Equity in Transit Planning: Are We Using the Right Metrics?

Cesar N. Yahia
Dec 3, 2020 · 8 min read

CapRemap — Austin’s restructuring of its transit service in 2018— raised several equity concerns and accusations of racial discrimination. In spite of CapMetro’s service equity analysis showing compliance with FTA’s policies, activists are still determined that the redesign violates Title VI requirements. So who got this right? and why were equity concerns an issue?

CapMetro and the FTA’s Perspective

Before analyzing the data in greater detail, it is worth mentioning CapMetro’s equity analysis that showed compliance with Title VI— in fact, CapMetro states that the benefit to minorities from the service adjustments far exceeds the potentially adverse impact!

As discussed in an MPO policy meeting, CapMetro evaluated each of the major service adjustments by studying the demographics of a 1/2 mile walk-shed that surrounds the changed routes. The key analysis approach is to first find out whether the % minority population around the walk-shed is greater than the average % minority population in the total service area. If that is the case, proceed to determine whether alternative routes can cover the minority block groups that lost service. This route level analysis is consistent with the FTA’s Title VI requirements.

CapMetro’s results shows that most areas with lost service will be covered by alternative routes and in many instances there will be new high-frequency options as well.

However, this analysis has some limitations:

  1. While it is often mentioned that there will be high frequency routes close to minority groups (similar to other studies that focused exclusively on those high frequency routes), a detailed service frequency analysis seems to be lacking. The addition of high frequency routes does not give the full picture of service changes on frequent and non-frequent routes.
  2. Forming 1/2 mile walk-sheds around routes (with 1/4 mile strips on each side) is a common method for measuring system coverage. Despite that, passengers board their buses at stops, and the 1/4 mile distance is based on the 85th percentile walking distance to those stops. Wouldn’t stop-based coverage, with a 1/4 mile radius around stops, be more appropriate in that case?
  3. The routes analyzed were restricted to those that had a greater than 25% change in geographic coverage or service characteristics, where this 25% threshold was set by CapMetro. Even after selecting the routes with major changes, they were only analyzed further if the % minority population in the walk-shed was greater than 35%. Does this exclude parts of the network that were adversely impacted?
  4. As shown in the figure above, the stops at Gardner Rd and Arthur Stiles Rd are removed, and their location will no longer be within a 1/2 mile walk-shed of any route. However, the minority block group in which they are located (blue) is assumed to be covered by the adjusted route. Clearly, passengers that previously used those stops will no longer be a short walk away from any transit line; but, they are considered to be covered due to the irregular shapes of census block groups. In particular, the analysis assumes that a block group has transit service if any part of its area overlaps with the walk-shed. A better equity analysis approach would restrict coverage to the area within a 1/4 mile distance from bus stops.

A Peak-Hour Stop-Based Analysis Approach

Focusing on the weekday morning peak service (7–10 a.m.), which targets essential home-based work trips, I implement a stop-level equity analysis of the service changes.

In contrast to the previous analysis: (1) The change in frequency is evaluated at each stop by measuring the difference in doors opening before and after CapRemap (2) A buffer with a 1/4 mile radius is created around each stop to determine the demographic characteristics of affected riders (3) The approach includes changes to all routes — not just ones that pass CapMetro’s thresholds for significant changes and disparate impact (4) The impact of the service change is restricted to the population within the buffer to avoid irregularities in census data and to accurately represent the coverage area.

An HTML file with interactive plots and code can be found here. Links to data sources (GTFS, census, shapefiles, etc.) and the analysis code are also available on github.

Where did the service change?

The following figure shows locations where stops were added or removed. The stops are represented as buffers with a 1/4 mile radius. The color bar represents the proportion of minorities within each census tract. It is evident that many stops were removed in areas with a high proportion of minorities — circled below.

That said, looking only at new or removed stops is not representative of the full service change. The removal of a stop that had low service is highlighted while major service reductions at other stops are not shown. Similarly, in their transition to a high frequency network, CapMetro may have significantly improved the frequency at existing stops without adding many new stops. To get a better illustration, the figure below shows stops that experienced an absolute change of more than 10 buses during the morning peak.

We can still see major service reductions in areas with a high proportion of minorities. However, it is also clear that CapMetro improved service at many locations throughout the network.

For a precise analysis of the change in service frequency and its impact on different demographic groups, some stop-level information is needed.

Stop-level demographic data

Getting stop-level sociodemographic data requires projecting variables from census tracts to the stop buffers. To do so, we can use the proportion of the buffer that lies in each tract. This mapping is best illustrated in an example. The following figure shows how demographic variables are computed for a buffer that overlaps with two tracts (25% of the buffer area is in tract 1). The term inter. area refers to the area of intersection between the buffer and the tract.

