New Evidence Links Transit Cuts With Poverty and Unemployment

When bus service was eliminated for five years in Clayton County, in the Atlanta metro area, residents endured substantial increases in poverty and unemployment rates. Especially for low-income people and communities of color, transit is “an essential part of the infrastructure, and for now it’s underprovided in most American cities,” Li says. “Reductions in transit service are definitely harmful to economic outcomes,” says Justin Tyndall.

April 18, 2023 •  Jared Brey


In Brief: Clayton County, Ga., saw poverty and unemployment rise during a five-year period when it had no transit access to Atlanta, according to a new study. The study’s findings show that poor transit access leads to fewer job opportunities and lower incomes for residents.

Researchers say there are strong links between transit access and economic outcomes, with important differences in bus, rail and other transit modes.

In 2010, amid budget pressures stemming from the 2008 housing market crash and ensuing recession, Clayton County, Ga., canceled its bus service. Clayton is a majority-Black county in the southern part of the Atlanta metropolitan area, with a poverty rate of almost 20 percent. It’s not connected to the city by rail, and before 2010, bus service was its only major means of public transit. The buses cost about $10 million a year to run and only collected about a fifth of that amount back in fare revenue, according to a report in the Los Angeles Times.

For the five years following the demise of the service, which was called C-Tran, Clayton County residents had no public transit access to Atlanta. MARTA, the city’s larger transit service, began running buses in the county again in 2015. In the half-decade interim, the county endured “substantial increases in poverty and unemployment rates” which are explained by the loss of bus access, according to new research published last month in the journal Urban Studies.

The cancellation of bus service was a blow to Clayton County. For researchers, however, it was a rare opportunity to study the links between access to public transit and economic outcomes like poverty and unemployment, says Fei Li, an assistant professor in the Urban Studies Institute at Georgia State University and lead author of the paper. The circumstances provided a kind of “natural experiment,” the paper says. It’s often hard to isolate the economic effects of certain events, like the pandemic or natural disasters, because they happen over large areas and affect lots of communities in the same way. But in the case of Clayton County, the researchers were able to compare census tracts that initially had bus access and then lost it with demographically similar tracts that weren’t affected by the cuts.

Li and her co-author, Christopher Kajetan Wyczalkowski, used a “difference-in-difference” method to observe how poverty and unemployment rates changed between 2010 and 2014 in different census tracts. They found that “losing all bus stops in a census tract leads to a 5.1 percentage point increase in the poverty rate and a 4.5 percentage point increase in the unemployment rate,” according to the paper.

The findings explore two divergent but overlapping theories in urban research. One is known as spatial mismatch, which holds that lack of transit access limits access to jobs and puts low-income residents at an extra disadvantage. The other, residential sorting, holds that lower-income households tend to move to areas with better transit access over time. The Clayton County experience suggests that both phenomena could be at play, but shows stronger evidence for the former, spatial mismatch, Li says. While it’s not possible to conclude exactly how the loss of bus access affected individual families, it’s likely that some residents lost access to existing jobs or job opportunities because of the cuts.

Especially for low-income people and communities of color, transit is “an essential part of the infrastructure, and for now it’s underprovided in most American cities,” Li says.

It can be tough to separate the effects of mismatch and sorting trends, but it’s really “a long-term/short-term question,” says Justin Tyndall, an assistant professor in the University of Hawaii Economic Research Organization. In the short term after service cuts, people lose access to jobs and opportunities; but over a much longer timeline, people tend to choose neighborhoods that suit their economic needs and that they can afford, Tyndall says.

Tyndall published a study in 2017 looking at how the temporary loss of the R subway line in New York City after Hurricane Sandy affected poverty and unemployment. His study found similar results to the Clayton County study, though the changes were smaller, possibly because New York generally has more transit options.

“Reductions in transit service are definitely harmful to economic outcomes. In these cases, people have already made location decisions based on the assumption of transit access,” Tyndall says.

The mode of transit service matters though. While poor families tend to live closer to bus service, the effects are reversed with flashier and more expensive services like urban light rail, Tyndall says. Higher-educated and higher-paid workers tend to congregate around light rail-accessible neighborhoods, which raises employment rates — and rents — in those areas, while lowering poverty. But the displacement impacts of light rail amenities can actually reduce overall employment across a wider area, by pushing lower-income people to neighborhoods with worse transit access, Tyndall has found.

In general, there’s strong evidence from a variety of research fields that transit access supports good economic outcomes, and that cuts to transit service are economically harmful. It’s important to highlight evidence of those links as U.S. transit agencies face some of their most dire financial challenges in years, Li says.

“This is a hard time for transit overall,” Li says. “I don’t think we can afford as a country or a region to have public transit fail because of the pandemic. I think that will have really long-term consequences.”


UCLA Policy Briefs

Fare-free? Reduced fares? What research tells us about strategies for pricing public transit 

By King, Hannah and Taylor, Brian D. Feb 2023

Many analysts have argued for transit fares to vary with distance traveled and time of day to better reflect the highly  variable costs of transit service provision on both efficiency  and equity grounds. However, proposals for variable fares  have garnered little traction among transit managers and their governing boards, who often worry that changing  fares may be even less popular with riders than raising them. Until recently, variable fares were also difficult to  implement from a technological standpoint. As a result, most fare experimentation has centered on “fare-free” or  reduced-fare programs. 

Free- and reduced-fare (FAR) programs have most commonly  been targeted at specific groups of riders, like students or  seniors. FAR programs may reduce the costs of collecting  fares. Because they are, essentially, flat fares, FAR policies  limit the ability of operators to charge different fares based  on trip costs rather than traveler characteristics. Even so,  FAR programs are increasingly being touted by advocates  in recognition of transit’s important social service role in  providing mobility to those unable to afford or otherwise access private mobility, such as older adults who may face both physical and financial barriers to automobile use. 

Key Findings 

FAR programs are likely to improve ridership, but sustainable funding for the FAR program must be found.  Identifying sustainable funding is the fundamental  challenge of FAR programs. Ridership increases are likely to be more pronounced on systems with previously high fares and those with higher proportions of low-income riders, and less pronounced on systems with already low fares and/or higher-income riders.  

The net fiscal impact of FAR programs on transit agency  finances is uncertain, particularly with respect to  increased costs that may be occasioned by increased  rider demand. Fully understanding how FAR programs  influence agency finances is a major research challenge but one worth undertaking. Without such information,  recommendations about the wisdom of implementing FAR  programs are necessarily speculative.  

(State level) farebox recovery requirements present a major barrier to the further expansion of FAR programs. FAR programs would almost certainly be much  more common than they are now if transit agencies were not bound by minimum farebox recovery requirements, such as those under California’s Transportation Development Act (TDA). However, eliminating or relaxing farebox recovery requirements would represent a significant move away from a user fee-funded transit system and toward one  that functions more like a park or school, where a baseline level of access is expected for every community member.  Accordingly, FAR programs have the potential to enable  some level of transit access for all.  

Service improvements are likely to be a  more effective use of resources than fare reductions,  even for low-income riders. The vast majority of transit research that compares fare elasticities with service elasticities finds that service elasticities are greater. This implies that, at the margin, increased spending to improve transit service is likely to attract more riders than similar expenditures to make transit cheaper, though there can be exceptions to this general rule.  

FAR programs may generate a host of societal benefits  to the extent that they decrease vehicle use. These  benefits include reducing vehicle miles traveled (VMT)  and associated greenhouse gas emissions. Mode shift and  environmental benefits are likely to be modest, however, because the most price-sensitive riders tend to have less  access to cars and trucks. Again, increased spending on transit service improvements may lead to more of these  benefits than FAR programs. 

For the foreseeable future, transit agencies that reduce the financial barriers to transit access will face challenges  related to holes in the social safety net. By reducing the financial barriers to transit access, FAR programs may risk increasing the presence of individuals engaging in antisocial behavior such as active, in-vehicle use of illicit substances, not maintaining acceptable hygiene standards, and not engaging other riders respectfully. Some transit  agencies, such as LA Metro, San Francisco’s BART, and Philadelphia’s SEPTA, are responding to these challenges by dedicating funding to (1) “transit ambassador” programs designed to both improve the experience of riding transit  and (2) increasing agencies’ abilities to support unhoused individuals and members of other vulnerable rider groups.  

One size does not fit all. If fare-free transit is to be adopted, the cost (in foregone fare revenue) is lower on systems that already recover a relatively small share of their operating costs out of the farebox. Such systems tend to operate in  less transit-friendly environments and carry larger shares of lower income and mobility disadvantaged riders. On systems with higher farebox recovery rates, especially those serving large downtowns, the opportunity cost of fare-free programs is much higher, and such systems often carry proportionally larger shares of non-poor riders. On  these systems, targeted fare-reduction programs aimed at particular rider groups (low-income, students, etc.) are a  less costly way of providing fare reductions to riders who need them most. 

This policy brief is drawn from the report “Considering  Fare-Free Transit in The Context of Research on Transit  Service and Pricing: A Research Synthesis,” prepared by  Hannah King and Brian D. Taylor at the UCLA Institute of  Transportation Studies. The report can be found here:  

For more information about findings presented in this  brief, please contact Hannah King at  Readers interested in learning more about FAR programs  in California can refer to “A Review of Reduced and Free  Transit Fare Programs in California,” prepared by Jean Daniel Saphores, Deep Shah, and Farzana Khatun at the UC  Irvine Institute of Transportation Studies. The report can be  found here: 

Research presented in this policy brief was made possible through funding received by the University of California Institute of Transportation Studies (UC ITS)  from the State of California through the Public Transportation Account and the Road Repair and Accountability Act of 2017 (Senate Bill 1). The UC ITS is a  network of faculty, research and administrative staff, and students dedicated to advancing the state of the art in transportation engineering, planning, and policy Project ID: UC-ITS-2022-08 | DOI: 10.17610/T6WC8Z 


2023  Georgia State University,, Li, Fei, “How buses alleviate unemployment and poverty: Lessons from a natural experiment in Clayton,  GA” (2023). USI Publications. 72. Page 1 of 33 

How buses alleviate unemployment and poverty: Lessons from a natural experiment in  Clayton, GA 


Many studies have documented the linkage between public transportation and economic outcomes,  though there is relatively little empirical evidence on the consequences of losing existing transit  services, especially bus services, which disproportionately serve low-income populations. We  investigate the impacts of bus access on poverty and employment using a natural experiment in  Clayton County, GA, where the local bus transit was terminated between 2010 and 2015. Using a  difference-in-difference approach, we find substantial increases in poverty and unemployment  rates in affected neighbourhoods during the five-year period. Our findings suggest both the spatial  mismatch hypothesis, which predicts the reduction in transit access can lead to reductions in job  accessibility and employment, and the residential sorting hypothesis, which states that poor  households gravitate toward neighbourhoods with better transit access, could be at play. Overall,  we find strong evidence that disruptions in bus transit could have significant adverse impacts on  neighbourhood economic outcomes. Our findings underscore the need for federal and local public  transportation funding to help improve job access, alleviate poverty, and maintain neighbourhood  stability. 


