Improving Cost-Benefit Analysis for the Bus Network

The MBTA strives to provide good service to riders, but what does providing “good service” really mean? In this post, we will explore this concept as it relates to bus service. For over a decade, we have been redefining our approach to “good bus service” to better capture the benefits bus service can provide for our riders. We have refined our process to understanding and measuring this by incorporating additional benefits that a bus route can provide to both passengers and the MBTA system as a whole.

In the 2010 Service Delivery Policy (SDP), we began calculating cost-benefit by comparing the ridership on the route to the net cost incurred to operate the route. This simple approach was designed to help service planners identify high-performing routes, but it left out benefits that aren’t captured solely by ridership. In the 2017 revision of the SDP, the MBTA took a more holistic approach to evaluating bus route cost-benefit by including transit-dependent ridership (the percentage of route riders that are transit critical) and a measure of route value to the network that assesses how well a bus route meets total travel demand. The most recent SDP revision further refines the 2017 approach to more effectively capture all of the benefits that our bus service provides to the entire transit network. This post explains the reasoning behind the changes we made to the measure and shares some of our results for the Fall 2019 and Fall 2021 ratings.

2017 Measure

The 2017 version of the measure critically acknowledges more of the benefits a bus route can provide to our transit network; this was accomplished by diversifying our approach to determining these benefits. The 2017 version of the measure added two additional benefit components: transit dependency and value to the network. These components allowed routes that may have scored worse based on ridership alone to score higher if they serve a population that relies more heavily on transit or if they serve a geography that has limited additional transit coverage.  

Transit dependency was defined as the proportion of fare payments that were made using MBTA reduced fare programs (Student, Senior, Transportation Access Pass (TAP), Blind CharlieCards, and Youth Pass) on each route. Value to the network contained three sub-components: transfer rate, unique access, and access to destinations. Transfer rate was defined as the percentage of taps on a route that were part of multi-stage journeys. Unique access refers to the residential population covered by a route. Areas covered by multiple bus routes had their population divided among the relevant routes. Finally, access to destinations refers to the number of jobs served by a bus route. These three sub-components were combined to get a single measure of value to the network.

After their calculations, the three benefit components were then scaled from 0 to 1 (to account for the different units of all measure components) and combined according to the weights set in the 2017 SDP: 70% for ridership, 15% for transit dependency, and 15% for value to the network. While ridership still receives the largest weight, the other two components ensure that low-ridership routes that provide critical service — through providing unique service to an otherwise transit-deprived neighborhood, serving a transit-dependent population, or connecting people to the rapid transit system — are able to score well in the measure.

2021 Measure

Between 2017 and 2021, we rethought our approach to the benefit measure to make further improvements. We acquired new data sources (mostly from Streetlight) that allowed us to refine how we measure the value a route provides our network. We also reconsidered how to combine the three benefit components in a meaningful way. Our updated benefit components are defined as follows:

  1. Ridership Generation & Propensity represents current usage of a route: in addition to measuring boardings alone, we include the average number of a route’s riders that transfer to other services in the “boardings” value. We then multiply this number by the sum of the percentages of current riders on the route that are low-income, people of color, or have limited access to a vehicle (to account for riders that tend to rely more heavily on the MBTA) to get our final value for this component.
  2. Seniors and People with Disabilities using Reduced Fares captures the percentage of validations on the route that are made with certain reduced fare products (e.g., Senior CharlieCards, TAP, and Blind Access CharlieCards).
  3. Access to the Network captures the potential number of trips that could be made using the service. Using location-based services data, this aspect of the measure attributes potential trips (bicycle, pedestrian, and all vehicle travel) to a route based on trip-making along the path of the route within each hour. These trip counts are summed to get the total volume of trips that could potentially be served by the route. In addition, we weight the volume by the total number of trips that are made by low-income people, people of color, and people with limited access to a vehicle. Finally, we weight the number of potential trips by the proportion of service that the route in question provides. For example, the 1 and the 66 both serve Nubian Square to Harvard at 8am. The 1 runs 8 trips during this hour and the 66 runs 7 trips. So, we assign 8/15 of the total travel demand between Nubian and Harvard for the 8am hour to the 1 and 7/15 of this demand to the 66. This process is applied to all routes for all times of day.

The weights of 70% for ridership generation & propensity and 15% for the other two components were kept the same. The changes since 2017 provide a more refined representation of the benefits that we are trying to capture with this measure. For example, including the transfer rate with the number of boardings captures the total ridership that a given route generates on the rest of the system in addition to the number of boardings on the route itself. This allows routes that play a role in connecting people to our transit network to score better on the measure than they would otherwise. The second component of the measure has been made more specific — rather than grouping all reduced fares programs together, we have focused on programs that are used by seniors and people with disabilities. This serves as a better estimate of riders that may have limited mobility.