In general, the number of minorities and the proportion of minorities within the 1/4 mile buffer can be determined as follows:

In fact, the above equations can be used to map any census demographic data to stop-level data.

Stop-level service change metrics

Then, let's define impact to be the change in service after implementation of CapRemap. From that, doors opening is defined as the impact at stops with improved service, and doors closing is defined as the impact at stops with reduced service.

Aggregate impact of CapRemap

Given the stop-level demographic information and impact , we can now define aggregate metrics that accurately describe the service changes.

The Expected Impact is the average service change experienced by a minority person. In other words, if a minority person was sampled at random from the service area, this is the change in service that they will experience.

The Frac. DO is the fraction of service improvements that went to minorities. Similarly, Frac. DC is the fraction of service reductions inflicted on minorities. In contrast to the Expected Impact metric, those measures are not dependent on the density of minorities in a particular area. For example, greatly improving service in a location that is dense with minorities while leaving out many minority areas unconnected would give a large positive Expected Impact, but this may be undesirable.

The aggregate metrics can be computed for any demographic group by replacing the minority and prop. minority with the census data of choice.

Results

The results show that, on average, Austin’s residents would see fewer buses passing during the morning peak! While CapRemap added frequent lines, this was at the expense of other non-frequent service. If we sample a minority person at random, we would find that she experienced a net loss of around 5 buses passing during the morning peak.

However, in terms of equity, there does not seem to be any bias against minorities

The fraction of service improvements that went to areas with Black people was low (only 9.8% of the total service improvements). At the same time, at 7.8%, the fraction of service reductions that was inflicted on areas with Black people was also low. Overall, minority areas were allotted 55% of the total service improvements (doors opening) and they received 52% of the total service reductions. Meanwhile, areas with White people were allotted 45% of the total service improvements and they received 47% of the total service reductions.

The results indicate that minority areas did not simultaneously receive a lower fraction of the service improvements and a greater fraction of the service reductions, which indicates that there is no apparent bias against minorities in the distribution of service modifications.

Thoughts

In response to complaints by activist Zenobia Joseph, the FTA stated that the total minority population close to frequent service substantially increased. They cited the fact that 50,000 additional minority persons will be close to such frequent service — again, disregarding changes to non-frequent routes.

The FTA dismissed the activists’ complaints, and in a discussion with the Austin American-Statesman CapMetro’s CEO said that “Not only do we not have disparate impacts, (the FTA analysis) says that we are providing even more service to low-income and minority populations in Austin. Sometimes facts matter in the conversation.

True, it is a fact that more minorities are closer to frequent service after CapRemap, but does that give an accurate representation of the service adjustments? While there’s no bias against minorities, the stop-level analysis suggests that, on average, transit users in Austin would experience a reduction in morning peak buses.

The stop-level analysis still relies on census data to project demographic information on stop buffers. An even more exact analysis would be to conduct on-board rider surveys before and after the network redesign. Surveys are an expensive option; however, if designed properly, they can precisely measure the change in ridership across demographic groups.

In summary, the transit community may need to further question the Title VI requirements and the approach used to satisfy them. There does not seem to be any discrimination in CapRemap’s case, but it is clear that the route-level (FTA compliant) analysis has its limitations. It is also apparent that focusing on the increase in frequent service may be misleading since people observed fewer buses on average. Having said that, CapMetro reported in May 2019 that their increase in ridership since CapRemap had been among the highest in the nation — so maybe they are right in overstating the importance of frequent routes?

The Startup

Get smarter at building your thing. Join The Startup’s +776K followers.

Cesar N. Yahia

Written by

optimization and analytics for transportation systems — UT Austin

The Startup

Get smarter at building your thing. Follow to join The Startup’s +8 million monthly readers & +776K followers.

Cesar N. Yahia

Written by

optimization and analytics for transportation systems — UT Austin

The Startup

Get smarter at building your thing. Follow to join The Startup’s +8 million monthly readers & +776K followers.

Medium is an open platform where 170 million readers come to find insightful and dynamic thinking. Here, expert and undiscovered voices alike dive into the heart of any topic and bring new ideas to the surface. Learn more

Follow the writers, publications, and topics that matter to you, and you’ll see them on your homepage and in your inbox. Explore

If you have a story to tell, knowledge to share, or a perspective to offer — welcome home. It’s easy and free to post your thinking on any topic. Write on Medium

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store