Public transportation, bus transit, poverty, unemployment, neighbourhood change Introduction 

Lack of access to jobs, and the subsequent unemployment or underemployment, can be a major  driver of urban and suburban poverty (Preston and McLafferty, 1999; Gobillon and Selod, 2021). 

Urban Studies

Despite the automobile dependence in most U.S. cities, many still ride transit to work, especially  minority, lower-paid workers who are more susceptible to poverty and unemployment (Burrows  et al., 2021; Clark, 2017). Access to public transportation (or lack thereof), therefore, could have  an important role in expanding access to jobs and alleviating poverty (Grengs, 2010; Lyons and  Ewing, 2021; Sanchez, 2008). 

Many studies have explored the linkage between public transportation and economic outcomes  (Lyons and Ewing, 2021; Pasha et al., 2020; Sari, 2015; Sanchez, 1999; Sanchez, 2008), though  few examined the causal effects of losing existing transit, especially bus services, on  unemployment and poverty. Since transit disruptions caused by natural disasters or public safety  crisis are often accompanied with confounding factors influencing economic activities, it can be  methodologically challenging to evaluate their economic impacts.  

We examine the causal link between bus access, poverty, and unemployment using a natural  experiment in Clayton County, Georgia. Clayton’s local bus service was abruptly terminated in  2010 due to county budget shortfalls (Fausset, 2010), leaving most residents with no transit access  until 2015. Using a difference-in-difference (DID) approach and propensity score matching, we  find that the termination of local buses led to significant increases in poverty and unemployment  rates in affected areas in 2010-2014. Our findings make an important contribution to the literature  and policy debates on the role of public transportation in job access and poverty. 


Bus transit in Clayton County 

Clayton is one of the core counties in the Atlanta metropolitan region, lying adjacent to the city of  Atlanta to the south. The Hartsfield Jackson International Airport, in Clayton’s northwest corner, 

Page 2 of 33 

Page 3 of 33 

Urban Studies

is the southern terminus of Metropolitan Atlanta Rapid Transit Authority (MARTA)’s heavy rail  line, but the rail station is inaccessible from outside the airport. Therefore, buses are the primary  form of public transportation serving the county. A local bus system, C-Tran, served Clayton  County and connected its residents to MARTA and Atlanta from 2001 to 2010. In March 2010, C 

Tran ended all bus routes due to county budget shortfalls (Fausset, 2010). On November 4, 2014,  voters in Clayton County passed a referendum with an 74-26 percent split to join MARTA and  dedicate a one-cent sales tax to expanding MARTA services to the county (Laing, 2014; Karner  and Duckworth, 2019). MARTA started operating bus routes in Clayton in March 2015 (MARTA,  2015). 

Clayton County is a working-class suburb with one of the highest concentrations of racial  minorities in the Atlanta region. Black residents represented 60.6% of the county’s population in  2005-2009 (US Census Bureau, 2021). By the time the county commission voted 4 to 1 to end C Tran service in October 2009, the bus routes had been an essential service for transit dependent  commuters in Clayton County (Fausset, 2010). Prior to the termination in 2010, C-Tran operated  five bus routes with over 2 million trips annually. 4% of the workers in Clayton County commuted  by transit in 2005-2009, the third highest in metro Atlanta following Fulton and DeKalb, the two  counties forming the City of Atlanta (US Census Bureau, 2021). A few private companies  attempted to fill the gap after the termination of C-Tran, often with limited service and higher  fares, though these attempts were largely unsuccessful and short-lived (Rankin, 2012; Martinez,  2011; Feigenbaum, 2014). The five-year gap in local bus service could have had tremendous  impacts on transit riders in Clayton County. 

Figure 1 shows the bus routes in Clayton County in 2007 (C-Tran) and 2016 (MARTA). The  census tract encompassing the airport (shaded area in Figure 1) is the only tract in Clayton with 

Urban Studies

direct access to MARTA rail, though it contains no residential population and is excluded from  our analysis. Six of the 50 census tracts in Clayton are also served by Georgia Regional  Transportation Authority (GRTA) Xpress, a regional commuter coach service throughout the  Atlanta region (Figure 1). Xpress was not affected by the decision to end C-Tran, and these tracts  maintained some regional bus access during 2010-2014. In robustness tests described in Findings,  we test the hypotheses without these tracts. Including or excluding these six tracts has no  substantial effects on our results.  

Please insert Figure 1 here 

A conceptual framework 

We draw from several groups of literature to form two theoretical links between bus access and  poverty or unemployment. The first one, based on the spatial mismatch theory and the transport  justice literature, posits that losing bus access will negatively impact the labour market and  economic outcomes of affected residents. The second link, drawing from the neighbourhood  change literature and the poverty decentralisation hypothesis, suggests that the lack of bus transit  may also lead to lower rates of poverty and unemployment via inter-neighbourhood migration and  residential sorting. 

Spatial mismatch, transport justice, and equitable accessibility. The spatial mismatch theory,  started in the 1960s in the U.S., links inner-city poverty to the separation of people and jobs in  suburbanization (Kain, 1968). Subsequent research extends the model to other countries and  contexts, with emphases on residential segregation and jobs-housing imbalance (Fan et al., 2014;  Gobillon and Selod, 2021; Holzer, 1991; Cervero, 1989). The focus on physical distance or travel  time often assumes commuting by car, thus overlooking the role of public transportation in job  accessibility (Grengs, 2010; Kain, 1992). Some researchers further attribute poor job accessibility 

Page 4 of 33 

Page 5 of 33 

Urban Studies

to lower levels of car ownership and advocate expanding access to automobiles for vulnerable  populations to address the “modal mismatch” (Grengs, 2010; Blumenberg and Manville, 2004;  Taylor and Ong, 1995; Patacchini and Zenou, 2005; Blumenberg and Pierce, 2014). 

While many consider public transportation inadequate in bridging the job access gap, researchers  find linkages between transit access and employment, although more empirical evidence focuses  on rail than bus transit. Sanchez (1999) finds that labour participation rates are negatively  associated with distance from rail or bus stops in Portland and Atlanta. Sari (2015) and Mayer and  Trevien (2017) report evidence from Bordeaux and Paris that new construction and extension of  public transit lead to increased employment. In a natural experiment similar to our study design,  Tyndall (2017) examines the effect of the R Train closure in New York City following Hurricane  Sandy, though the transit disruption was limited to one train route in a transit rich city and a much  shorter time frame. One year after the hurricane, Tyndall (2017) finds significant increase in  unemployment rates along the R Train, suggesting the R Train closure had an adverse impact on  commuters and job searchers in these neighbourhoods. An inter-city cross-sectional analysis by  Lyons and Ewing (2021) argues that transit may affect unemployment and poverty indirectly by  supporting compact urban forms. 

A broader literature on transport justice further establishes the importance of transit in ensuring  equitable accessibility, including access to economic and employment opportunities (Tiznado Aitken et al., 2021; Adli and Chowdhury, 2021; Liu and Kwan, 2020). Jiao and Dillivan (2013)  use “transit deserts” to describe areas where transit services do not meet transit demand. Cai et al.  (2020) find 11% of low-income commuters in Wuhan, China live in transit deserts. In Atlanta,  Wyczalkowski et al. (2020) find low transit connectivity in poor, minority concentrated 

Urban Studies

neighbourhoods. The reimplementation of bus transit in Clayton County was a public response to  transit inequity and move towards transit justice (Karner and Duckworth, 2019).  

The residential sorting and poverty decentralisation hypotheses. Transit has long been seen as a  potential driver of rising property values (Stokenberga, 2014; Bowes and Ihlanfeldt, 2001; Hess,  2005; Grass, 1992) and gentrification (Dawkins and Moeckel, 2016; Grube-Cavers and Patterson,  2015; Baker and Lee, 2019), which may lead to the displacement of low-income residents. As in  the spatial mismatch literature, more attention has been paid to rail and bus rapid transit (BRT)  than regular bus, although the latter forms the bulk of public transit systems in the U.S. (APTA,  2019). The few studies finding a positive linkage between bus transit and property values are  largely from outside the U.S. (Ibeas et al., 2012; Yang et al., 2019), likely reflecting the higher  share of racial minorities among American bus riders and the related stigma and poor service  quality (Clark, 2017). Taking service frequency into account, Pasha et al. (2020) find increased  bus access is associated with increased employment as well as poverty but not increased rental  values in Cuyahoga County, Ohio, suggesting buses have not caused displacement of poverty. 

Other studies showing a positive association between transit and poverty point to the importance  of transit access in poor households’ residential choices (Brueckner and Rosenthal, 2009; Glaeser  et al., 2008). Glaeser et al. (2008) argue that the concentration of poverty in cities is largely  attributable to better transit access. Likewise, recent research on suburbanization of poverty  suggests the expansion of transit may have played a role in housing decisions of the poor. Wang  and Woo (2017) report increasing racial and income diversity in suburban transit rich  neighbourhoods in metro Atlanta. Moreover, they find a stronger association between poverty and  transit ridership in suburban transit rich neighbourhoods than those in downtown or inner-city.  Pathak et al. (2017) examine the longitudinal changes in suburban poverty in metro Atlanta in 

Page 6 of 33 

Page 7 of 33 

Urban Studies

relation to the expansion of bus transit, concluding that growing transit coverage has expanded the  housing options of the poor and facilitated the decentralisation of poverty. 

