Ten years ago I blogged that smartphones would not have a revolutionary effect on public transportation. I think the trend since then has vindicated me – routing apps have not had visible impact on ridership, whereas traditional investments in better service have. I bring this up because a brief conversation with an Israeli public transportation activist reminded me of how the British and American focus on data corrupts institutions elsewhere, to the point that people culturally cringe toward the generally wealthy UK and US and learn from them even on matters where they fail, that is public transportation.
What is data?
Data, in the context of transportation, is any information about how people travel or could travel. It is mostly collected through personal surveillance, whether by smartcard giving the agency exact origin and destination data linked to a specific person, or by a smartphone app doing same. It can also be collected from census data on travel, but the trend is to seek non-census sources and prefer surveillance apps for more granular information.
What are the uses of data?
Data can be used to plan networks more precisely. For examples, granular origin-destination data can be used to plan bus networks, time of day data can be used for schedule planning, and demographic crosstabs can be used to see whether there are patterns in ridership that require addressing (maybe people who don’t speak the language struggle with monthly passes?).
This can also be done in public, hence the fascination with open data. The idea behind open data is to some extent about transparency, but it’s adopted far more widely in places with poor general transparency, like the UK and US, than in places with good transparency, like Sweden. The justification in the US at least is less about informing the public and more about creating a pool of open data, like GTFS, that can be used for third-party apps, such include trip planners and next-bus apps, as well as for data visualization. The same data can also be used analytically, and thus for example TransitCenter has the Equity Dashboard showing unequal access by various demographic categories.
Has this worked?
Not really. An app showing me that the bus in Boston will not arrive for another 17 minutes is not going to make me ride the bus (I took a taxi that time; the public bus tracker was down but there was some dodgy third-party app). A schedule in which the bus shows up every 6 minutes without variation is.
Unsurprisingly to me, apps have made no difference in modal choice or in overall travel numbers. What visible effects there are come from the growth of TNCs, and even they have had a very small effect on overall mass transit ridership. This was surprising to other people – that post I wrote 10 years ago got a lot of criticism, and Reihan Salam dubbed it “bad Alon – backward-looking, dismissive – rather than good Alon – analytical.” But to me, it wasn’t even a particularly controversial claim. Innovating in an industry requires a lot of knowledge about where its current technological frontier is, and the sort of people pushing open data as the solution are, with few exceptions, incurious about recent success cases.
So that’s for apps. What about open data’s use for planning? That, too, is limited. I think Uday’s work showing how the Fairmount Line in Boston does provide as good job access as the subway is really good illustration of Boston’s transport planning failures. But it is less important to illustrate failure than to fix it. The planners who moved the Orange Line from black Roxbury to white Jamaica Plain didn’t need data; they needed to be fired for racism and replaced with people who don’t bustitute service to black neighborhoods. The Fairmount situation is likewise much less about data and more about a combination of racial sensitivity and understanding global (i.e. non-North American) best practices regarding mainline rail frequency.
Okay, so data is insufficient, but perhaps it’s necessary?
Nope. Important aspects of planning require either very coarse information, readily available not just from conventional present-day census sources but often also from the state of data analysis of the 1920s and 30s.
If anything, more recent (say, post-1970s) innovations in public transportation planning have made granular data less important rather than more important. The frequency-ridership spiral, not yet understood in the postwar era back when trips were CBD-centric and preexisting frequency was so high small cuts didn’t spiral, means that frequency must depend on minimum guidelines and not on granular time-of-day travel data. Changes in the nature of work also mean that split shifts are harder to sustain for the labor force now than then, which makes flat schedules better. In effect, how first-world rail transit works today is that costs depend almost entirely on the peak, and midday off-peak service is almost free to provide up to the point where it matches the peak.
Bus network redesigns have had a similar effect. Carlos Daganzo was adamant on not relying on current travel data in network redesign, because it only reveals how people travel today, not how people would travel on a redesigned network. It’s of course useful to know the major activity nodes, population density levels, etc., but matching origins to destinations is not useful.
At regional and intercity scale, the growth of integrated timed transfer networks is telling as well. The Swiss planner has no need for detailed surveillance app data to figure out how to precisely match where trains from St. Gallen, Biel, or Zug go and at what time. Instead, the national ITT system means trains run hourly with timed connections to everywhere; the decision of where the one-seat rides goes can be based on special patterns, but it’s a second-order effect. It’s not how Flixbus plans its service, but the modal splits of Switzerland are not achieved elsewhere in Europe, let alone in app-oriented North America.
Learning worst-industry practices
Britain and the US are complex, wealthy societies. London and New York also have high mass transit ridership, by virtue of size, and are globally familiar, especially given that they use English. It’s very easy to overlearn from them, to look at TfL’s open data and say “we want that,” even when the impact of such learning is limited. It’s harder to synthesize the real innovations in scheduling, signaling, fare payment, and construction.
It’s fortunate that there are parts of the world that don’t automatically think everything done in the core Anglosphere is the bee’s knees. Israel is among them – its idea of a normal country is pan-Western European – but even there it’s so easy to err and adopt worst industry practices just because of the cultural cachet of London and New York.