit who never says this.)
But journalists are asking me this and I need an answer. Laura Bliss’s recent piece in Citylab really captures the problem. It’s a smart read, but in short: Bliss interviewed a bunch of experts on this, including me, and she got lots of smart speculation, mostly grounded in anecdotes. (“Pick a Culprit” was her sub-headline).
Everyone seems to agree on the same long list of culprits.
- Ridehailing services like Lyft and Uber, especially to the extent that this industry may be undermining transit through unsustainable predatory pricing.
- Stagnating or declining transit service. Even transit agencies that are not shrinking are mostly declining in service/capita, as the population grows but they don’t have the resources to keep up.
- Cheap driving. Previous studies about the impact of cheap gas thought this relationship was mild, but those are less useful now, because gas is so cheap that we are off the scale of those studies’ analysis.
- Fares static or rising as other options get cheaper. To be clear: I’ve seen no cases where cutting fares triggered so much ridership that the agency broke even. Transit agencies have very little room to more financially here. But there may be correlations. (Always check transfer penalties, too; they often matter more than base fare.)
- Crisis situations in certain agencies. Lots of transit agencies are in financial trouble, which creates trouble of all other kinds. The travails of Washington DC’s subway get all the press, maybe because national journalists and policymakers experience it personally. But many transit agencies are facing crises — especially deferred maintenance in older transit agencies. And no, not all transit agencies are victims. I see a lot of obsolete management and planning habits, in some agencies, that hold transit down.
- Some shifts from transit to other non-single-occupant-car modes, which can be OK. These may include ridesharing, improved cycling infrastructure, greater urban density (which is putting more trips within walking distance) and better pedestrian amenities.
And I would add a couple of others to the list.
- Bad data. Do we even know how bad the problem is? A few weeks back TransitCenter published a table purporting to compare 2015 and 2016 ridership at many US metros, showing drops in many agencies. But most transit agencies I talked to said the table was wrong, and instead admitted to problems in their own reporting and analysis. Transit data is often a mess — as I’ll discuss in another post — though it’s improving fast. Still, almost every data element is prone to methodological problems.
- Noisy data. Transit ridership is so volatile that it takes time to see long trends. I’d conclude nothing from a one year drop; it’s only because we’re now seeing multi-year drops that I’m deciding this is real. That makes me very late to the party but it’s the only way I can know I’m not chasing phantoms. And it’s a huge pitfall for transportation journalists, whose deadlines require them to write stories before we can really know.
The problem is, we really don’t know the relative importance of these things, and neither does anyone else who’s speculating in the media.
Bottom line: We need research! Not the sort of formally peer reviewed research that will take a year to publish, but faster work by real transportation scholars that can report preliminary results in time to guide action. I am not a transportation researcher, but there are plenty of them out there, and this is our moment of need. Here are my research questions:
- Which global causes seem to matter? Straight regression analysis, once you get data you believe. Probably the study will need to start with a small dataset of transit agencies, so that there’s time to talk with each agency and understand their unique data issues.
- What’s happening to the quantity of transit? If ridership is falling because service is falling, this isn’t a surprise. If ridership is falling because service is getting slower — which means lower frequency and speed at the same cost — well, that wouldn’t be surprising either.
- How does the decline correlate to types of service? Is this fall happening in dense areas or just in car-based suburbs? Is it happening on routes that are designed for high ridership, or only on those that are designed for coverage purposes (services retained because three sympathetic people need them rather than because the bus will be full). Is it correlated to frequency or span changes? Heads up, local geeks! A lot could be done looking at data for your own transit agency — route by route and even (where available) stop by stop, to analyze where in your metro the fall is really occurring.
One more note: It’s easy to analyse this “bus vs rail,” because that’s how the National Transit Database is structured, but nobody knows if that’s the real distinction that matters. As Laura Bliss’s piece notes, rail ridership and bus ridership are not trending any particular way relative to each other — a good hint that this is just the wrong place to look for an explanation.