These theoretical models have various implications on the termination of C-Tran in Clayton  County. On one hand, the spatial mismatch hypothesis predicts increased poverty and  unemployment due to diminished job accessibility, implying that low-income workers have low  residential mobility due to the limited availability of affordable housing. The poverty  decentralisation hypothesis, on the other hand, suggests decreased poverty due to residential  sorting. The two processes could work in tandem, with the transit disruption first worsening the  economic outcomes of transit riders in affected neighbourhoods and then forcing some of them to  relocate to areas with better transit access. 


The difference-in-difference (DID) approach 

The termination of C-Tran, caused by fiscal deficits arguably exogenous to the public transit  system and with contained effects on Clayton County residents, constitutes an ideal setting for a  quasi-experimental examination of the impacts of losing bus transit on poverty and unemployment  rates. If all other factors were held equal, we can expect poverty and unemployment to increase or  decrease in Clayton after the bus termination in 2010 and then regress to their baseline levels  following the reinstatement of bus routes in 2015. However, those “other factors” rarely stay  constant in real-world settings and could confound our analysis. Notably, the study period of 2005- 2019 coincides with the Great Recession and subsequent economic recovery. Even if the transit  disruption had no effects, we can still expect to see poverty and unemployment rates rise in 2010  and subside after 2015. It is therefore important to situate any observed changes in the  macroeconomic environment to isolate the transit impacts. 

Urban Studies

The difference-in-difference (DID) approach addresses this issue and identifies causal effects by  comparing observed temporal changes in “treated” versus “control” areas (for a more detailed  discussion of DID, see Lechner, 2011). Specifically, in this study “treated” areas are census tracts  in Clayton County that lost bus access in 2010, and “control” areas are census tracts from the rest  of the Atlanta region that were not affected by the termination of C-Train but otherwise similar to  the treated tracts. We compare poverty and unemployment rates in three five-year periods: 

∙ Period 1 (P1): 2005-2009, when C-Tran operated in Clayton. 

∙ Period 2 (P2): 2010-2014, during which Clayton had no local public bus services while  other counties only experienced marginal or incremental changes in transit coverage. ∙ Period 3 (P3): 2015-2019, when MARTA implemented new bus routes in Clayton. 

The effects of the termination of C-Tran will be observed in P2.  

Figure 2 demonstrates the DID approach in three hypothetical scenarios. In Figure 2(a), with no  linkages between transit access and unemployment or poverty, both treated and control tracts  follow the same macroeconomic trend in three time periods. Figure 2(b) illustrates the scenario  where the spatial mismatch theory was dominant, i.e., the termination of bus services led to an  increase in poverty or unemployment in P2 in addition to that caused by the Great Recession (as  seen in control tracts). The widened gap between treated and control tracts in P2 represented the  effects of the transit disruption. In Figure 2(c), residential sorting was a stronger force, and treated  tracts saw a smaller increase (or even decrease) in poverty or unemployment as compared to  control tracts. In all three scenarios, controlling for macroeconomic trends using the DID approach  helps identify the impacts of the transit disruption. 

Please insert Figure 2 here 

Page 8 of 33 

Page 9 of 33 

The study area 

Urban Studies

Following the Atlanta Regional Commission1, we define the Atlanta region as the 11-county area  including Cherokee, Clayton, Cobb, DeKalb, Douglas, Fayette, Forsyth, Fulton, Gwinnett, Henry,  and Rockdale counties. In addition to the 2007 C-Tran map, we obtained transit maps for the 11- county Atlanta region in 2010 and 2016, including MARTA, GRTA Xpress, and local transit  services in other counties, such as Cherokee Area Transportation System (CATS), CobbLinc or  Cobb Community Transit (CCT), and Gwinnett County Transit (GCT). Since no large-scale  expansion or termination of transit services occurred outside Clayton County between 2005 and  2014, we use the 2010 maps to derive bus access measures for both P1 and P2.  

We develop two measures for bus access: 1) a binary indicator, HASSTOP, with 1 indicating at  least one bus stop within the census tract and 0 indicating none, and 2) a continuous variable,  PCTBUS, measuring the percentage of land area in a census tract within a 0.5-mile buffer from  any bus stop. The continuous measure complements the binary indicator and allows us to  differentiate tracts with extensive bus coverage from those with minimal coverage. The Atlanta  region has limited rail coverage only in Fulton and DeKalb, with rail stations and surrounding  neighbourhoods all well served by buses. Therefore, our bus access measures would also capture  rail transit access where it exists. 

We are primarily interested in areas affected by the termination of C-Tran, so we exclude all census  tracts outside Clayton County that experienced substantial changes in bus access between 2010  and 2016 from the panel dataset. We define a substantial change as either 1) having no bus stops  in 2010 and one or more bus stops in 2016, or 2) having one or more bus stops in 2010 and no bus  


Urban Studies

stops in 2016. In other words, HASSTOP is time invariant in all control tracts in our panel dataset.  We further use propensity score matching to select control tracts, as explained in the following  section. 

Propensity score matching 

Propensity score matching (Rosenbaum and Rubin, 1985) is a statistical technique widely used in  quasi-experimental studies for constructing control groups comparable to the treatment group.  Matching is based on the “propensity” for treatment as predicted by a set of covariates, thus  balancing these covariates between the treated and control subjects. We create two matching  scenarios: one defines all census tracts in Clayton County as treated, and the other defines only  Clayton tracts with one or more bus stops (i.e., HASSTOP = 1) in P1 as treated. Table 1 shows the  matching variables, including the two dependent variables, the two measures of bus access, and a  number of control variables. All variables were measured in P1 for matching purposes. 

Please insert Table 1 here 

For each of the two scenarios, we use nearest neighbour matching to create three matched samples,  containing one, three, and five matches for each treated tract. Nearest neighbour is a matching  algorithm that identifies the closest control unit(s) in propensity scores (based on the matching  variables in Table 1) for each treated unit. Table 2 compares the means of all matching variables  between the treated and control tracts in P1, as well as all untreated tracts in the Atlanta region, for  each of the six matched samples. As Table 2 shows, propensity score matching significantly  improves the comparability between treated and control tracts, especially in the two 1:1 matching  samples. Figure 3 shows the locations of these matches in relation to the treated tracts, as well as  tracts excluded due to substantial changes in bus access. Most of the matched controls are 

Page 10 of 33 

Page 11 of 33 

Urban Studies

geographically close to the treated tracts or in similar suburban locations of the Atlanta region,  although geographical proximity is not considered in the matching process. 

Please insert Table 2 here 

Please insert Figure 3 here 

Model Specification 

Equation (1) shows our main model specification: 

�� (1) ���� = ������������ + �������� + ���� + ������ 

where is the outcome variable ( �� POVERTY or UNEMPLOYMENT) in census tract i and period ���� t, represents one of the two measures of bus access in tract ������ i and period t, is a vector ���� ������ containing all control variables in Table 1, and is a full set of census tract fixed effects so we ���� can isolate temporal changes in poverty and unemployment within tracts. The parameter captures �� the association between bus access and poverty or unemployment. A positive will indicate that �� the termination of C-Tran has led to reduced poverty and unemployment, and a negative suggests �� the opposite. 

To test the residential sorting hypothesis more directly, we use a slightly modified set of models  to examine neighbourhood change and residential mobility in the affected tracts. These models use  the following dependent variables: 

∙ Total population (natural log); 

∙ White population (natural log) 

∙ Black population (natural log); 

∙ % population that moved into their current units (at the time of survey) within five years; 

Urban Studies

∙ % population that moved within the same MSA in the past 12 months; ∙ % population under poverty that moved in the past 12 months. 

The first three variables test socio-demographic changes as related to the termination of C-Tran.  We use population counts instead of percentages to examine population growth and inter neighbourhood migration. Unlike poverty or unemployment status, race typically does not change  over time, so changes in the number of whites or Blacks in five years would be a stronger indicator  of residential sorting than changes in poverty or unemployment rates. The last three variables test  residential mobility. The model takes a similar form to Equation 1:  

�� (2) ���� = ������������ + �������� + ���� + ������ 

where is the one of the neighbourhood change or residential mobility variables, and is a ������ ������ reduced set of control variables, including only housing costs (LNRENT and LNVALUE) and age  (H30OLD) as factors of residential choices. According to the theory that the poor and  disadvantaged gravitate to transit rich neighbourhoods, we should see positive estimates of for �� Blacks and negative estimates for whites. If, on the contrary, the termination of C-Tran has  diminished the desirability and economic vitality of the affected neighbourhoods, we expect to see  the opposite signs in . �� 

The three residential mobility measures have inherent limitations due to the design of the American  Community Survey (ACS). None of them distinguish movements within census tracts from those  between census tracts, so increased mobility does not necessarily suggest population inflows from  other tracts. Moreover, as the ACS estimates only describe people living in a census tract at the  time of the survey, these measures cannot capture people who have moved out of the tract. Still, if 

Page 12 of 33 

Page 13 of 33 

Urban Studies

the disruption in bus access has led to substantial socio-demographic changes through migration,  we expect to see more movements in affected neighbourhoods, or a negative in those models. �� 


No matching 

Table 3 presents the results of Model 1 without matching, i.e., using all census tracts in the Atlanta  region other than those excluded in Figure 3. All four models show a strong negative relationship  between bus access and the dependent variables. Specifically, losing all bus stops in a census tract  leads to a 5.1 percentage point increase in the poverty rate and a 4.5 percentage point increase in  the unemployment rate. These are significant effect sizes, considering the baseline levels of  poverty and unemployment in Clayton were 14.9% and 10.9%, respectively, in P1. Most of the  control variables are also strong predictors of poverty and/or unemployment. Overall, these results  lend strong support to the spatial mismatch or job access hypothesis. 

