Guest Post: An Application of ODX Data

Catherine Vanderwaart graduated from MIT’s Master’s in City Planning and Master’s in Transportation programs earlier this year. She currently works for our friends at WMATA. The below is a guest post she wrote for the Data Blog about her thesis, which used data from the ODX model.

When I did my Master’s thesis at MIT, I was fortunate to be able to work with the MBTA and use its data for my research. ODX data, described in a recent post, had recently become available, but everyone was still trying to figure out the best ways to use it. Since ODX provides more detail on passengers’ travel behavior than previously existed, I focused on how the new data could be used in service planning, particularly long-range bus planning.

I examined a few locations around the city in detail. I chose locations that have changed a lot (and which therefore are likely to be underserved by bus) or that previous research had identified as needing attention. These included Kendall Square, Longwood Medical Area, and an area covering parts of Roxbury, Dorchester, and Mattapan (I called this the RDM study area). With the data on full trips in ODX, I could look at a variety of statistics for each location: the number of journeys and journeys per capita around the Boston region, average travel times and speeds, the proportion of journeys by different modes, typical numbers of transfers coming from different directions, and the distribution of trips across different times of day. With all of this data, I was able to identify some of the structural weak points of the current MBTA network and propose some improvements.

One interesting piece of this analysis was a look at how often the same CharlieCards and CharlieTickets were used in each location.1

The charts below show these travel patterns for the month of October 2014, with the different columns showing how the trips starting or ending at Kendall relate to all travel on the card.2 The left-hand bar is occasional visitors to Kendall who usually use the T to travel to other parts of the city. The right-hand column shows those travelers who mainly travel to and from Kendall and rarely travel elsewhere in the system.

The second chart shows the same thing, but for the total number of trips to and from Kendall instead of the unique cards used. Here we see that most of the trips to and from Kendall are being made by people who rarely travel elsewhere; many of these are likely commuters who work in the area.

The interesting part comes when these results are compared to results from other parts of the city. Here are the results for the RDM study area:

Distribution of transactions on cards in RDM Area
Distribution of journeys on cards inRDM area

The pattern here is similar, though less pronounced. Most of the travel is by the same group of frequent travelers (many of them likely residents of the area), though there are more occasional and irregular visitors.

And here are the results for the Longwood Medical Area:

Distribution of transactions on cards in Longwood
Distributions of journeys by card in Longwood

Here we see a completely different pattern. There are many more visitors to Longwood for whom the Longwood travel is an exception, with most of their travel in other places. And much higher percentages of journeys are made by infrequent or irregular visitors than for Kendall or the RDM study area.

The fare types used on all these trips also varied by location:

Nearly 90% of travel to Kendall paid the regular adult fare, and many of the rest paid a regular fare as part of a commuter rail pass. In the Roxbury, Dorchester, and Mattapan area, by contrast, less than two-thirds of journeys paid the regular adult fare, with about a quarter of all trips made by students (students here mean middle and high school students with school-sponsored passes, not college students). More people with disabilities (users of the Transit Access Pass card) travel in the RDM study area and Longwood than in Kendall, so the MBTA might target these locations when upgrading bus stops to improve accessibility.

What are the implications of these results for planning? There are many, but one example is this: Bus planners often need to balance efficiency, or providing exactly the service that people need when and where they need it, with simplicity. Routes that serve one destination in the peak but only run a shortened version at other times can be more efficient, but may confuse riders who only use the route occasionally. This kind of service is likely to be welcome in Kendall, where most transit users are regular commuters who are able to get to know the routes and schedules quite well. In Longwood such an arrangement may cause too much confusion or need additional explanatory signage, since so many of the passengers using the route are out of their normal routines.

  1. Nearly all origin locations are inferred by the ODX algorithm, while at the time I did this research only about 60% of destinations were. In order to avoid complex adjustments for this asymmetry, I doubled the number of journeys originating at each location to approximate the number of journeys that either started or ended there. [Back]
  1. This analysis looked at unique CharlieCard and CharlieTicket serial numbers, which may not always correlate exactly to the number of passengers, since riders can use more than one card and multiple people can ride on the same stored-value card. For simplicity in this explanation, though, I’ll assume that the relationship of cards to people is one to one. [Back]