I don’t pretend this is easy, but it’s needed. Scholars! Come to our rescue!
My guess is that research question #3 will be most productive.
Regression is very tricky because few people change their commuting habits immediately in response to changes in cost & convenience. There is a range of time delays from zero to years and even decades (as housing turns over near transit lines).
Not data, just a wild ass guess… I see:
1) many more people riding bikes these days compared to 5-10 years ago, though surely this is a small slice of people who get around by means other than a personal car. I don’t know if it’s statistically significant.
2) transit is too slow (time-in-transit) compounded with agencies moving like icebergs to implement faster service either through BRT or by simply just removing bus stops that are too close together (bus spends valuable time stop-and-go). People are so busy working multiple jobs to pay for student loans, healthcare, rising cost of rent, etc that they often need a car simply to move around fast enough to all the places they need to go. Most transit in the US cannot meet their speed needs.
3) similarly — mobility options often being A) take transit for $4 and arrive in 30 min or B) take Uber for $7 and arrive in 9 minutes — most people would choose B, especially if you’re with another person to split the cost then Uber is cheaper.
Although it’s odd that Uber is still not profitable, despite taking a 20% cut on millions of fares. It seems impossible to me that they aren’t massively profitable since they are just providing the technical backend, and drivers assume the risk. Who knows where the money goes.
So I don’t see Uber surviving once the VC money runs out. Many drivers already hate them due to the low pay. Maybe the trend will reverse when Uber is gone?
Uber is subsidizing the drivers with special bonuses, since the fares alone are well below a living wage.
This might be of interest to you, a presentation on some of the research that is currently being done at the UCLA Institute of Transportation Studies for the State Transportation Agency and SCAG on why there is declining public transit trends:
Which of these agencies have open lines of communication with riders, so riders can give feedback through a simple web form or a text message? Ours doesn’t, so they haven’t heard what I’d tell them, about why I don’t ride it.
Call their customer service line or general office line. Find their employees on LinkedIn or the transit agency’s own website, and figure out who to talk to about your specific issue. Go to their operational and/or board meetings. Both should be open to the public. Talk to your city councilor. Connect with local transit advocacy groups. Remember to be persistent but respectful. There are many ways to make your voice heard.
(A “We want to hear from you!” link that goes to a structured survey where they selected the questions being asked, does not count.)
In the Chicago region we are seeing population shifts in which people move away from areas of high poverty and high transit usage. There is some national research showing a shift in poverty towards the suburbs – certainly that isn’t conducive for transit.
Michael, that’s an excellent point: many transit-conducive neighborhoods are either (1) depopulating due to continued crime and disorder (parts of Chicago, Baltimore, St. Louis, etc.) and dispersing to areas with little or no transit or (2) gentrifying with people (at far lower densities; the “unslumming” process Jacobs described) who don’t use transit for the reasons Alon described:
Almost all US transit outside New York is pretty bad: what if the first decade of the 2000s was actually the anomaly, and now we’re merely returning to the slow and steady decline that’s ground on since the 1960s? Jobs and people continue to disperse, so most transit networks – let alone the even smaller portions of those networks that comprise the FTN – access an ever-declining proportion of diluting metro destinations and people.
This decline is easily hidden in the metrics, which is why I think most metrics are GIGO: I’m amazed at how crudely some metrics are conducted to “show” fantastic transit access, in which crude “buffers”, for example, are generated around peak-only routes to create the statistical impression of lots of service to/in low-density areas: “We reach 88% of the region’s population!” In reality, the area within those buffers is almost always unwalkable, which means the true “buffer” radius is laughably small, plus the transit extending into those areas doesn’t actually exist for large periods of the day.
“Look how many more people we’re reaching! How can our transit use be declining?!” Ah, but those huge areas that you serve, can people get on that transit without checking a schedule? Is it within comfortable walking distance or are people being flattened jaywalking across “boulevards?” Is the route reasonably fast and direct, or does it meander for more coverage, ironically making it less useful? Once those qualitative analyses are added to the metrics, then they no longer look so great. That’s when we always discover that the FTN – the only part of any system that is actually useful – has remained unchanged or even shrunk over the years even as the population has diluted away from that stagnant FTN. “But the system map looks so great: we get into every corner of the region!”
