The MBTA reports its ridership as a vital indicator of how we’re doing. As with any data, it’s important when we examine trends in ridership and compare data that we are making sound comparisons to ensure we are comparing apples to apples. For this reason, we try to report ridership during an average weekday, Saturday, or Sunday, over a period of time, rather than simply reporting out the total number of trips that passengers took. This post will examine why this is so important, and how changes in the calendar and one-time events can distort ridership totals if they are not accounted for.
In the National Transit Database’s monthly module, every transit agency reports their ridership total as unlinked passenger trips (UPT) for each month. While this is a good database of wide-ranging data, and collecting it in one file certainly saves researchers time, the monthly data is highly vulnerable to natural variation in the calendar, and it can be easy to misuse the data.
Let’s do an exercise. Assume for the moment that the MBTA’s ridership was steady each day, and we provided 1.2 million UPT on our bus and subway system each weekday, 700k each Saturday, 450k each Sunday, and 500k each day of Holiday service. If we applied these theoretical numbers to the 2016 and 2017 calendar, our expected monthly ridership would be the following:
As you can see, even if there were no seasonal changes or overall trends in ridership, ridership could vary widely simply based on the calendar — even between months that have the same number of days. In the above chart, the expected ridership in December 2016 is 4.7% higher than the expected ridership in December 2017, simply because there were 2 extra days of weekday service.
These effects can even add up over the whole year. Using the above method, the ridership for the entirety of 2016 would be 2.25 million more trips than 2017, even if demand and service were entirely steady each day.
Like any major city, Boston holds a number of big events throughout the year. Getting people to and from these events efficiently and safely is a key goal of the MBTA, and plenty of planning and hard work goes into this on our end.
These events can also influence ridership. While event attendees count just as much as anyone else, to measure the demand on the system and for future planning purposes, we need to account for these events separately. Ridership from special events is not only dependent on that event happening at all (congratulations SEPTA!) but also dependent on when they occur, which is largely out of the MBTA’s control.
Sports championship parades and the Boston Marathon are a good example of this. One might think that a parade, or the Marathon is a boon to ridership, as hundreds of thousands of people come into the city who normally would not, and many of them use the T. But looking at the data, the effect of this seems to depend on which day of the week the parade occurs. Parades that are held on weekdays (like the Patriots’ championship parade on Tuesday, February 7, 2017) add ridership, especially on commuter rail, but they also displace ridership as some people stay home (or some people who go to the parade would normally be going to work and so they take the same number of trips on the T that day). Similarly, the Boston Marathon brings in hundreds of thousands of visitors to the city, but since many people in Massachusetts have the day off, our overall ridership is not higher than a regular weekday. But parades that are held on the weekend, such as the Red Sox’s celebration in 2013, add a significant number of trips to the system, since people are taking trips that they wouldn’t normally make.
Here’s the total number of taps (and ticket validations or cash-on-board payments) on bus and subway for Saturdays in October and November 2013, and for comparison, the nearest Saturdays in 2014, when the Red Sox missed the playoffs. Note that “taps” are not the same as ridership, since they don’t include behind-the-gate transfers or people who don’t interact with the farebox or gate. We also believe the gates undercount slightly when stations are extremely busy, and occasionally we open the gates entirely for safety reasons, during which time none of those entering would be counted. For this purpose though, the data should be usable:
*denotes Red Sox home playoff game on that date
While other events could have influenced ridership on these days, it seems from the data that the Red Sox parade added at least approximately 170,000 trips to the bus and subway system over what we saw on surrounding Saturdays. While extra home games likely do add ridership, their effect is relatively small and as we can see, not really noticeable in the normal variance.
When the Patriots won the Super Bowl in 2017, the parade was held on the following Tuesday. Since this was a weekday, the MBTA’;s ridership was already high, and while we saw higher ridership than normal, the effect was not as pronounced as we saw with the Saturday Red Sox parade. The following chart shows total taps on the Tuesdays in January and February of 2016 and 2017:
While the parade day had the highest number of taps in the set, other days came fairly close, and the parade day had only 75,000 more taps than the average (ignoring the first Tuesday in January, where ridership is low).
Finally, here is the total number of taps for Mondays in April for the last 4 years. Marathon Mondays are starred and in yellow:
Ridership on Marathon Mondays is about 19% lower on average than ridership on the surrounding Mondays (Again, this does not count Commuter Rail ridership, which certainly brings in a lot of people on those days). But, since we run weekday service, this gets counted as a weekday and lowers our average ridership for April.
So, how do we account for all the above? While we do report ridership on a “total trips” basis in certain reporting, whenever we are comparing ridership over time, we try to group days with other days of the same service type (Weekday, Saturday or Sunday), take an average or a median of these days over a time period (for weekdays, usually a month) and then make sure to look at trends over the long term to account for seasonality. We also try to think about any possible influences on ridership besides the demand of passengers and make sure to note them: for example, bus diversions due to maintenance work, particularly bad or good weather, or times when an entire station was closed for renovation.
For an example of this, take a look at our Ridership Quarterly Update presentation. For the main charts, we grouped ridership by day type and displayed the rolling average over 12 months to account for seasonality and smooth out the effects of one-time events. While not perfect (especially with changes in service on weekends from repair work), we think this gets pretty close to the actual level of ridership and can help us determine how things are changing.