Please insert Table 3 here 

Propensity score matching 

We then proceed to test the four models in Table 3 using each of the six matched samples illustrated  in Table 2 and Figure 3, generating 24 sets of estimates. Figure 4 summarises the key coefficients  in these models along with the 95% confidence intervals.2 Bus access consistently demonstrates a  negative association with poverty and unemployment in the six matched samples, with an average  effect size of 4.6 percentage point increase in the poverty rate and 3.7 percentage point increase in  the unemployment rate in tracts affected by the termination of C-Tran.  

2 Full modelling results available upon request. 

Urban Studies

Please insert Figure 4 here 

Clayton-only analysis 

A second strategy of accounting for potential confounding factors, or systematic differences  between treated and untreated tracts, is to limit the analysis to Clayton County only. The Clayton only analysis considers only those tracts with bus access in P1 as treated, and tracts within Clayton  that either never had or maintained (via GRTA Xpress) bus access from P1 through P3 as controls.  While these control tracts may be ranked lower in propensity score matching, this analysis allows  us to rule out the possibility of unobserved factors unique to Clayton County influencing the  previously reported results. We run the same four models for the Clayton-only sample, including  all census tracts in Clayton except the airport and one with missing LNVALUE. Considering the  smaller number of tracts (37 “treated” and 11 control tracts) in this sample, we test the models  both with and without census tract fixed effects, with standard errors clustered by tract in all  models. Table 4 presents the results.  

The relationship between bus access and poverty or unemployment persists within Clayton  County, suggesting the previous findings are not driven by county-specific factors. Interestingly,  the last two models in Table 4 show counterintuitive relationships between the unemployment rate  and some of the control variables. The percentage of non-white population (NONWHITE) and 25+  adults without a high school diploma (HSDROP), often linked with higher unemployment  (including in Table 3), are negatively associated with unemployment rates in these models. These  associations disappear or become statistically insignificant in models without census tract fixed  effects (Models 5 and 6 in Table 4), indicating they are largely driven by temporal changes within  census tracts, not cross-sectional variations between census tracts. In other words, decreases in  NONWHITE or HSDROP had concurred with increases in unemployment rates in certain census 

Page 14 of 33 

Page 15 of 33 

Urban Studies

tracts. These observations, in addition to the strong evidence suggesting the termination of C-Tran  caused higher poverty and unemployment, lead us to further investigate the socio-demographic  changes and residential mobility in treated tracts. 

Please insert Table 4 here 

Socio-demographic changes and residential mobility 

Figure 5 illustrates key results from models on socio-demographic changes and mobility.3 We do  not find strong evidence that losing bus access is associated with a decline or slower growth of the  overall population (the top left panel of Figure 5), with only one coefficient significant at the 0.05  level and one marginally significant (p = 0.081). This is further corroborated in the residential  mobility models for the population as a while (the top and middle right panels of Figure 5), which  show no effects of the transit disruption on overall residential mobility in the past five years or 12  months. The two race models (the middle and bottom left panels of Figure 5), however, paint a  different picture: reduced bus access, especially the continuous measure of bus coverage, is  associated with a net increase of whites and a net decrease of Blacks in a census tract. These  findings are consistent with the residential sorting hypothesis (Pathak et al., 2017; Glaeser et al.,  2008), which nevertheless runs counter to our findings regarding poverty and unemployment. 

Interestingly, the residential mobility among those living under poverty (% moved within the past  12 months) seems to have slightly increased in census tracts that lost bus access (the bottom right  panel of Figure 5). More frequent moving among the poor could be evidence for the residential  sorting hypothesis, corroborate the racial changes discussed above or the counterintuitive  associations between NONWHITE or HSDROP and UNEMPLOYMENT in Table 4. Specifically,  

3 Full modelling results available upon request. 

Urban Studies

the termination of C-Tran could have led to increased housing instability for the poor, including  an exodus of Blacks and those with lower levels of educational attainment, as poverty and  unemployment rates rose due to worsening economic outcomes of the low-income residents who  stayed in the neighbourhood, creating a spurious positive association between the two processes.  Nevertheless, as aforementioned, the residential mobility models should be interpreted with  caution, since these measures do not differentiate movements within the census tract or from other  census tracts, nor can they reflect displacement from the treated tracts.  

Please insert Figure 5 here 

Robustness tests 

To further test the robustness of our results, we exclude the six census tracts in Clayton County  that were served by GRTA and therefore maintained regional bus services during P2, as described  in the Background section, leaving only those served by local bus routes as “treated” tracts. A new  set of matched samples is generated for this subset of treated tracts. In regression analyses like  those presented in Figure 4, 23 of the 24 models show significant negative relationships between  bus access and poverty or unemployment rates, with a consistent effect size averaging -0.047 on  POVERTY and -0.039 on UNEMPLOYMENT. We also test the models with spatial autocorrelation  in dependent variables or residuals. Neither spatial lag nor spatial error models report significant  spatial autocorrelation, with key estimates consistent with those reported.4 

Discussion and Conclusion 

Public transit provides an essential service for lower income individuals to access jobs and other  opportunities. Even in low-density, automobile-dependent American cities and suburbs, better  

4 All modelling results not fully presented here are available upon request. 

Page 16 of 33 

Page 17 of 33 

Urban Studies

transit coverage expands housing opportunities for transit riders and mitigates residential  segregation. To date, more research attention has been focused on rail and Bus Rapid Transit  (BRT), although in the U.S., bus systems serve the vast majority of low-income transit riders  (APTA, 2019). Bus transit, meanwhile, is more vulnerable than rail to disruptions and service cuts  due to financial, environmental, and public health crises such as the COVID-19 pandemic. To  better understand the implications of bus access and disruptions in bus services on poverty,  unemployment, and neighbourhood change, we examine a natural experiment in Clayton County,  Georgia, where the local bus service was terminated between 2010 and 2015.  

Using a difference-in-difference approach and propensity score matching, we find that the  termination of local buses led to substantial increases in poverty and unemployment in affected  neighbourhoods, supporting the spatial mismatch theory that lack of public transportation can limit  access to jobs, especially for low-income workers. Most of our models suggest that the termination  of local buses had a greater impact on the poverty rate than on the unemployment rate, indicating  that the economic impact of transit disruptions could go beyond those workers who lose their jobs  – for example, to their dependents or those who have to work fewer hours or accept lower pay due  to longer commute or reduced employment options. These findings are robust to alternative  measures of bus access and matching schemes. While some of the census tracts retained regional  bus services during this period, excluding these tracts does not qualitatively change our conclusion. 

We find that the termination of local buses also led to an increase in the number of whites and a  decrease in the number of Blacks. As poverty and employment status can and do change over time,  we consider racial composition a stronger indicator of inter-neighbourhood migration. These  findings are consistent with the residential sorting hypothesis that racial minorities and  disadvantaged groups tend to live closer to transit (Pathak et al., 2017; Glaeser et al., 2008), though 

Urban Studies

it would not explain the strong linkage between bus access and poverty or unemployment we find  in the opposite direction. We also observe an increase in residential mobility among the poor with  no significant changes among the population as a whole, although the residential mobility  measures do not capture emigration from the neighbourhood and could simply indicate increased  housing instability for the poor, which can be an additional adverse effect of the transit disruption.  

Our findings suggest that both theoretical models on the relationship between transit and poverty  or employment could be at play. While transit access can be an important factor in low-income  households’ residential choices, losing existing bus service predominantly impacts transit dependent riders in the area, leading to net increases in poverty and unemployment, at least in the  relatively short term of a five-year period. Moreover, the potential exodus of low-income  households in seeking of better transit access, which is not captured in our analysis, would only  add to the adverse consequences of the transit disruption in Clayton County. While suburban  communities in Atlanta often oppose public transit for its connection with the poor (Givens, 2017),  the termination of C-Tran in Clayton County had most definitely not attracted enough affluent  residents to offset the negative economic impacts on affected neighbourhoods. 

This study uses strong quasi-experimental design to provide solid evidence on the causal effects  of bus access on poverty and unemployment. The significance of bus transit in job access and  poverty has important policy implications. On one hand, it calls for federal and state investments  and improvements in bus transit to reduce the accessibility gap. On the other hand, it highlights  the risks of transit disruptions for low-income communities. The widespread transit disruptions  and service cuts during the COVID-19 pandemic, for example, likely exacerbated the economic  impacts of the pandemic on vulnerable populations. Maintaining consistent public transit access,  particularly for low-income communities, should be a policy priority in future disasters and crises. 

Page 18 of 33 

Page 19 of 33 

Urban Studies


We thank the Metropolitan Atlanta Rapid Transit Authority (MARTA) and the Atlanta Regional  Commission (ARC) for providing historic maps of bus routes and stops. 

Conflict of interest 

Fei Li declares no conflicts of interest. Christopher Wyczalkowski is an employee of the  Metropolitan Atlanta Rapid Transit Authority (MARTA), the public transit authority in metro  Atlanta. Other than providing bus transit maps, MARTA has no direct involvement in the research.  All views and opinions are our own. 


Adli SN and Chowdhury S (2021) A Critical Review of Social Justice Theories in Public Transit  Planning. Sustainability 13: 4289. 

APTA (2019) 2019 Public Transportation Fact Book. American Public Transportation  Association

Baker DM and Lee B (2019) How does light rail transit (LRT) impact gentrification? Evidence  from fourteen US urbanized areas. Journal of Planning Education and Research 39: 35- 49. 

Blumenberg E and Manville M (2004) Beyond the spatial mismatch: Welfare recipients and  transportation policy. Journal of planning literature 19: 182-205. 

Blumenberg E and Pierce G (2014) A driving factor in mobility? Transportation’s role in  connecting subsidized housing and employment outcomes in the moving to opportunity  (MTO) program. Journal of the American Planning Association 80: 52-66. 

Bowes DR and Ihlanfeldt KR (2001) Identifying the Impacts of Rail Transit Stations on Residential  Property Values. Journal of Urban Economics 50: 1-25. 