Weren’t modest gains in transit ridership seen in the recessions of the 70s and 90s? Since the 2005-2010 economic stagnation was far worse than previous slumps, maybe that’s why the corresponding bump in transit ridership was larger? Did we generally see ridership stagnation outside the periods of economic stagnation?
“Death Spiral” caused by congestion?
I know City Observatory had an old article tracking average bus speeds in US metros for (IIRC) only a decade or so. It’d be interesting to see if the declining average speeds they observed could be extended back to see if the trend has been underway since the autos-vs-streetcars era. If so, then that makes me wonder if we’re still – apart from the occasional anomalous period – in a half-century “death spiral” of ever-declining transit vehicle speeds caused by ever-increasing congestion. Declining speed, more expensive to provide the same amount of service, your budget is stagnant so you cut service to cut costs, which makes the system less useful, which increases congestion, and on till death.
Bad data goes both ways! Using unlinked trips instead of linked trips can create fictitious extra ridership as a bus agency transitions to a transfer-based grid; this is probably what happened in Houston – if I remember correctly, linked trips are flat, unlinked trips are up 2%.
A few more points:
– In New York, the impact of Uber and other TNCs seems to be negligible. The key here is that in New York, subway ridership fell for the first time in years, while bus ridership fell, as it had in previous years, but at a lower rate. TNCs induce car traffic, so they’d wreck bus ridership more than rail ridership.
– In cities with bad transit, i.e. all of the US outside New York, there’s a secular trend toward more car ownership as poor people get cars and leave the city. Additional transit investment is required just to stop this trend. This is what Portland has achieved: its metro area mode share is the same now as in 1980, but that of peer cities dropped. Even metro areas with decent transit have at best flat ridership when there’s no growth in service: Vancouver’s ridership rose after the Canada Line opened, but after 2012 it stopped growing.
– Low-intensity crises. Washington, DC is in freefall, and since US transit usage is dominated by a few cities, DC’s crash is enough to create a national trend. But New York has had problems with lower average speeds and greater delays, just not at the level of DC.
Researchers: feel free to contact me at [email protected]
I work for GoTriangle, the regional transit agency in the Triangle region of North Carolina (Raleigh, Durham, Chapel Hill) and we also manage GoDurham, the municipal system in Durham. Our region would be a good one to study for a number of reasons:
1. We have no rail, so no distractions there.
2. Our region is one of the fastest growing in the country, so plenty to analyze in terms of population and job growth and how that relates to ridership trends.
3. We have different types of systems to study: university-focused ones, local systems that have a higher low-income ridership, and a regional system that is more commuter oriented.
4. We have been able to add some bus service in two of our counties (Durham, where Duke is, and Orange, where Chapel Hill/UNC is), but not in the third and largest (Wake, where Raleigh/NC State is) because Durham and Orange passed transit sales tax referenda in 2011 and 2012. Wake just passed one last fall, but no new service can be added until this summer. Allows for some interesting comparisons. This is mirrored by parking prices, which are much higher in the higher-density areas of Durham and Orange Counties than in Wake County.
5. We have plenty of data and our data sources are pretty well aligned across all major agencies in the region.
6. We are seeing similar ridership drops as the rest of the country, but from what I have seen at the regional/local level and gathered from the other agencies, it hasn’t hit each agency the same way.
A couple of observations we made for the GoTriangle (regional) and GoDurham (local) systems in recent annual reports that may go beyond the usual macro trends of low gas prices and higher employment:
1. Traffic congestion has truly hit the Triangle in the past few years. It was always a matter of “when” and when is now. With no dedicated transit lanes, this has created some difficult tradeoffs for the agencies we manage. Either cut something from the route (which lowers ridership), add more time to the route (which takes dollars from adding frequency or span), or lower frequency (which lowers ridership). Even where we have had some funding to add service, it sucks to have to add it to maintain previous frequencies, but that’s what we’re having to do in some circumstances and there will be no fixed-guideway transit in our region for many years to come.
