Declining MBTA Bus Speeds

It’s no secret that traffic congestion in the Boston area is getting worse. In the context of OPMI’s deep dive into ridership, we wanted to evaluate how MBTA bus speeds have changed over time and what relationships that may have with ridership. While the results from the ridership study are forthcoming, we can share some of the variables that we examined and discuss them.

Overview of Methods

In order to calculate bus speeds we used data from the MBTA’s automatic vehicle location (AVL) systems. Each bus is equipped with a GPS device that records the bus’s location, both each minute and each time the bus passes through a designated timepoint (this system was recently upgraded to report location much more frequently, but that new system was not in place for any of the data analyzed here). These timepoint “crossings” data and “heartbeat” locations are combined with the announcement data from TransitMaster (the software that runs on buses and among other things, provides route and stop information to the on-board systems) in post-processing to create a table that has the time we estimate each bus arrived at and departed from each stop. These data may be inaccurate for some inferred stops or stops with missing data, but for measuring the run-time of an entire route they should be accurate. Most routes start and end at timepoints when the AVL system would almost always record an actual crossing time. Read How the MBTA Tracks Vehicles for more information.

To get the run-time of each trip, the time of departure from the first stop was subtracted from the arrival time at the last stop. We then used this time interval and the length of the route (determined from GTFS) to calculate the average speed along the route for the trip. These trips were then grouped by time of day and by route. Finally, a yearly median speed was determined for each route and time period from 2013 through 2017. For consistency, the time span analyzed each year was limited to the period from September 1 through December 31.

Some additional important notes on the data:

• Since the time span of the route includes its entire length, it will include dwell time at any stops (apart from the first and last stop of the route). This should, however, more accurately reflect the passenger experience, which includes such dwells.

• Since these speeds are for the whole route, they may not represent the passenger experience in certain cases; for example, a route that had a longer section with lower levels of traffic, but a relatively shorter section that experienced higher levels of traffic and ridership may have a higher speed along the entire route than most people experience.

• Variants are unaccounted for here (all trips are recorded as the base route) so a route with a high number of variants, or where the variants have changed over time, may be inaccurate. However, from 2013 through 2017 there was not a significant amount of change to most MBTA bus routes.

General Decline

The average bus speed on local buses and key bus routes is shown below:

Line chart of Average Speed of Key Bus Routes on Weekdays, year over year from 2013 to 2017. Average speeds are highest in off peak at night and lowest in PM peak. Across all categories, average speeds lowered a bit from 2013 to 2014, but rose in 2016 and lowered again in 2017.
Line chart showing Average Speed of All Non-Express Bus Routes on Weekdays, year over year from 2013 t0 2017. Compared to the Key Bus Routes chart, the average speed peak in 2016 is higher.

In the above, a few things stand out. First, it is clear that bus speeds have fallen, especially during peak periods. For the busiest routes, the Key Bus routes, speeds have dropped over 1.5 miles per hour at peak and just over 1 MPH for all weekday trips.

The second interesting reveal is that the drop in speeds is not steady. There is a significant drop from 2013 to 2014, and then for most time periods, either a steady speed or a speed increase in the years from 2014 through 2016, before another drop in 2017. It seems unlikely that the increase in 2016 is due to less congestion in the region in that year, but there could be short-term reasons why buses were slightly faster in this year. It could also be a problem with the data. We will continue looking into this to validate our data and see if we can figure out what might have happened in 2016, but the trend seems clear, especially when we look at the national data.

System and Nation-Wide

We also checked the systemwide speeds at the top 27 transit agencies in the country. In order to find speed at the route level, we’d have to check internal agency data, and comparing apples to apples would be pretty challenging, but we can take a crude look at speeds by looking into National Transit Database Statistics. The NTD has statistics on total vehicle revenue miles provided, as well as total vehicle revenue hours provided. These tell us the total number of hours when transit vehicles were in revenue service, and the total number of miles they traveled while in revenue service (excluding “deadhead” travel when the bus was not accepting passengers). We looked at the speed change from FY 2009 to FY17:

Name20092017Change from ’09-’17
MTA Maryland11.5413.2815.1%
Metro Transit (Minneapolis)11.6312.154.5%
SacRT (Sacramento)11.1111.231.1%
MARTA (Atlanta)12.4712.41-0.5%
MTS (San Diego)11.0510.95-0.9%
RTC (Las Vegas)11.3411.21-1.2%
Milwaukee County Transit System12.7012.50-1.6%
TheBus (Honolulu)13.2012.92-2.1%
Metro Transit (St. Louis)13.5913.25-2.5%
SEPTA (Philadelphia)10.239.90-3.3%
Port Authority (Pittsburgh)13.6013.07-3.9%
GCRTA (Cleveland)11.8611.37-4.1%
MTA NYC7.747.40-4.3%
Miami-Dade Transit12.0011.47-4.4%
WMATA (DC)10.7210.14-5.4%
VTA (San Jose)12.3911.61-6.3%
DART (Dallas)13.6312.71-6.8%
King County Metro (Seattle)11.3210.52-7.0%
Metro (Houston)14.6013.55-7.2%
Muni (SF)7.917.33-7.3%
TriMet (Portland, OR)12.0711.11-7.9%
RTD (Denver)13.9712.75-8.8%
LACMTA (Los Angeles)11.7810.69-9.3%
Valley Metro (Phoenix)12.9611.11-14.3%
UTA (Salt Lake City)18.5513.87-25.2%

As you can see, the speed of all bus travel in the agencies studied dropped by about 5% in the designated time period, and dropped about 7.7% at the MBTA. Note that this is for all times of the day, week and year, so it’s likely that (as we saw in the MBTA route-level data) some time periods have dropped more than others. For agencies at the very top or bottom of this chart, it’s likely they changed their service provision. For example, if the MTA in Maryland provided more commuter routes that traveled on highways during this time, that could mean their average speed increased because they added service on routes that simply travel faster. Conversely, if UTA in Salt Lake City added more service in denser parts of their region, you would expect to see average speeds drop overall regardless of other factors.

Reversing the Trend

What can we do to improve bus speeds?

MBTA bus speed is a notable problem and is something that we take seriously and want to improve. Much of what slows buses is outside of the MBTA’s control, but the new fare collection system, which will allow all-door boarding, should significantly decrease dwell times. However, as we noted in All Door Boarding on the Silver Line, dwell times are most improved by all-door boarding at the busiest stops, so its effect on the overall speed of the route may be limited.

There were a few routes that seemed to buck the decline in 2017, during AM peaks:


These three routes most benefitted from the City of Everett’s morning bus-only lane. As you can see, the improvement in speed wasn’t dramatic but, as noted above, these speeds are for the entire route in both directions. Any improvement in 2017 is a strong data point in the bus lane’s favor. With the implementation of a similar lane on Washington St. in Roslindale this spring, we hope to see improvement on routes 34, 34E, 35, 36, 37, 40 and 50 as well, and as additional bus lanes are implemented in Cambridge, Arlington and Watertown, we hope to see similar impacts. The MBTA will continue to work with municipal partners to pilot and implement similar bus interventions throughout the region, and we’ll look at these changes in a future blog post once the data comes in.