Defining and measuring

public transit service equity

Alex Karner, PhD

Postdoctoral researcher ● CEDEUS

Defining and measuring

public transit service equity

Alex Karner, PhD

Postdoctoral researcher ● CEDEUS

@AlexKarner

Transportation equity What is it?

3

• Freeway revolts, urban

unrest (1960s)

– McCone commission

report

• ISTEA (1991)

– “Planning factors”

• Subsequent guidance,

legislation, etc. Watts, Los Angeles, 1965

Transportation equity What is it? • Transportation equity is a civil and human rights

priority. Access to affordable and reliable transportation widens opportunity and is essential to addressing poverty, unemployment, and other equal opportunity goals such as access to good schools and health care services. However, current transportation spending programs do not equally benefit all communities and populations. And the negative effects of some transportation decisions— such as the disruption of low-income neighborhoods — are broadly felt and have long-lasting effects. Providing equal access to transportation means providing all individuals living in the United States with an equal opportunity to succeed.

4

http://equitycaucus.org/sites/default/files/Testim

ony%20for%20the%20Record%20from%20Equ

ity%20Caucus%20for%20Jan142014House%2

0TandIHear....pdf

…a civil and human rights priority. Access to

affordable and reliable transportation

widens opportunity. Current transportation

spending programs do not equally benefit

all communities and populations. Negative

effects of some transportation decisions are

broadly felt and long-lasting.

The Leadership Council on Civil Rights

http://www.civilrights.org/transportation/

Transportation equity What is it? An illustrative example

• 2004: Minority Citizens Advisory Committee proposes adoption of four EJ principles:

1. Creation of an empowering public process

2. Collection of data to analyze inequities in transportation funding

3. Changing discretionary investments to mitigate such inequities as are found

4. Mitigation of disproportionate project effects prior to being approved for funding

5 Transportation 2035: MTC’s 2009 RTP

Regional equity analysis

• Critical review

– Geographic aggregation

– Future vs. existing equity

– Treatment of race

6

Karner, A. and D. Niemeier (2013). “Civil rights guidance and equity analysis methods for regional

transportation plans: a critical review of literature and practice.” Journal of Transport Geography 33: 126-134.

Rowangould, D., A. Karner and J. London. “Identifying environmental justice communities for transportation

analysis.” Under review at Transportation Research: Part A.

San Francisco County, California

Photo credit:

Paloma Pavel

Regional equity organizing:

Equity, environment and jobs Scenario

Transit service equity

8

9

Why do we have transit?

• Image and aesthetics

• To use federal funds (FTA’s New Starts)

• Economic development (Chatman and Noland, 2013)

• Congestion and air quality mitigation (Anderson, 2014)

• To provide basic mobility for transit dependent populations (Grengs, 2005; Garrett and Taylor, 1999)

Who uses transit?

10

Transit dependents

(bus users)

Choice riders

(commuter rail users)

“Simply put, the bus is the mode of the poor.”

Median income $22,500 $62,500

source: Taylor and Morris, 2015 using 2009 NHTS data

Transit goals in tension

• Rail transit service has expanded faster than bus service over the past 25 years

• Bus patronage declined from 2001-2009 as rail ridership grew

• Bus and rail service and patronage converging over time

• Shift to serving choice riders with premium service

11 sources: Taylor and Morris, 2015; Wells and Thill, 2012; Grengs, 2005

• IT ISN’T JUST ABOUT BUS V. RAIL –

CONSIDER ADDING A SLIDE WITH LIT

REVIEW ABOUT OVERALL DISPARITIES

BY NEIGHBORHOOD, EVEN FORLOCAL

BUS (WELLS AND THILL).

Title VI of the 1964 Civil Rights Act

Legal basis for transit equity

12

“No person in the United States shall, on the

ground of race, color, or national origin, be

excluded from participation in, be denied the

benefits of, or be subjected to discrimination

under any program or activity receiving

Federal financial assistance.”

Title VI of the 1964 Civil Rights Act

13

Fund recipients may not discriminate “with

regard to the routing, scheduling, or quality of

service … furnished” to patrons.

