Illustration by Alysheia Shaw-Dansby

Modeling the Effects of Public Transit Investments on Equity in Access to Employment in Washington, DC

Data@Urban
7 min readMar 21, 2024

The $1.2 trillion Infrastructure Investment and Jobs Act (PDF) offers an exciting opportunity for localities to make critically needed transit investments to close gaps in access to transit for people of color. However, the competitive funding applications ask for data-driven equity analyses, which research has shown poses a barrier for counties with fewer transportation staff, in part leading to less funding for those counties. Public agencies need easy-to-use tools that can help model the equity impacts of transit projects and prioritize investments.

Case study: DC bus priority lanes and bus frequency

Responding to this need, the Urban Institute developed a new method to model the impact of transit investments on equity in access to opportunity. In partnership with the Washington Metropolitan Area Transit Authority (Metro), we piloted this approach on two investment scenarios. First, we measured the potential accessibility impacts of Washington, DC’s proposed priority bus lanes, which are designed to speed transit service by allowing buses to bypass traffic congestion. Second, we estimated the impacts of increasing the frequency of the existing bus lines along routes where the lanes are being proposed instead. Improving bus service is particularly critical for equity in access to opportunity, as Metro estimates (PDF) that 81 percent of bus riders are people of color and 46 percent of bus riders come from households with low incomes.

To model each of these scenarios, we created transit data that reflect the public transit network with the proposed investment. We then used the modified data to calculate transit travel times to jobs by census block group and compared estimated outcomes by neighborhood racial and ethnic demographics. The demographics of neighborhoods through which bus routes run — the data we examine — may be different than those of the demographics of bus riders, which we do not investigate.

General transit feed specification data

Public transit data is commonly published by transit agencies in the General Transit Feed Specification (GTFS) format, which allows it to be used by a wide variety of software applications. A GTFS feed, which contains transit schedule information for a given period, is composed of several relational text files that each describe a particular aspect of public transit (e.g., stops, trips, routes) in a single zip file. To model a public transit investment, we created a modified GTFS data file for the investing agency that reflects the new investment. The image below shows a subset of the GTFS.txt files that we used most in our analysis (with hypothetical data):

Because of its complex relational structure, GTFS data can be difficult for agencies to produce or manipulate, which can be a barrier to modeling the impacts of potential new investments. In our work with GTFS data, we used the gtfstools R package developed by Daniel Herszenhut, Rafael H. M. Pereira, Pedro R. Andrade, Joao Bazzo, Mark Padgham, and Marcus Saraiva at the Institute for Applied Economic Research (IPEA). This package contains several helpful functions to ease the process of reading, writing, and manipulating GTFS data.

Modeling bus priority lane construction and enforcement

To model bus priority lane construction and enforcement, we first performed a spatial join between a shapefile of bus priority lanes provided by Metro and bus stop locations to identify the stops that fall on each lane. We then used the get_trip_segment_duration function from gtfstools to calculate the trip times between each stop, or set of two consecutive stops, for the dozens of daily trips that pass along the lane.

We then calculated updated travel time durations for all segments along priority lanes at different levels of enforcement. We did so by randomly assigning whether each trip segment receives the expected priority lane speed gains using the enforcement levels. To model 75 percent enforcement, each segment has a 75 percent probability of receiving the speed gains (i.e., the priority lane is not blocked by other vehicles and the bus can travel more quickly) and a 25 percent probability of receiving no speed gain (i.e., the priority lane is blocked and the bus needs to travel in regular traffic). For segments that receive the speed gains, we calculated the updated segment duration by reducing the original segment duration by the trip speed multipliers below.

The bus stop times published in Metro’s baseline GTFS data reflect bus trips taking longer during periods of higher traffic, such as morning or evening rush hour. Therefore, the most significant gains in speed from using bus priority lanes, which aim to enable uninterrupted travel, will occur during peak traffic hours. We based the specific multipliers on the literature and the published bus schedules by Metro (its adjustments during the day help provide an estimate for how quickly a bus could move if there were no traffic). These numbers are best-case-scenario speed increases on the assumption that buses will be able to reach the scheduled speeds of the night time period. This assumption can only happen with perfect enforcement, no additional delays related to passenger boarding and alighting (which are likely to be significant during high-demand periods), and no blockages of priority lanes. For comparison, New York’s Metropolitan Transportation Authority estimates that bus routes using priority lanes improve their travel times by an average of 19 percent (PDF).

