Syndicate content

Taking the Bus to Opportunity: Guest post by David Phillips

In the 1960s, black and white individuals in the United States had radically different labor market outcomes. In 1962, the unemployment rate for African-Americans was 13 percent while it was only 6 percent for whites. Fifty years have passed, enough time for Martin Luther King to go from movement leader to monument, but as of 2010, the unemployment rate in the U.S. was 18 percent for blacks and 10 percent for whites. For all that has happened, there is still the same gap in economic outcomes. Why hasn’t the gap narrowed? There’s not likely to be one lone answer, but the housing patterns of U.S. cities suggest one contributing factor. 

Many U.S. cities are de-facto segregated, and few employers decide to locate in the African-American part of town. The resulting “spatial mismatch” of these individuals from available jobs could make the job search process more difficult, leading to adverse labor market outcomes (Kain, 1968; Wilson, 1997).  In my job market paper, I test this hypothesis using a randomized field experiment which subsidized public transit access for a group of active job-seekers from job-scarce neighborhoods of Washington, DC.

As the map above shows, Washington provides a prime example of a city that, despite the fall of legal segregation decades ago, exhibits de-facto segregation with African-Americans clustered to the east and Whites to the west. Meanwhile, jobs are clustered downtown (center of the map) and in the western suburbs. This generates a potential problem for the job search process of poor, minority individuals living on the far Southeast edge of the city. For someone who has been unemployed and looking for minimum wage employment for several months, the cost of repeatedly riding the bus across town can be prohibitive. Empirical evidence based on natural experiments has generally confirmed that this type of spatial mismatch hurts labor market outcomes for minority individuals (Holzer et al, 2003; Zax and Kain, 1996), but when researchers have tested for spatial mismatch with randomized experiments, the results have come up nil. Most prominently, the Moving to Opportunity (MTO) experiment randomly assigned offers for free housing vouchers that required recipients to move from high-poverty public housing to lower poverty census tracts, but moving out of high poverty neighborhoods had no measurable impacts on labor market outcomes (Kling et al, 2007). One way to rationalize these conflicting results is to note that housing moves are a complex bundle that can decrease distance to available jobs but also have negative impacts like severing social networks.

To test the importance of spatial mismatch directly, I tried something much less ambitious than the MTO experiment but focused more directly on access to employment. In cooperation with a local non-profit employment agency called Jubilee Jobs, I designed and implemented a field experiment in which a group of their clients were randomly selected to receive subsidized access to the public transit system. The sample was composed of 468 clients of the partnering organization, all of whom received standard job search assistance (vacancy information, interview skills training, etc.). The sample consists mainly of individuals who face major barriers to labor market participation (e.g. low education, criminal history), and the median individual has been unemployed for 10 months at the time of recruitment. On top of this, treatment group members were offered fee-reducing public transit cards (WMATA SmarTrip). The card itself reduces the per-trip bus fare from $1.70 to $1.50 while also functioning as a store of value that can be used to pay for future trips. Individuals assigned to treatment were offered two cards loaded with a total balance of $50. 

For this modest subsidy, I find that being assigned to treatment results in a large short-run decrease in unemployment durations. For instance, being assigned to receive the subsidy increases the probability of finding employment within 40 days by 9 percentage points, from 26 percent to 35 percent. This difference narrows to a statistically insignificant 5 percentage points by the end of the 90 day follow-up period. For a different take on the same results, quantile regressions indicate that the 30th percentile of the distribution of unemployment durations falls from 49 to 35 days , while the 45th percentile falls by a statistically insignificant 12 days from 85 to 73 days (Note: quantiles at and above the median are censored due to the short follow-up).

From these results, I conclude that public transit subsidies speed up the job search process in the short-run with uncertain long-run effects. Of course, short run does not mean unimportant. In a world of frictional unemployment, equilibrium unemployment rates fall if people match with available jobs more quickly, and the results are consistent with a search model/frictional unemployment view of the labor market in which transit subsidies reduce search costs for employable individuals, allowing these individuals to find work more quickly. I find some (for the most part statistically insignificant) support for this mechanism in the data. The subsidy group searches more intensively, both applying to jobs more frequently and traveling further from home to look for work. Altogether, these results confirm that spatial mismatch exists at a magnitude that matters for persistent differences in unemployment rates. In the policy arena, “place-based policies” that try to revitalize poor areas and housing policies like Moving to Opportunity that move people “up and out” gain most of the attention. My results indicate that when we think about policy responses to spatial mismatch, we might want to consider cheap interventions that improve access to existing transit systems more seriously.

Of course, this study doesn’t close the book on spatial mismatch. The clearest limitation to my work is its scope. The sample is small; the follow-up period is short; and with some back-of-the-envelope calculations you can probably see that my budget would be a rounding error for many other RCTs. As a result most of my tests have relatively low power; I cannot make any statements about long-run effects; and the design was limited to only one treatment package. These shortcomings prevent a cleaner analysis of the mechanism leading to the observed effects. Context matters as well. My sample consists of individuals who are actively seeking employment, are receiving complementary job search services, and live in a city with a reasonably effective public transit system. While my results indicate that targeted transportation subsidies could be very effective, these contextual factors suggest that dropping metro cards on a random sample of unemployed individuals may be less likely to succeed.

As someone who also does work in development, I think it’s also useful to think how this study relates to the developing world, where spatial mismatch seems even more likely to show up in quickly growing cities with less-developed haphazard transport infrastructure. We have some good evidence on the impact of paving roads (Gonzalez-Navarro and Quintana-Domeque, 2011) but little experimental evidence on how spatial mismatch and access to transportation plays out in the urban labor markets that contain an increasing fraction of the world’s poor. In the U.S. experience, residential segregation has persisted and I provide some evidence on how this spatial mismatch can contribute to on-going disparities in the labor market of one city. How this plays out in the rest of the U.S. and the world remains to be seen. 

David Phillips is a Ph.D. candidate at Georgetown University and is on the job market.