Urban Studies

Brueckner JK and Rosenthal SS (2009) Gentrification and neighborhood housing cycles: will  America’s future downtowns be rich? The review of economics and statistics 91: 725-743. Burrows M, Burd C and McKenzie B (2021) Commuting by Public Transportation in the United  States: 2019. American Community Survey Reports. U.S. Census Bureau. 

Cai M, Jiao J, Luo M, et al. (2020) Identifying transit deserts for low-income commuters in Wuhan  Metropolitan Area, China. Transportation Research Part D: Transport and Environment 82: 102292. 

Cervero R (1989) Jobs-housing balancing and regional mobility. Journal of the American Planning  Association 55: 136-150. 

Clark HM (2017) Who rides public transportation. American Public Transportation Association  (APTA). 

Dawkins C and Moeckel R (2016) Transit-induced gentrification: Who will stay, and who will go?  Housing Policy Debate 26: 801-818. 

Fan Y, Allen R and Sun T (2014) Spatial mismatch in Beijing, China: Implications of job  accessibility for Chinese low-wage workers. Habitat International 44: 202-210. Fausset R (2010) Clayton County loses vital bus service, link to Atlanta. Los Angeles Times.  (Accessed Jan 14, 2022) 

Feigenbaum B (2014) U.S. Regulations Limit Private Transit Services. Reason Foundation. Givens D (2017) Anti-Transit Design in the Atlanta Suburbs: Aiming for Exclusion, and Failing. Glaeser EL, Kahn ME and Rappaport J (2008) Why do the poor live in cities? The role of public  transportation. Journal of Urban Economics 63: 1-24. 

Gobillon L and Selod H (2021) Spatial mismatch, poverty, and vulnerable populations. Handbook  of regional science: 573-588. 

Page 20 of 33 

Page 21 of 33 

Urban Studies

Grass RG (1992) The estimation of residential property values around transit station sites in  Washington, DC. Journal of Economics and Finance 16: 139-146. 

Grengs J (2010) Job accessibility and the modal mismatch in Detroit. Journal of Transport  Geography 18: 42-54. 

Grube-Cavers A and Patterson Z (2015) Urban rapid rail transit and gentrification in Canadian  urban centres: A survival analysis approach. Urban Studies 52: 178-194. 

Hess DB (2005) Access to employment for adults in poverty in the Buffalo-Niagara region. Urban  Studies 42: 1177-1200. 

Holzer HJ (1991) The spatial mismatch hypothesis: What has the evidence shown? Urban Studies 28: 105-122. 

Ibeas Á, Cordera R, Dell’Olio L, et al. (2012) Modelling transport and real-estate values  interactions in urban systems. Journal of Transport Geography 24: 370-382. Jiao J and Dillivan M (2013) Transit deserts: The gap between demand and supply. Journal of  Public Transportation 16: 2. 

Kain JF (1968) Housing segregation, negro employment, and metropolitan decentralization. The  Quarterly Journal of Economics 82: 175-197. 

Kain JF (1992) The spatial mismatch hypothesis: three decades later. Housing Policy Debate 3:  371-460. 

Karner A and Duckworth R (2019) ‘Pray for transit’: Seeking transportation justice in metropolitan  Atlanta. Urban Studies 56: 1882-1900. 

Laing K (2014) Atlanta transit expansion overwhelmingly approved. The Hill. (accessed March 1,  2022). 

Urban Studies

Lechner M (2011) The estimation of causal effects by difference-in-difference methods.  Foundations and Trends® in Econometrics 4: 165-224. 

Liu D and Kwan M-P (2020) Measuring spatial mismatch and job access inequity based on transit based job accessibility for poor job seekers. Travel Behaviour and Society 19: 184-193. Lyons T and Ewing R (2021) Does transit moderate spatial mismatch? The effects of transit and  compactness on regional economic outcomes. Cities 113: 103160. 

MARTA (2015) 2015 MARTA year in review. Available at: (Accessed Feb 22, 2022) Martinez L (2011) The Impact of Transit on Property Values: A Case Study of the Elimination of  C-Tran in Clayton County, Georgia. Atlanta, GA: Georgia Institute of Technology. Mayer T and Trevien C (2017) The impact of urban public transportation evidence from the Paris  region. Journal of Urban Economics 102: 1-21. 

Pasha O, Wyczalkowski C, Sohrabian D, et al. (2020) Transit effects on poverty, employment, and  rent in Cuyahoga County, Ohio. Transport Policy 88: 33-41. 

Patacchini E and Zenou Y (2005) Spatial mismatch, transport mode and search decisions in  England. Journal of Urban Economics 58: 62-90. 

Pathak R, Wyczalkowski CK and Huang X (2017) Public transit access and the changing spatial  distribution of poverty. Regional Science and Urban Economics 66: 198-212. Preston V and McLafferty S (1999) Spatial mismatch research in the 1990s: progress and potential.  Papers in regional science 78: 387-402. 

Rankin NR (2012) Transportation: The regulation and funding of Metropolitan Atlanta Rapid  Transit Authority (MARTA) and transportation equity. Department of Political Science. Atlanta, GA: Clark Atlanta University. 

Page 22 of 33 

Page 23 of 33 

Urban Studies

Rosenbaum PR and Rubin DB (1985) Constructing a control group using multivariate matched  sampling methods that incorporate the propensity score. The American Statistician 39: 33- 38. 

Sanchez TW (1999) The connection between public transit and employment: The cases of Portland  and Atlanta. Journal of the American Planning Association 65: 284-296. 

Sanchez TW (2008) Poverty, policy, and public transportation. Transportation Research Part A:  Policy and Practice 42: 833-841. 

Sari F (2015) Public transit and labor market outcomes: Analysis of the connections in the French  agglomeration of Bordeaux. Transportation Research Part A: Policy and Practice 78: 231- 251. 

Stokenberga A (2014) Does bus rapid transit influence urban land development and property  values: A review of the literature. Transport Reviews 34: 276-296. 

Taylor BD and Ong PM (1995) Spatial mismatch or automobile mismatch? An examination of  race, residence and commuting in US metropolitan areas. Urban Studies 32: 1453-1473. Tiznado-Aitken I, Muñoz JC and Hurtubia R (2021) Public transport accessibility accounting for  level of service and competition for urban opportunities: an equity analysis for education  in Santiago de Chile. Journal of Transport Geography 90: 102919. 

Tyndall J (2017) Waiting for the R train: Public transportation and employment. Urban Studies 54: 520-537. 

US Census Bureau (2021) American Community Survey 5-Year Data 2009. Wang K and Woo M (2017) The relationship between transit rich neighborhoods and transit  ridership: Evidence from the decentralization of poverty. Applied Geography 86: 183-196. 

Urban Studies

Wyczalkowski CK, Welch T and Pasha O (2020) Inequities of Transit Access: The Case of Atlanta,  GA. JCULP 4: 654. 

Yang L, Zhou J, Shyr OF, et al. (2019) Does bus accessibility affect property prices? Cities 84:  56-65. 

Page 24 of 33 

Page 25 of 33 

Table 1. Variables definition. 

Urban Studies

Matching Variables Description Dependent POVERTY (%) Poverty rate 

Variables UNEMPLOYMENT (%) Unemployment rate 

Independent HASSTOP (%) Whether there is any bus stop within the census tract (Bus Access) PCTBUS (%) % of the census tract within a 1/2 mile radius from a  


bus stop 

POPDEN (‘000s) ‘000s of people per square mile 

NONWHITE (%) % non-white population 

HSDROP (%) % population 25 and over with less than high school  

Control  Variables+ 


COLLEGE (%) % population 25 and over with a bachelor’s or higher  degree 

LNRENT the natural log of median rent (2019 dollars) LNVALUE the natural log of median housing value (2019 dollars) H30OLD (%) % housing units built 30 or more years earlier++ 

+ All control variables are derived from American Community Survey (ACS) 5-year estimates. The 2005- 2009 estimates (P1), based on 2000 census tracts, are reweighted to 2010 census tracts for consistency  with the other two periods using the relationship file from the Census Bureau.1 

++ Age of the housing stock is included as a factor of income-based sorting, as suggested by Brueckner  and Rosenthal (2009) and Pathak et al. (2017). 


Urban Studies

Table 2. Key variables for the “treated” tracts, all other tracts in the Atlanta region, and matched  tracts in P1 (2005-2009). 

(a) All Clayton County tracts as “treated” 

VAR Clayton County Nearest Neighbour Matches All Other  

Page 26 of 33 

Census Tracts 

Tracts 1:1 3:1 5:1 

N 49 673 49 147 245 POVERTY (%) 14.9 13.4 15.8 16.2 15.7 UNEMPLOYMENT (%) 10.9 8.4*** 10.8 11.2 10.9 HASSTOP (%) 67.3 61.1 67.3 67.3 62.9 PCTBUS (%) 41.0 47.1 42.5 44.6 45.2 POPDEN (‘000s) 2.6 3.0 2.4 2.6 2.6 NONWHITE (%) 75.5 45.2*** 73.8 68.8** 65.7*** HSDROP (%) 17.0 12.6*** 16.0 17.4 16.5 COLLEGE (%) 16.9 38.4*** 19.5* 20.7*** 22.7*** LNRENT 7.0 7.1** 7.0 7.0 7.0 LNVALUE 11.9 12.5*** 12.0 12.0*** 12.1*** H30OLD (%) 45.4 39.2* 42.6 42.6 41.7 (b) Clayton County tracts with HASSTOP = 1 in P1 (2005-2009) as “treated” 

VAR Treated Tracts All Other Nearest Neighbour Matches Tracts 1:1 3:1 5:1 

N 33 689 33 99 165 POVERTY (%) 16.3 13.3** 20.4* 18.9 19.6** UNEMPLOYMENT (%) 11.3 8.4*** 12.5 12.2 12.4 HASSTOP (%) 100.0 59.6*** 100.0 100.0 100.0 PCTBUS (%) 58.6 46.1*** 62.2 65.8 71.2** POPDEN (00s) 47.2 52.1 2.9 3.0 3.3* NONWHITE (%) 77.9 45.8*** 79.5 77.7 76.9 HSDROP (%) 18.0 12.7*** 19.7 19.4 18.9 COLLEGE (%) 16.1 37.9*** 16.0 19.1** 20.8*** LNRENT 7.0 7.1*** 6.9 7.0 7.0 LNVALUE 11.9 12.4*** 11.9 12.0*** 12.0*** H30OLD (%) 50.6 39.1*** 54.9 49.9 52.4 Two tailed t-test between treated and matched/other tracts: ***: p<0.01; **: p<0.05; *: p<0.1. 