2. For GoDurham, ridership dropped on weekdays and Saturdays, but increased on Sundays. Of note, the demographics of those who ride on Sundays skew lower income, with less auto availability. We were able to extend Sunday service to 9pm (a pretty big deal in these parts) and saw a nearly one-to-one ridership-to-hours added ratio after only a few months! So Sunday service is doing just fine and there was clearly some latent demand. This supports a theory that those on the margins between taking the bus and driving a car switched to the car, while those who still cannot afford a car are using transit just as much, if not more.
3. For GoTriangle, the stark difference in ridership trends was around destination. There are three big destinations in our region for transit – Chapel Hill (and UNC), downtown Durham (and Duke), and downtown Raleigh (and NC State). For trips going to/from Chapel Hill or Durham, ridership has remained flat or even increased slightly. Remember that those are in the two counties that had successful referenda back in 2011/12, so we’ve been able to add some bus service to those destinations. They are also much more parking constrained. Trips going to Raleigh saw big drops in ridership across nearly every route. Until this summer, we were not able to improve service to Raleigh and parking is much cheaper than at the other major transit destinations, particularly at the largest employer in downtown Raleigh – the State Government (insert your favorite joke about Republicans being against “social engineering” here).
My personal experience is that the Saturday/Sunday switch also happened in Paris, France, the last 20 years(but especially in the early 2000s). It’s especially striking on sundays. Trains used to be small and empty, they are now big and full, with similar frequencies. And the population is very different from the weekdays. People seen in the train on weekdays take the car on weekends because there is less congestion – and suddenly car goes quicker than the suburban train. People who take the train on sundays don’t have cars at all.
In other words, people prefer the car when they can afford, or when transit is not better. Which also fits my experience in Montpellier, France. The centertown is very attractive to tramway because it’s nearly impossible to park there. Other destinations are flooded with cars.
I live in Seattle, where transit share is up. Most of these factors exist here, such as rideshare services and increased cycling facilities, cheap gas, etc.
So what’s different? We have been investing in transit and our rail network is new and so doesn’t have the maintenance and upgrade costs of new systems. There also might be more unrealized potential from the previous decades of underinvestment, so that there’s more low-lying fruit vs. the declining returns of cities with older systems. We also have a huge population boom (almost all of which has been captured as new riders).
There could be other factors, but a study of what if working here vs. elsewhere could be instructive.
My guess is that at least half of the variation could be explained by gas prices and economic factors (transit does best when the economy is good enough – people have jobs and need to go to work – but not so well that the very poor buy cars).
One aspect unmentioned so far is immigration trends. Immigration numbers, both legal and illegal, have been dropping for a while (this was true pre-Trump); as local transit tends to be heavily used by immigrants, when fewer of them are around more seats run empty. Additionally, some states are now giving away drivers licenses to illegal immigrants, which depresses transit ridership as they can now drive ‘legally.’
Is your city investing in public transport infrastructure and service?
Is your city investing in driving infrastructure?
I’m willing to bet money that this is the answer to the falling ridership question. Here in Auckland we are investing lightly in transit and already seeing 20% year on year rail growth and 5%-10% year on year total transit growth. Transit is becoming an attractive choice, not just a last resort of those who can’t drive.
A category of data that may be useful is looking at car ownership/vehicle availability patterns. The American Community Survey has data that can be mapped down to the census tract level, but this requires using the five-year estimates. ACS data indicate that Americans let go of a lot of cars during the recession and rebuilding period to live car-free and car-lite. One-year estimates indicate that this trend dramatically reversed starting in 2014 for households living car-lite and in 2015 for households living car-free .
Just read that Trump reduced spend on public transit. We should be increasing public transit spend to slow down global warming.
