Do Employers’ Neighborhoods Predict Racial Discrimination?

This paper uses evidence from a large field experiment to explore whether the racial composition of employers’ neighborhoods, as well as other neighborhood and business characteristics, predicts racially discriminatory employment decisions. Over 15,000 fictitious job applications were sent, in otherwise-similar black and white pairs, to low-skill job postings distributed throughout New Jersey and New York City. Overall, white applicants received 23% more callbacks than equivalent black applicants. The white advantage was much larger in whiter and less black neighborhoods. We interpret this pattern to suggest some form of in-group preference and/or irrational stereotypes; the pattern cannot readily be explained by “rational” statistical discrimination. In prior work on Ban-the-Box laws, we showed that when employers lack access to criminal records they appear to make exaggerated negative assumptions about the likely criminality of black applicants. We now show that this effect too was driven by employers in less black neighborhoods, and conversely that this apparent stereotyping pattern can explain some (but not most) of the effect of neighborhood composition on the black/white callback gap. Real-world black applicants are presumably more likely to apply to jobs in black neighborhoods (and white applicants to jobs in white neighborhoods) because they are more likely to live nearby. Through simulations that reweight our results geographically to mirror a real-world population distribution by race, we show that this geographic selfsorting will likely greatly magnify the net disadvantage that black applicants face, rather than mitigating it. This is because within each of these jurisdictions, job availability and overall callback rates are lower in nonwhite neighborhoods. Other local characteristics, including partisan vote share, crime rates, income, and poverty rates, did not predict racial discrimination rates once racial composition was controlled for, nor did observable business characteristics or the other varied characteristics of our applicants. The white advantage was much larger in New Jersey than in New York City, even after accounting for neighborhood-level differences.