Page 27 of 33 

Urban Studies

Table 3. Poverty, unemployment, and bus access in Atlanta region (2005-2019). Model 1 Model 2 Model 3 Model 4 


HASSTOP -0.051***  


PCTBUS -0.064***  (0.019) 





POPDEN 0.003  (0.003) 

NONWHITE 0.075***  (0.021) 

HSDROP 0.216***  (0.059) 

COLLEGE -0.101***  (0.031) 

LNRENT -0.038***  (0.011) 

LNVALUE -0.086***  (0.008) 

H30OLD -0.011  (0.010) 



0.075***  (0.021) 

0.217***  (0.059) 

-0.098***  (0.032) 

-0.038***  (0.011) 

-0.087***  (0.008) 



-0.003*  (0.002) 





-0.128***  (0.024) 



-0.041***  (0.007) 

-0.099***  (0.008) 

-0.003*  (0.002) 





-0.125***  (0.024) 



-0.042***  (0.007) 

-0.098***  (0.008) 

N 2112 2112 2112 2112 R2 0.236 0.234 0.239 0.237 

***: p<0.01; **: p<0.05; *: p<0.1. All models include census tract fixed effects. Figures in  parentheses are cluster-robust standard errors. 

Urban Studies

Table 4. Poverty, unemployment, and bus access in Clayton County (2005-2019). Model 1 2 3 4 5 6 7 8 DV POVERTY RATE UNEMPLOYMENT RATE 

Page 28 of 33 

HASSTOP -0.043***  


PCTBUS -0.071***  (0.021) 













POPDEN 0.002  (0.008) 

NONWHITE 0.152**  (0.074) 

HSDROP 0.161  (0.125) 

COLLEGE -0.172  (0.116) 

LNRENT -0.215***  (0.040) 

LNVALUE -0.047*  (0.026) 

H30OLD 0.031  (0.030) 

Constant 2.121***  (0.343) 

Census tract  



0.152**  (0.073) 



-0.216*  (0.125) 

-0.232***  (0.044) 

-0.045*  (0.027) 



2.230***  (0.355) 









-0.212***  (0.072) 

-0.054**  (0.021) 











-0.213***  (0.060) 

-0.056***  (0.019) 









-0.333***  (0.082) 















-0.379***  (0.094) 











-0.117*  (0.067) 

-0.205**  (0.079) 

-0.178*  (0.098) 





-0.084***  (0.029) 





-0.266***  (0.086) 

-0.239**  (0.092) 



-0.030*  (0.016) 

-0.086***  (0.026) 

fixed effects No No Yes Yes No No Yes Yes N 144 144 144 1445 144 144 144 144 R2 0.515 0.515 0.455 0.502 0.300 0.285 0.502 0.519 ***: p<0.01; **: p<0.05; *: p<0.1. Figures in parentheses are cluster-robust standard errors. 

Page 29 of 33 

Urban Studies

Figure 1. Bus routes in Clayton County in P1 and P3 

Urban Studies

Figure 2. Conceptual models of the difference-in-difference (DID) estimation 

Page 30 of 33 

Page 31 of 33 

Urban Studies

Figure 3. Treated tracts and matches in the two matching scenarios. 

Urban Studies

Figure 4. Regression coefficients on bus access across the six matched samples

Figure 5. Bus access, socio-demographic changes, and residential mobility, 2005-2019.

** At least 28 million people in the U.S. are transit-dependent and rely on trains, buses and ferries to get to jobs, school, community services and to see friends and family.

One finding, he said, is that pretty much everybody disliked the in-station kiosks for obtaining, loading and reloading Key Cards. It wasn’t much of a surprise that SEPTA would prioritize an array of methods for riders to pay, from smartphone wallets to credit and debit cards.

“They wanted a simpler process similar to what they see in other major cities,” Rosen said. “We’re basing the design on simplicity. People don’t want to have to wait at machines; they want to be able to basically get on the train or bus and go.”

As of last December, SEPTA has had a smartphone option, called Key Tix, that requires users to download prepurchased tickets to the phone by QR code.

Internally and in the request for proposals document, SEPTA has been calling the project Key 2.0, though officials said there’s been no decision about whether that would be the new system’s consumer-facing brand. “The marketing people may want to come up with something a little more catchy,” Grieshaber said.


In 1959, for instance, politicians still forced Baltimore’s fading private transit company, the BTC, to divert US$2.6 million in revenues annually to taxes. The companies retaliated by slashing maintenance, routes and service.

Local and state governments finally stepped in to save the ruins of the hardest-strapped companies in the 1960s and 1970s. Public buyouts took place only after decades of devastating losses, including most streetcar networks, in cities such as Baltimore (1970), Atlanta (1971) and Houston (1974).

These poorly subsidized public systems continued to lose riders. Transit’s share of daily commuters fell from 8.5% in 1970 to 4.9% in 2018. And while low-income people disproportionately ride transit, a 2008 study showed that roughly 80% of the working poor commuted by vehicle instead, despite the high cost of car ownership.

There were exceptions. Notably, San Francisco and Boston began subsidizing transit in 1904 and 1918, respectively, by sharing tax revenues with newly created public operators. Even in the face of significant ridership losses from 1945 to 1970, these cities’ transit systems kept fares low, maintained legacy rail and bus lines and modestly renovated their systems.Tax policies and subsidies have promoted highway development across the U.S. for the past century, creating car-centric cities and steering funding away from public transit.

Today, public transit is under enormous pressure nationwide. Inflation and driver shortages are driving up operating costs. Managers are spending more money on public safety in response to rising transit crime rates and unhoused people using buses and trains for shelter.

Other cities are using more targeted strategies to make public transit accessible to everyone. For example, “Fair fare” programs in San Francisco, New York and Boston offer discounts based on income, while still collecting full fares from those who can afford to pay. Income-based discounts like these reduce the political liability of giving free rides to everyone, including affluent transit users.

History shows what works best to rebuild public transit networks, and free transit isn’t high on the list. Cities like Boston, San Francisco and New York have more transit because voters and politicians have supplemented fare collection with a combination of property taxes, bridge tolls, sales taxes and more. Taking fares out of the formula spreads the red ink even faster.


Spin study shows commuters, low-income riders use e-scooters the most

RYAN DETO   | Tuesday, April 11, 2023 6:07 p.m.6091050_web1_Scooters_movepghTRIBUNE-REVIEWSpin scooters are pictured at the MovePGH launch in Downtown Pittsburgh in July 2021. EMAIL NEWSLETTERS 

TribLIVE’s Daily and Weekly email newsletters deliver the news you want and information you need, right to your inbox.

Pittsburgh has had electric scooter sharing for nearly two years now. A new study from the e-scooter provider Spin reveals more about who rides them, how often and for what purpose.

One of the biggest takeaways from the recently released Spin study is that ridership is larger and more varied than popular conceptions that e-scooters are mostly just used for recreation.

Spin recorded more than 480,000 e-scooter trips in 2022. With a rotating fleet of about 750 to 1,500 scooters in Pittsburgh, that means that each scooter averages at least one trip a day. This is down compared to 2021, when the company logged more than 576,000 scooter trips, according to an audit by Pittsburgh’s Department of Mobility and Infrastructure.

Spin rides start at $1 to unlock a scooter and then 39 cents per minute riding.

About 30% of riders report having ridden a scooter for recreation and only 12% report recreation as their sole reason for using scooters — like riding around Point State Park for fun.

More riders are using scooters for commuting purposes (48%) or meeting family/friends (45%). About 23% use Spin scooters for dining out/shopping purposes and 19% use them for essential errands.

Jason Shaffner, Spin’s general manager based in Pittsburgh, said the company’s Pittsburgh ridership figures exceeded his expectations.

“It is a very wide demographic of people that ride scooters for a wide variety of reasons,” Shaffer said. “This is a legitimate form of transportation to a lot of people.”

Who rides scooters?

Spin’s ridership follows closely with Pittsburgh’s racial demographics, which is about 65% white, 23% Black, 5.6% Asian and 3.5% Latino, Shaffer said. He said a disproportionate amount of young people (ages 18-24) ride scooters and that group makes up 50% of Spin’s ridership.

Approximately half of riders fall under Allegheny County’s median household income level, which is about $66,000 a year. And 33% of scooter users report having no access to a vehicle, and 18% of them report having occasional vehicle access.

Students remain heavy users of Spin scooters, but less so than in 2021. Part- and full-time students accounted for about 33% of scooter users in 2022, which is down from 40% in 2021.

While overall scooter rides are down compared to 2021, Spin’s 480,000 trips still easily outnumber POGOH, Pittsburgh’s bike sharing service, which recorded about 80,000 trips in 2022.

Where people ride

Spin said Downtown and Oakland have the highest use of scooters. Shaffer said Spin has been working to expand the stretch of the scooters, and that nearly every Pittsburgh neighborhood has experienced a presence of scooters at some time.

About half of Pittsburgh’s 90 neighborhoods experienced more than 10 trips a day between May and October of last year.

Scooter trip lengths average between three-quarters of a mile and a mile, and about 80% of trips are under 1.5 miles, according to the study.

To reach low-income neighborhoods, Shaffer said Spin also provides Spin Access zones in places like Manchester, the Hill District and Hazelwood where all rides originating in those neighborhoods are provided at discounted rates.

This program is an expansion of Spin Access, which allows riders who use government assistance programs to have access to deeply discounted rides. However, according to the 2021 audit, only 0.1% of unique spin users signed up for Spin Access that year.

Will the scooter pilot be renewed?

The Spin service was first authorized in July 2021 after the state government created a pilot program for Pittsburgh to become the first city in Pennsylvania to allow e-scooters.