49 CFR §21.5 Appendix C(a)(3)(iii)

14

Equity analysis practice

• Transit agencies evaluate the equity of “major” service changes according to FTA guidance FTA Circular 4702.1B

• Process has been contentious in Los Angeles and the Bay Area

• FTA requires specific data and methods but these may not reflect actual ridership and afford wide agency discretion

15

image source: Metropolitan

Transportation Commission

Oakland Airport Connector

• $500 million project

• Doubles fare, no intermediate stops

• $70 million withdrawn by Federal Transit

Administration for civil rights violations

Local bus service cuts

• AC Transit disproportionately serves riders of color

• 2008-11: 8% fewer service miles, 12% fewer trips, fares

increased 11%

17

April 24, 2012: Metro responds: http://goo.gl/XNMcVx

July, 2010 Metro budget cuts 387,500 bus service hours

Nov., 2010 Bus Riders Union files administrative complaint

with Federal Transit Administration

April 23, 2012 Metro found in violation

April 5, 2013 New Metro service equity analysis

June 27, 2013 Metro found in compliance

Los Angeles

How well do existing methods

work?

18

Study area: Phoenix, Arizona

• 6th largest city in US (1.4 million people)

• 12th largest metropolitan area (4.2 million people)

• Urbanized area increased sevenfold from 1950 – 2000

• 2.4% of workers commute using transit (half the US average rate)

19

Congestion on Interstate 10 in Phoenix

image source: ADOT

20

Rapid

Express Local

Light rail

Typical analysis: 1. Establish service area demographics

• Based on proximity to stops

• Results may not reflect ridership and do not reflect the importance of each line to protected populations

• Academic literature takes a similar approach (REFS) – no systematic analysis of differences

21

total population people of color household

income < $25K

Valley Metro system

demographics (buffers) 1,710,309

891,990 177,640

52% 28%

Valley Metro system ridership 242,687 136,729 122,532

56% 50%

Typical analysis: 2. Establish affected population demographics

22

Route 39 – 40th St. white people of color household income < $25K

77% 23% 17%

3. Compare affected and service area populations white people of color

household income <

$25K

Valley Metro system

demographics (buffers) 48% 52% 28%

Route 39 – 40th St. 77% 23% 17%

77% > 48% and 17% < 28% Potential impact

Typical analysis: limitations

23

• Either census data or ridership can be used

• Most analyses based on demographics proximate to stops

• Academic literature has taken a similar approach (Wu et al., 2003; Minocha et al., 2008; Mavoa et al., 2012;

Al Mamun and Lownes, 2011)

Accessibility-based analysis?

• Accessibility measures the potential to

meet desired needs (Wachs and Kumagai, 1973; Handy and Niemeier, 1997)

• Essential for understanding transportation

system benefits (Martens, 2012; Martens et al. 2012)

• Use to supplement demographic analyses

24

Research questions

1. How consistently do existing methods

characterize the equity of transit-related

decisions?

2. How can new data sources aid with

equity determinations?

– Incorporate accessibility

25

Observed ridership and census

demographics

26

Data and methods

• Map of buffers, with an inset of the census

blocks overlaid

27

• Census demographics

– 2010 SF1 (race)

– 2008-2012 ACS (income)

• Ridership

– 2010-2011 Valley Metro

On-board survey

Karner, A. and A. Golub (In press). “Comparing two common

approaches to public transit service equity evaluation.”

Transportation Research Record.

28

BRT

Rapid

Neighborhood

circulator

Rural

Express

Local

29

BRT

Rapid

Neighborhood

circulator

Rural

• White ridership higher than census on premium modes

• Black ridership higher than census on local modes

• Latino ridership lower than census in all cases

30

31

Modeling results

dependent variable (ridership proportion)

White Black Latino Asian < $25K > $50K

census 2.1 2.1 24.8

N 92 92 92 92 92 92

R2 0.26 0.024 0.33 0.20 0.003 0.005

32

• Census demographics have some

relationship with ridership for some groups

Modeling results

dependent variable (ridership proportion)

White Black Latino Asian < $25K > $50K

census 2.1 2.1 24.8

N 92 92 92 92 92 92

R2 0.26 0.024 0.33 0.20 0.003 0.005

33

dependent variable (ridership proportion)

White Black Latino Asian < $25K > $50K

census 1.89 12.43 1.76 27.6 2.61

total ridership -0.0419 0.0343 0.0303

mean walk score along

route 0.00829 -0.0184 0.0252

premium mode dummy 0.739 -1.479 -0.546 -1.77 1.58

light rail dummy 1.516 -1.418 -1.258 -1.94

N 92 92 92 92 92 92

R2 0.59 0.54 0.47 0.26 0.52 0.68

• Adding quality-of-service variables improves fit

• Relationships differ by racial category

Modeling results

34

Implications for FTA equity analysis Rapid routes example

Comparison populations

white people of

color

household

income < $25K

6,524 7,158 1,396

48% 52% 48%

Reference population (buffers)