The third component of the measure constitutes the largest change from the 2017 version. Rather than using residential population and total number of jobs to derive a measure of access, the new version of this component is based on the number of trips (on all modes) being made along the path of the route. This gives the route credit for serving origin-destination connections with high volumes, rather than just those with high populations or high numbers of jobs.


Figure 1 shows the high-level results for bus routes in the Fall 2021 rating. Unsurprisingly, most of the routes with high benefit scores are key bus routes – having more trips per day not only gives these routes higher ridership, but it also gives them more opportunities to capture potential travel volume in the access to the network component. Key bus routes (e.g., 111, 66) tend to have high relative costs — this is also a function of their high number of trips per day.

Scatter plot showing cost-benefit with each point representing a bus line. Benefit on x-axis and cost on y-axis, 1:1 reference line shows if cost is greater than benefit or vice versa. Majority of points are clustered at bottom left corner, less going up to top right corner. Lines with higher benefit include 17, 21, 114. 116. 117. Lines with higher cost include 57, 71, 111, 66.
Figure 1: Scaled cost displayed against scaled benefit for MBTA bus routes in the Fall 2021 rating.

Calculating the cost-benefit measure (using the same methods) for multiple ratings allows us to understand how a routes’ performance may have changed, whether because of the pandemic or other changes to ridership patterns. Figure 2 compares the benefit rank of routes in 2019 and 2021. This graph highlights routes that have had more durable pandemic-era ridership (e.g., 109), as well as routes with percentages of senior/TAP riders that have increased more than the system average (e.g., 14). It also shows routes where relatively low Fall 2021 ridership has caused a drop in the benefit rank (e.g., SL1/SL2).

Scatter plot with each point representing a bus line showing 2019 vs. 2021 rank. 2019 on x-axis and 2021 on y-axis, 1:1 reference line shows if 2021 is greater than 2019 or vice versa. Points mostly evenly spread along 1:1 line, about even amount better in 2019 vs. 2021. Bus lines better in 2019: 11, 18,9, SL1/SL2, etc. Lines better in 2021: 222. 119. 14, 109, etc.
Figure 2: Benefit rank of routes in 2019 compared to benefit rank of routes in 2021. Only routes that ran service in both years were included in this plot.

Figure 3 shows the benefit scores for Key Bus Routes. As explained above, each benefit category was rescaled from 0 to 1, with 1 being the highest possible score for that category. Routes with high ridership tend to perform well on the measure, as ridership makes up the majority of the overall benefit score. However, this figure highlights how non-ridership components can help increase a route’s overall score.

Bar chart with key bus routes on x-axis and benefit value on y-axis. Horizontal bars are 3 colors, red to show ridership score, blue to show mobility score, and yellow to show network access score. Shows benefit results ranging from .5 to 2. Lowest benefit score: SL1/SL2, 71, 73, 57. Highest benefit score: 114, 116, 117, 28, SL4/SL5, 1.
Figure 3: Bar chart showing benefit scores on Key Bus Routes. Bars are ordered by weighted total benefit score.

As an example, we can walk through the calculations for the SL4/SL5, which has a high proportion of riders with limited mobility. According to Automated Fare Collection (AFC) data, around 24% of the transactions on those routes are made using Senior/TAP fare products — much higher than the median value of about 12%. This gives those routes a very high scaled Mobility value. SL4/SL5 unweighted ridership during Fall 2021 was around 8,200 boardings per day, of which 39% were part of multi-ride journeys.  According to the 2015-2017 MBTA Systemwide Passenger Survey, 61% of riders on the SL4/SL5 are people of color, 36% are members of low-income households, and 60% have limited access to a vehicle. To calculate the final value for Ridership Generation & Propensity, boardings are weighted by the transfer rate and the sum of the percentages of people of color, low-income riders, and riders with limited access to a vehicle. In the case of the SL4/SL5, this means that we arrive at a Ridership Generation & Propensity value of 8,200 * (1 + 0.39) * (0.61 + 0.36 + 0.6) = 18,000 — one of the highest on the system but slightly below other Key Bus Routes. While the SL4/SL5 serves several origin-destination (OD) pairs with high travel volume, some of these OD pairs are served by other MBTA services — like the Orange Line between the Financial District and Chinatown. So, the SL4/SL5 only receives partial credit for serving these ODs, slightly deflating its Network Access score. Taken together, the SL4/SL5’s very high Mobility value, high Ridership, and slightly lower Network Access score place it relatively far up the 2021 rankings.

The results of the bus routes cost-benefit analysis help Service Planning identify specific characteristics of high performing bus routes. It can also help identify potential ways to redesign low performing routes – for example, a route with a high percentage of transit critical riders or senior/TAP riders but a low network access score may benefit from rerouting to improve access to important destinations. While this measure is intended to evaluate existing bus routes, we used a similar approach in developing new and modified routes in the Bus Network Redesign. We will evaluate those changes in the our SDP report as they are finalized and implemented. The MBTA will continue to refine this measure as we acquire new data sources in order to thoughtfully understand and measure good bus service.