An international perspective here. Some but not all NZ cities are experiencing falling PT patronage. Auckland is up quite a lot (total patronage rising at 6-7% per annum), Wellington is up a little (~3% per annum), but Christchurch is down and most of the smaller towns are down as well over the last couple of years. Some data here: http://transportblog.co.nz/tag/patronage/
This is happening against a backdrop of rapid population growth, rising traffic and congestion, and lower petrol prices. The differentiating factors seem to be:
* Rapid transit availability / investment – Auckland has recently invested to significantly improve its rapid transit system, while Wellington has a long-established (but fairly static) regional rail system. A majority of Auckland’s patronage growth has been on rail or busway.
* Geographic constraints – Auckland and Wellington have geographies that funnel a lot of travel demand through pinch-points, which experience concentrated traffic congestion as a result. Where rapid transit or bus lanes are available, this is an incentive to shift modes. Christchurch is a flat city on a plain.
* Land use disruptions – Christchurch is still dealing with the impacts of a major earthquake that hollowed out the city centre.
The big Australian cities seem to be on the same trend as Auckland and Wellington: https://chartingtransport.com/2017/01/28/update-on-trends-in-australian-transport/
So if you’re looking into the US experience, I’d start by investigating some cities where PT patronage *hasn’t* declined, and see how they differ from the rest. That might give some preliminary indication of what factors matter, and what factors don’t, and motivate a more in-depth analysis.
Lastly, rather than calling for pro-bono / ad hoc research, I think that it would be worthwhile trying to organise some funding for a proper study, even if it had to be done in a tight timeframe.
One notable response by Prof David Levinson at the University of Minnesota:
The TTC is struggling to explain a plateau, or even a drop, in ridership as well lately:
I would be interested to see if demographics are playing a role — in particular the effects of millennials — and how the decline looks in a longer-term (backwards) perspective. From memory, the past few years saw gains for transit ridership and drops for auto-related indicators. Much of the transit growth / auto-related drops were attributed to millennials backfilling travel demand left by retiring boomers, and the speculation of millennials having different preferences and aspirations from boomers — returning to the city, living a car-free or car-light lifestyle, etc. It may be that millennials’ travel choices are evolving with time (e.g., growing discretionary income with increased job experience; starting families — keep in mind that both of those life cycle events generally seem to have occurred later for millennials than for previous generations).
The fact that ridership decreases have been experienced across many agencies with very different contexts (ridership levels and modal split; different fare and service levels and structures; large and small systems; different parts of the country (and even U.S. vs Canada); cities experiencing different economic conditions) suggests to me that it is a fairly large-scale, universal factor at play. Demographics is certainly one. The only other one I can think of is the relative price of gas.
I look forward to seeing a post on bad transit data.
One of the limitations of the National Transit Database (NTD) data is how it handles “subcontracted service”. For example, in the Seattle area, Sound Transit (“Central Puget Sound Regional Transit Authority”) reports its commuter bus service (“CB”) to the NTD as directly operated (“DO”) but this service is then sub-contracted to three agencies: Pierce Transit, King County Metro Transit and Community Transit (who then sub-contracts its Sound Transit commuter bus service to First Transit). Therefore, the total bus service operated by King County Metro Transit, for example, is underreported in the NTD data yet the Sound Transit commuter bus service comprises the majority of the service operated by one of King County Metro’s seven bus bases (East Base).
This concern also applies to rail although you can figure out the NTD data in the Puget Sound region if you understand who owns and who operates what system and the type of service, i.e. commuter rail (“CR), light rail (“LR”) or streetcar rail (SC”).
One of the huge issues that goes on in the USA is the ongoing sprawl that is inherently difficult to serve with anything other than driving. Jarrett Walker lives in the Portland, Oregon area. Take a look at places such as Happy Valley, the north end of Washougal and Camas extending all the way to Battle Ground, and people commuting all the way from Newberg, Woodburn, and Salem.