48% 52% 28%

35

Comparison populations

white people of

color

household

income < $25K

1,337 434 205

76% 24% 11.5%

Reference population (ridership)

44% 56% 50%

Census demographics Ridership

48% = 48% and 48% > 28%

No impact under service

improvement

76% > 44% and 11.5% < 50%

Potential impact under service

improvement

Conclusions

• The demographic data used (census or

ridership) can affect the conclusions drawn

regarding equity

• FTA considers both sources valid

• Future work to understand when model

results can be more widely applied

36

Equity in accessibility

37

Challenges

• Data on transit accessibility difficult to

acquire

• Based on regional travel demand model

outputs

• Coarse and aggregate, zone-based

• For FTA, need route- and stop-level

information

38

Maricopa Pinal

Maricopa Association of Governments TAZ system

~1 sq. mile

~1/4 sq. mile

~1/10 sq. mile

Data

39

Pedestrian network

OpenStreetMap

Block-level demographics

Longitudinal Employer Household Dynamics

Transit stops

Transit routes

GTFS

Transit schedule

Methods

1. Calculate pedestrian service areas around stops (1/4 mi. bus, 1/2 mi. rail)

2. Develop service area demographics

3. Calculate travel time between all stop pairs (64 minute cutoff, ~95% of observed trips)

1. 2 hour morning peak, 24 random departures (22 GB, ~7 hours on consumer hardware)

2. ESRI network analyst with “Add GTFS to a Network Dataset”

4. Calculate stop- and route-level accessibility

40

• In the prior slide we need a piece about EQUITY.

• Disproportionate burden is typically based on comparing sub-system ridership to systemwide numbers (%s and absolute)

• It might involve looking at the same numbers for middle and higher income groups…yeah, that makes sense, to see how those things change, but we know that lower income people are more transit dependent.

41

Origin

Stopi

Stop1 Stop2 Stopj

Workersw Jobsw

Travel time and geography

General Transit Feed Specification

Accessibility analysis would better

match proposed changes to actual

effects on transit-dependent

populations.

…

Independent from the concentration of minorities / low-income people

Develop an indicator for each stop – showing those that provide high levels

of access. Could be used as a decision support tool, instead of an after-the-

fact analysis

Demographics

Longitudinal Employer-Household Dynamics

Unemployment effects

Block group economic data

Differential transit use rates

Cost per unit of accessibility – would be cheaper for bus.

Jobsw Jobsw

How much more likely are transit

users to be low-income people?

Transit route k

𝐴𝑖𝑤 = 𝐸𝑗

𝑤𝑒−𝛽 𝑡𝑖𝑗

𝑗

Territorial accessibility

𝐴𝑖𝑤 = 𝑊𝑖

𝑤 𝐸𝑗𝑤𝑒−𝛽 𝑡𝑖𝑗

𝑗

Worker-weighted accessibility

Sum over all workers

for overall accessibility

Take mean over stops

on a route for route-

level accessibility

ti1 t12 t2j

Results: total accessibility

42

Territorial accessibility Worker-weighted accessibility

43

Route-level territorial accessibility

44

Compare ACS and

LEHD data on light

rail buffers – are

there more high

income people

there?

Is it the effect of

unemployed inflating

the low-income

numbers?

Route-level worker-weighted accessibility

Scottsdale

Neighborhood

Circulators

Glendale Urban

Shuttles

Scottsdale

Neighborhood

Circulator

Glendale Urban

Shuttles

Rapid routes

45

Implications for FTA analysis

> average low-income riders

worker-weighted route-level accessibility to low-wage jobs

Limitations

• Coarse (and

unchanging) LEHD

thresholds

– Low-wage jobs

definitely low

– Mid-wage jobs less

clear

• No consideration of

unemployed

46

• LEHD also contains race, occupational category

• Updated annually

• Possible to open source the methods to some degree

Opportunities

Conclusions

• Valley Metro routes appear mostly equitable

• New data allow for the development of

refined indicators of public transit accessibility

• Their application in concert with traditional

demographic measures is likely to improve

public transit decision making

47

Contact Alex Karner

alex.karner@gmail.com

http://www.alexkarner.com

@AlexKarner

Acknowledgements Funders

Walton Sustainability Solutions Initiatives

Centro de Desarrollo Sustentable Urbano

California Endowment

Resources Legacy Fund

Sustainable Transportation Center, UC Davis

Colleagues

Deb Niemeier Parisa Fatehi-Weeks

Aaron Golub Lindsay Imai

Jonathan London Richard Marcantonio

Sam Tepperman-Gelfant