How buses alleviate unemployment and poverty

Between March 2020 and April 2021, the Metropolitan Atlanta Rapid Transit Authority (MARTA) suspended half of its bus routes due to the COVID-19 pandemic, reallocating services to reduce crowding on busier routes. While MARTA strived to maintain services on key routes serving essential workers and healthcare facilities, many communities that previously relied on bus transit lost their only access to public transportation. Even in a car-centric city like Atlanta, there are communities where a substantial percentage of the population do not have a private vehicle, and the absence of bus transit could have had significant effects on the daily lives and accessibility of affected riders.

Metro Atlanta is not alone in transit disruptions during the pandemic. Although buses provide the bulk of public transit services in the United States, they are also disproportionately subject to delays, schedule changes, and service cuts due to disasters, staffing issues, and revenue shortfalls. Considering that low-income populations and ethnic minorities more often ride buses, these disruptions can have significant implications on transportation equity and life outcomes for affected individuals. Moreover, public transportation research tends to pay more attention to rail and bus rapid transit (BRT), providing little empirical evidence on the consequences of disruptions in regular bus services.

To bridge this gap, we examined a natural experiment in Clayton County, Georgia, where the local public buses were suspended due to a budget decision between 2010 and 2015. Clayton County is a predominantly Black working class suburb to the south of the City of Atlanta, with one of the highest percentages of workers commuting by transit in metro Atlanta. Understandably, the termination of local buses adversely affected these commuters and other transit users. Specifically, we tested two hypotheses: 1) the spatial mismatch hypothesis, based on the loss of job accessibility by bus, predicts a negative relationship between bus access and indicators of deprivation, including poverty and unemployment rates; and 2) the residential sorting hypothesis, which posits that low income individuals gravitate toward transit rich neighbourhoods, and the termination of bus transit could lead to an outflow of the poor, resulting in lower poverty and/or unemployment rates.

Using propensity score matching and a difference-in-difference approach, we found support for both hypotheses. Reduced bus access was associated with an increase of white people and a decrease of Black people in affected census tracts, suggesting possible selective migration between neighbourhoods. Meanwhile, the poverty and unemployment rates significantly increased in census tracts that lost bus routes, indicating that the spatial mismatch effect due to lost job access had overwhelmed any residential sorting effect in the five-year period. These observations are consistent among alternative matching schemes and model specifications, and the overall effect sizes are around 3.7 percentage points for the unemployment rate and 4.6 percentage points for the poverty rate. These findings, with the strong quasi-experimental design, provide solid empirical evidence for the important role of bus transit in job access and poverty mitigation.

Our study underscores the need of more research, funding, and support for public transit, especially buses, in the United States. On one hand, it calls for federal and state investments and improvements in public transit systems to reduce the existing 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 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.

Read the accompanying article on Urban Studies OnlineFirst here.