If you add 10,000 people to a city of 10,000 people using this geometry, then things are still all pretty close together. Add another, 10,000 to that and the distances between everything is further if all the growth is kept to the fringes, as happens in most cities in the USA.
A few of these areas might have mildly tolerable transit connections for peak hours, but for the vast majority of trips there is not going to be a good transit alternative.
Another factor is on-line shopping, affecting both shoppers’ travel and staff travel. Some of this is replaced with other trips, but regional warehouse mailing centers usually are located in areas with little consideration for walking to or from bus stops.
I live in Windsor, ON, and for our transit system, we have the highest ridership since 1992. The largest reason was that the U-Pass was just signed last year, and because of that, many more students are packing the bus. ON. The ridership per capita has been the higher than 1995-2004. Link: http://www.cbc.ca/news/canada/windsor/windsor-transit-ridership-1.3996443
I honestly think though, we should also be asking why was it increasing in the first place.
Regarding the proposed/suggested research: could agent-based modeling be a suitable approach?
Here’s a webinar on the topic of ridership and productivity trends, coming up on Thursday, March 30:
Webcast March 30: Public Transit Ridership and Productivity Trends – What We Do and Don’t Know
Please join us for a FREE CUTR webcast Public Transit Ridership and Productivity Trends: What We Do and Don’t Know on Thursday, March 30, 2017 from 12 to 1pm (ET). This webcast is hosted by the Center for Urban Transportation Research at the University of South Florida.
The link to join the webcast and the toll-free phone number are below. Pre-registration is not required. For those unable to attend, the webcast will be recorded and available at http://www.cutr.usf.edu.
Visit the CUTR website to add the webinar to your calendar.
About the Webcast:
This webinar will share information on national public transit trends in ridership and productivity and discuss causes and implications of current trends.
About the Presenter: Steve Polzin, Ph.D., Director of Mobility Policy Research, Center for Urban Transportation Research, University of South Florida
I do transit and community transportation for a living; plus I recently returned from a trip to Asia, visiting very first world and quite third world cities. I’ve also been to wonderful transit destinations in Europe and, of course, London. Where the pedestrian infrastructure is good and transit service is frequent, reliable, clean, and goes everywhere, transit is used and it is popular. Where any of those parts are missing, and generally the trajectory of those factors tend to bunch together (poor pedestrian infrastructure and infrequent bus service as an example), transit becomes a service the desperate a/k/a transportation-challenged populations. I will add a word for clear and easy information at stops and stations. Unclear, complicated information at bus stops is one reason, I believe, why rail tends to attract more people.
While the fall on ridership is a worthy subject for investigation, I think there needs to be more time before recognizing it as a long-term trend. Too many ‘degrees of freedom’ on any credible model to make it powerful enough.
It is a similar situation to the one that led many in the transit community trumpeting peak car and the start of a multi-decade delay on car usage in 2010-11, based only on 3-year data of what are otherwise persistent patterns, and then become dismayed at the rebound on car usage in 2014-16.
Maybe we’re measuring the wrong thing. # of boardings is just one measure, and potentially not the most important one (or at least, not the one actually being prioritized by the agencies).
What’s the trend in ridership measured in passenger-miles?
From a research design perspective, it might be possible to use the change in transit ridership as a dependent variable, use change in service, car ownership, or demographics, and then a range of control variables. Regarding the metro area issue: metro areas are useful for time series analysis, as they provide constant geography. For correllational analyses, likely better off using the FHWA urbanized area boundaries.
Here’s a relevant article from Australian experience. Interestingly, the PT mode share based on employment location drops rapidly with distance from the city centre. In the belt 1-2 km from the centre (about 1 mile) PT mode share approximately halved relative to the centre. So the entire city’s PT usage was dependent on the number of jobs in a small area at the centre of the city. Decline of PT mode share in some Australian cities was related to a relative decline in central city employment, while strong increases in mode share in Melbourne and Sydney were related to increases in central city employment. There was also a relationship with the strength of the PT service in suburbs experiencing residential intensification.