Shaffer is hopeful that Spin’s ridership data show the pilot has been successful and will encourage the state legislature to extend the use of e-scooters in Pittsburgh.

He was pleased to see that 29% of riders were using Spin scooters to connect with other transit options like public transportation. The study said that number increased from 15% in 2021.

“We expect to get renewed. We have established something here that would be disappointing to a lot of people if removed,” Shaffer said.

Some local advocacy groups are opposed to scooters in Pittsburgh, and believe the pilot shouldn’t be renewed and legislators should focus their efforts on other mobility priorities such as fixing the city’s sidewalks.

Laura Wiens, executive director of Pittsburghers for Public Transit, said the numbers show Spin scooters are merely providing more mobility options for people that already have multiple options. While 30% of spin scooter trips are replacing vehicle trips, she said that 65% are replacing walking, biking or public transit, all of which have a better carbon footprint.

“We should be focusing on the sidewalk design and replacement, and instead are talking about a product that serves a very niche audience, which I think is the biggest problem with the thing,” she said.

Disability advocates have also criticized scooters for blocking sidewalk access when they are illegally parked.

Shaffer said that 3% of scooter trips end in parking jobs that block sidewalk access. He said the company is working to lower that, and has implemented warnings, fines and suspensions for riders who park the scooters improperly.

He said the warnings have been effective. Of the people who get warnings, 92% fail to reoffend.

Pittsburgh City Council is hosting a public forum on Spin e-scooters at 2:30 p.m. Wednesday in council chambers.

Ryan Deto is a Tribune-Review staff writer. You can contact Ryan by email at or via Twitter .


The Federal Aid Highway Act of 1956 was the largest peacetime public works program in American history. Approximately 42 billion tons of dirt were moved to create the interstate highway system, the equivalent of 116 Panama Canal projects. One million Americans were displaced from their homes as well, while annual federal spending on highways quadrupled in the years that followed.

Clayton Nall, an associate professor of political science at University of California, Santa Barbara, has sought to show that the interstate system also reshaped the political landscape of America, fueling extreme partisan polarization. In his 2018 book The Road To Inequality, he dives into historic survey data to show that predominantly Republican voters were far more likely to take advantage of further-flung locales opened up by new highways, creating a regional polarization that proved especially strong in Sun Belt areas that hadn’t substantially suburbanized before the 1950s.

Nall’s book also shows that as regional polarization grew, certain aspects of transportation policy became more contentious. Transit and other non-auto sources of spending increasingly polled like means tested anti-poverty programs, which are often unpopular among Republicans and conservative voters. (He finds that the more a transportation policy proposal would help city residents, or is specifically targeted to lower-income people, the less popular it is.) Highway spending, meanwhile, is beloved by all, like the universal welfare programs Social Security and Medicare.

Governing talked with Nall about why this process was so much more dramatic in the Sun Belt, the Obama-era surge in polarization around transit, and whether transportation reformers should hope that their policy issues don’t get attention in Congress.

Governing: You argue that the interstate highway system facilitated residential sorting and political polarization. How did the fruits of the 1956 act differ from the suburbanization processes that had already occurred, from streetcar suburbs to places like Levittown which had already been built before the mid-1950s?

Clayton Nall: Suburbanization was already happening prior to the construction of interstate highways. But the interstates were, in terms of volume, much more significant. They enable a degree of greenfield development in the outlying parts of metro areas that wouldn’t have occurred on nearly the same scale absent federal funding and the construction of a large-scale highway system.

One of my case studies looked at the growth of suburbs in Milwaukee. A great example of this style of development that differs from the streetcar suburbs is the I-94 corridor west of Milwaukee. If you look at presidential election results across the period from the 1950s to the 1970s after I-94 went in around 1963-64, you see a sudden explosion of population along that corridor in Waukesha County, which is now one of the most important counties for Republicans nationwide.

Prior to that point there had been some suburbanization. But it wasn’t anywhere near the scale enabled by the interstates. The goal of my project is not to argue that there’s a single cause driving suburbanization or urban-suburban polarization, which did exist prior to the interstates. But interstates catalyzed that to a much greater degree than anybody had seen before.

Milwaukee. After I-94 was built around 1963-64, the metro area saw a sudden population explosion along its corridor in Waukesha County, which is now one of the most important Republican counties nationwide.

Governing: You find that this process was most marked in high-growth areas, like the Sun Belt, and especially in the South. What explains that? There’s still plenty of segregation and polarization in the industrial Midwest or the urban Northeast. Wisconsin was just your prime example there!

Nall: Even by the standards of the Rust Belt, Milwaukee is an unusually segregated city. Racial segregation is highly correlated with partisanship, and has become more correlated with partisanship, which helps explain why Milwaukee is such a significant case. But otherwise, particularly in the northeastern parts of the Rust Belt, there was a lot of suburbanization that had already taken place.

There wasn’t any of that pre-existing development in, say, suburban Atlanta. Even in the 1950s, the Atlanta metro area was nothing compared to what it has become. Same with all these cities across the Sun Belt. The interstates are a lot more pivotal for the development of those areas. Take a look at a place like Minneapolis, which had already seen a bunch of development into the suburbs around Minneapolis and St. Paul. By the ’60s, ’70s, ’80s, you’re already seeing more expansion outward, where the inner suburbs are becoming more Democratic, the housing stock is depreciating and you’re seeing some filtering out where more affluent households are moving out into more remote suburbs. The remaining housing left in these inner suburbs is becoming more diverse, more likely to be occupied by working-class families. Some of that is being picked up in the state and regional political differences.

Governing: How did this process affect partisan attitudes toward transportation policy? 

Nall: There are two things that are happening. The book relies heavily on data from this really terrific historical archive called the Roper Center for Public Opinion Research, which has all sorts of historical surveys from Gallup and Roper. But they only started asking questions about transportation policy, transit and highways in the ’70s. Prior to that, we don’t know a lot about how people thought about highway construction or housing development, let alone what the U.S. should do about declining transit agencies or railroads.

We’ve seen a few things in the aftermath of the construction of interstate highways, though. One of the things that’s gotten a lot worse in recent years is that some of these issues have become very polarized, perhaps as a result of almost no Republicans living in really urbanized areas anymore. Republicans no longer have a direct attachment to urban concerns.

The survey data that we have show that Republicans have become much more opposed to funding for transit. When you ask them about specifics around things like funding bus lines to low-income housing projects, they’re extremely negative toward any approaches to transportation that involves targeting benefits to urban areas. However, there’s still bipartisan consensus around highway spending.

Why would that be? Republicans are no longer urban and therefore have pretty much removed themselves from any direct concern with urban transit funding. Both Democrats and Republicans live in pretty suburban areas, both parties are relatively suburbanized and many subgroups of both parties support more spending on highways. You’ve seen this in the politics of recent highway bills. Democrats have talked a big game about climate change and developing alternatives to more road building. But when the chips are down, they’re still supporting highway bills that reflect this consensus on highways.

Governing: You did not find distinctive rural, suburban or urban opinions around highway spending. Instead, you found broad support for it across Americans of all residential types. Where polarization really hits, increasingly in recent decades, is in Republican opposition to non-highway spending. Transit is becoming polarizing. 

Nall: I wonder if we would be looking at such a scenario if we were still fighting out these battles within cities. If fights over school desegregation and the other fights of the 1960s had not been resolved as a result of white flight? If these kinds of conflicts were still playing out in cities, you might see Democrats and Republicans engaged more directly on these issues, and there would be a lot more direct open conflict. Instead in federal transportation bills, there’s a log roll that’s happened. Roughly 20 percent of the Highway Trust Fund now goes to transit programs, which are predominantly urban liberal places getting big grants to fund transit projects. Those urban areas [with] their own [votes] can never secure funding for transit services, so they’re now vested in the perpetuation of the Highway Trust Fund. This log roll actually helps perpetuate the deal where highways get most of the money.

Democrats and Republicans have become more geographically polarized as a function of population density. Then, during the Tea Party backlash against Obama, elite messaging on transit and trains really turned against Obama early in his term. Folks like Scott Walker rejecting federal high-speed rail money. A bunch of Republican state leaders rejected those programs. You see that appear in some of our survey data as Republican voters started to respond to elite messages about how trains and transit are not something Republicans support. They wouldn’t have got that message in the 1970s from the Republican leadership.

Chicago’s elevated subway. During President Obama’s administration, survey data showed Republican voters responded to messages about how trains and transit are not something they support. This was not the case 30 years earlier.

Governing: But this increasing polarization around transit hasn’t changed the 80/20 split of the Highway Trust Fund or other federal transportation policies. Why has there been such stasis at the national level, why hasn’t some enterprising Republican turned this into a cause célèbre on Fox News? 

Nall: There have been some attempts at such crusades. Freedom Caucus members have introduced bills to zero out transit’s share of the Highway Trust Fund. In the 2012 Republican Party platform, there was a line put in by activists saying that Democrats want to use our transportation budget to engage in social engineering. Of course, the construction of the interstate highway system was social engineering too, but the kind Republicans like.

If you look at legislative activity, there are still a lot of bills that are passed on a bipartisan basis. This is just informed speculation, but one of the reasons transit has held on to funding is it’s not a fight that excites the party bases. Democrats are trying to build majorities really dependent on the suburbs, so they’re not going to go out of their way to demand less money for highways.

The perfect symbol of this is Democrats pushing through legislation to provide things like national electric car charger networks on the interstate system. It’s not going to offend swing suburban voters to see a bunch of electric chargers on the highways. But it also really doesn’t do very much to facilitate changes in land use, doesn’t do much to help urban residents who might want to not rely on driving or might not be able to drive. So, neither party has a strong interest in smashing the bipartisan deal where transit continues to get a cut of the action.

Governing: It seems like if there’s more attention paid to these issues by prominent Democrats — like Obama with high-speed rail — the more it becomes a polarizing issue that Republicans will organize against. Part of the reason this hasn’t been polarized at the national level is that no one’s really paying attention.