Metro currently considers bus routes with 12 minutes or fewer between trips to be a “high-frequency route,” so these updated frequencies represent a huge increase in bus service. Across the 72 affected routes, this increased the total daily weekday trips from 6,417 to 14,483–or a 125 percent increase in the number of daily trips. In the cases where multiple bus routes have significant overlap in their paths (such as the 52 and 54 buses that both travel largely north-south on 14th Street NW in DC), we staggered the departure times of the overlapping buses to achieve the shortest possible headways between bus arrivals on shared stops.

We then generated updated stop times for the routes based on the frequencies.txt file using the frequencies_to_stop_times function from gtfstools. This process creates new entries in the trips.txt and stop_times.txt files to reflect the bus frequencies defined in frequencies.txt, which we used to replace the original data for those routes. Finally, we combined the modified data for the relevant routes that have 20 percent of stops in the priority lanes with the unmodified data for all other routes and saved the updated GTFS file.

Calculating access to employment

To model access to employment, we calculated the number of jobs (both raw count and competition-adjusted) that are accessible within a 30-minute public transit trip for a worker living in each census block group in DC. We used the r5r package — also developed by researchers at IPEA — to calculate these travel times. The package uses GTFS data and OpenStreetMap road data as inputs to calculate the fastest trip between start and end points via various modes of transportation. To account for the variability in transit travel times caused by just making or missing a transit connection, the travel_time_matrix function calculates the travel time departing every minute in a user-specified time window around the departure time (we used 20 minutes) and returns the median travel time by default.

We calculated the travel times in the baseline scenario (unmodified Metro GTFS data), and then in each of our transit investment scenarios using the modified GTFS files to understand how each investment scenario affects equity in access to employment. Though our analysis focuses on spatial access to employment, we recognize there are many other factors affecting equitable access to employment that are outside the scope of our analysis, such as skills mismatch, discriminatory hiring, job quality, and others. We also estimated some operational effects of the different investment scenarios to begin to explore relative impact on operating costs and environmental impact.

Results

We found enforcement had a significant effect on the impact of priority lanes on access to employment. At 50 percent and 75 percent enforcement, increasing current bus frequency offers a greater increase in access to jobs, but with 100 percent enforcement, we found that bus priority lanes offer a substantially larger increase in jobs accessible within 30 minutes than increasing bus frequency. This finding is particularly significant, as bus priority lanes could reduce operating hours and thus costs, whereas increasing transit frequency would increase agency operating costs. This analysis doesn’t capture all benefits of increased frequency, such as planning flexibility for riders and service reliability. Also, the calculated cost savings of priority lanes are based on the best-case-scenario speed increases as discussed above and our assumption that jobs and workers don’t move in response to the transit investments.

Although all of DC’s wards would experience positive changes across all scenarios, Ward 7 and Ward 8 — which have populations that are 85 percent and 83 percent Black and together represent half of DC’s Black residents — experience the smallest gains. It’s also important to consider that the vast majority of bus riders are people of color and that Metro survey data (PDF) has found that many bus lanes in predominantly white neighborhoods have majority riders of color.

Therefore, while investments in bus service are likely to close racial access gaps, our findings highlight the need to invest in additional bus routes and priority lanes in Wards 7 and 8, create high-quality jobs in Black communities, and invest in affordable housing in job opportunity rich areas. For more information on our results, see our Urban Wire blog post.

Next steps

We hope to build a public-facing tool that will enable transit agency staff and community members to model the impact of transit investments in their own communities and prioritize uses of public resources toward transit investments that will have the greatest impact on advancing equity in access to opportunity. We also hope to expand this work to incorporate additional types of public transit investments, such as infill stations and route changes.

-Alena Stern

-Yonah Freemark

-Tina Stacy

-Manuel Alcala Kovalski

Want to learn more? Sign up for the Data@Urban newsletter.

--

--

Data@Urban
Data@Urban

Written by Data@Urban

Data@Urban is a place to explore the code, data, products, and processes that bring Urban Institute research to life.

No responses yet