Nall: Planning and transportation policy can be really boring, and frequently doesn’t involve clear-cut class or redistributive dimensions. It can be hard to see the clear winners or losers in the decisions that are made in the various congressional infrastructure committees.

Historically, the entire transportation bureaucracy up until the 1970s was white male engineers concerned with building roads, often dominated by rural interests. That was what it meant to be in transportation policy in the United States. Now there’s more diversity and a realization within transportation departments that what they are doing is actually social policy. But it hasn’t fully penetrated these agencies. They continue to be heavily dominated by people who think transportation policy equals highway building.

On the other side, progressive activists, other than those who are specifically really interested in transportation policy, have not really challenged the status quo. I didn’t see a lot of activism from, say, the Sunrise Movement around transportation infrastructure bills of recent years. They’ve been focusing on the climate bill, but they haven’t focused on legislation that is part of the routine operation of government that’s having tremendous effects on carbon emissions.

Governing: But your research seems to show that they shouldn’t do that. A lot of transportation reformers were unhappy with last year’s infrastructure act and its focus on highways, but things could have been a lot worse. If these trends continue, future infrastructure bills could zero out transit funding, or at least cut it, especially if this becomes a focus on Fox News or talk radio. Maybe it’s better for those advocates to quietly accept the status quo.

Nall: There’s another school of thought. Martin Wachs was a giant in transportation policy and civil engineering, one of the leading academics engaging with the politics of transportation. He thought about how federal transportation programs create a lot of perverse incentives for states to spend a lot of money on big projects that are wasteful, and ultimately, detrimental to effective and equitable transportation policy.

There’s a libertarian, devolution argument that says perhaps now that the interstate highway system has been built, it should be on state highway departments to maintain and expand it.

Get rid of the federal gas tax, and just leave it to states to fund their own highways. The roads are already maintained by, and property of, the states. There’s an enormous moral hazard that has resulted from such generous highway funding that in large share goes to just build more and more lanes. If states had to fund their own highways, they probably would not fund such extensive expansion. They would be more strongly considering various types of conservation measures, like congestion pricing, as opposed to just building more lanes.

It’s an argument I’ve heard from both the left and the right: Perhaps a better solution is to not provide all this money for states to do gold-plated transit projects and highway expansions. Some of the projects built by state governments using federal transportation money are frequently gold-plated infrastructure projects, because the money is restricted to capital spending. They’ll spend money on equipment that they maybe don’t need. They’ll build fancy new rail stations, big transit centers that don’t actually have decent transit service. The Salesforce Transit Center in San Francisco is a great example of this.

One of the things that happens is federal money comes down and then is allocated and spent by metropolitan planning organizations (MPOs) that are responsive to a set of local pressures that can end up being biased against providing for transportation equity needs. That’s partly because the incentives that are built into federal transportation funding do not provide for a way to provide high-frequency reliable bus service between cities.

Governing: What would be a strong avenue for transportation reform advocates to push for change? It doesn’t seem like major change at the federal level is likely. But in many Northeastern and Midwestern regions a lot of the suburbs are increasingly Democratic. Maybe a more advantageous route for change is to push for regional funding. Urban representatives could ally with partisan counterparts at the suburban level.

Nall: One of the ways out from political polarization is that the inner suburbs, in many metro areas, have become a lot more Democratic. If you look at the start of the housing life cycle, a lot of these areas have older housing stock which tends to be occupied by lower socioeconomic status groups. Immigrants who are moving to the U.S. and instead of settling in cities, they’re settling in suburbs. They’re diversifying those areas and making them a lot more Democratic.

There is definitely a coalition emerging in a lot of metro areas between inner suburbs and the central cities. Maybe those inner suburbs see their fate tied up with having a healthy and productive central city, with a more equitable transportation split, and providing transportation services that actually allow more efficient land use and frequent transit service. Maybe actually making transit something that’s appealing to the middle class!

I think one of the main reasons people are opposed to certain types of transit spending is that in a lot of places it’s treated as something like welfare. Providing bus service to low-income housing projects, the partisan split on that survey item was so huge. We’re talking partisan attitudes toward welfare huge.

For a lot of Americans outside of cities like New York, D.C., San Francisco, Chicago, most people don’t use transit. Riding buses is something poor people do. But if it’s understood that it’s the Democratic position to support funding for transit, and that’s something that all good Democrats in the suburbs believe they should be doing, that’s a way forward for people who want to assemble a winning coalition for better transit funding.

Related Articles

Can Pete Buttigieg Really Change Transportation Policy?

Rolling Coal vs. Biking. How Politics Has Changed Transportation.

A new paper shows that ideological commitments increasingly determine attitudes towards transportation policy.

“Historically, it’s been apolitical,” says Kelcie Ralph, associate professor at Rutgers University’s Bloustein School of Planning and Public Policy. “But over the past 15 to 20 years, and amplifying even more recently, you’ve seen partisan rhetoric particularly from the right.”

The paper Ralph published with her colleagues Nicholas Klein, Calvin Thigpen and Anne Brown delves into the results of a survey of 600 Americans. They were given a series of questions about transportation policy and asked to rate themselves on an ideological spectrum from very liberal to very conservative.

The results paint a complex picture of polarization in the transportation realm. The researchers found strong majorities in support of some ideas that break with the dominant policies of the mid-to-late 20th century. Sixty-nine percent of respondents supported mixed use neighborhoods, while 63 percent were in favor of shifting more car trips to transit, walking and biking.

There are contradictory results too. Only 32 percent said that the goal of transportation policy should be to reduce driving, as opposed to making it more convenient. In every case, liberal respondents were more likely to favor policies that encouraged alternatives to driving while conservatives were more supportive of the car-dominated norm.

“Even when we control for [a host of factors], we find a big effect for the very conservative respondents being very reluctant to give up our status quo of a society that favors driving,” says Nicholas Klein, assistant professor at Cornell University’s College of Architecture and Planning.

These findings fit within “The Big Sort” theory that Bill Bishop made famous in his 2008 book by the same name. He argued that Americans are increasingly living near people who believe the same things they do, and that partisans are more likely to settle around those with similar voting records.

More progressive and Democratic voters tend to cluster in cities, where their political power is diluted by some of the anti-democratic aspects of the American political system. More rural and exurban areas tend to be overwhelmingly conservative and Republican, a trend that seems to grow stronger with every election cycle. This partisan distribution has helped the GOP build political power even with consistently smaller numbers of voters (thanks in large part to how the electoral college and the U.S. Senate allow more rural, less populous states to retain significant political power).

A litany of studies and polls find that most conservative Americans express a desire to live in areas with big houses that are spaced farther apart, with commercial areas kept at a distance. More-liberal Americans are likely to want denser, more walkable, mixed-use communities.

Petro-Masculinity Versus Walkability

This summer, a Pew Research Center survey found that ideology was the biggest distinguishing factor between those who want to live in a walkable community and those who want to drive to do anything. That was a stronger predictor than factors like age or education.

As these identifiers have risen in salience, politicians and pundits have made grandstanding on transportation a part of their platform.

“There’s a sense on the right that walking, biking and transit are for people who do not vote for them,” says Ralph. “Petro-masculinity is a thing now, huge trucks rolling coal. That is a signal that you’re a conservative man. We’re turning up the dial on these differences.”

Evidence of this turn in American politics can be seen in the GOP platform in 2016 and 2020, where it was proposed that federal support for transit be eliminated entirely.

As a transit advocate in northeastern Ohio, Akshai Singh has been on the frontlines of transportation reform. He says that until recently, Republicans in the region have been relatively supportive of transit.

A pickup trick spewing carbon emissions.
An example of a pickup truck “rolling coal.”

Former Cleveland mayor and Ohio senator George Voinovich helped secure funding for the Greater Cleveland Regional Transit Authority. GOP Congressman Steve LaTourette co-sponsored a bill that would have given transit systems operational support from the federal government.

“That’s the politics that we historically have had, but in these last 10 years especially we’ve seen larger amounts of partisanship,” says Singh, national transit justice organizer with Alliance for a Just Society.

Singh argues that the way to fight against these trends is to focus on local self-interest, not partisan divides. He notes that the Center for Transportation Excellence has tracked the success rate for local transit funding campaigns over the last decade and found that the overwhelming majority win voter support every year.

“People are invested in their communities and are happy to invest in a system that they know is going to better connect them to their communities, which is the most direct translation of self-interest,” says Singh.

In their paper, Ralph, Klein and their colleagues argue that transit planners should attempt to win converts, in part,

  • through pilot programs and tactical urbanism — showing residents the results of programs that offer alternatives to car usage and the benefits that stem from them.
  • Positive messaging that focuses on advantages for local residents, not global challenges or partisan struggles, would help too.
  • They argue for a robust campaign to educate Americans on the realities of induced demand, the fact that expanding roadways only very briefly eases traffic congestion. They found that just 45 percent of liberals and 24 percent of conservatives knew that traffic relief after highway expansion would be temporary. They also saw that support for changing the auto-oriented status quo was 25 percentage points higher among those who understood the concept. This speaks to another complicating factor of their survey:
  • Unlike most countries, America is almost entirely car dependent. Automobile ownership is nearly ubiquitous, with over 90 percent of households owning one. Most Americans use a motor vehicle for most trips regardless of their ideological proclivities. “It can be hard to study these kinds of questions in the United States, because car ownership is largely universal,” says Klein. “If you’re doing a national study, there’s not going to be many people in the sample who live in places like South Philadelphia or walkable parts of Chicago.”
  • The few respondents who didn’t own a car or primarily traveled without one were more likely to embrace transportation reform. Meanwhile, the most likely opponents of reform were those who defined themselves as very conservative — they were 41 percentage points less likely than moderates to support the notion of shifting trips to biking, walking or transit.

Ralph and Klein say that as GOP politicians increasingly pander to the most conservative elements of their base, the polarization of transportation policy may sharpen still further. That’s a recipe for a kind of doom loop, where rank-and-file voters’ passions become more heated about an issue as politicians and ideological media focus on it.

“I suspect that we’ll see continuing ramping up of that and that on the 2024 campaign trail, transportation will be a more central issue,” says Ralph.