Every year, BARI updates and releases data from the American Community Survey on the Boston Data Portal. To demonstrate the value of these data, our researchers wrote a series of blog posts, highlighting some maps and insights they’ve been able to glean from the data.
by Riley Tucker
In recent years, a number of articles published in the media claim that Boston is among the country’s most racially segregated cities. Notably, these articles all analyzed Boston’s metropolitan statistical area, a nearly 3,500 square mile geographic unit that expands deep into New Hampshire. As such, there is an opportunity to gain a more nuanced understanding of racial segregation in the Boston-area by focusing in closer to the city. To probe this issue, I’ve conducted a series of analyses using census tract data from the 2013-2017 American Community Survey, released publicly by BARI this week.
First, I’ve limited our area of interest to Suffolk County and the three counties that border it: Essex, Middlesex, and Norfolk. Across these counties, there is a significant degree of segregation at the census tract-level. Compared to whites, 62% of Black residents and 58% of Hispanic residents would need to move to a census tract with more white residents for all groups to be equally distributed across space. To understand where these groups are living and where they would need to re-locate, I’ve constructed a “hot spot” map which shows census tracts where there are high concentrations of white, Black, and Hispanic residents relative to the rest of the area. The contrast is quite striking even between minority groups; while central Suffolk county has high populations of both Black and Hispanic residents, Hispanic populations have generally formed north of the city in Middlesex and Essex counties, while Black populations have instead emerged in Norfolk county. In contrast, there is a large swath of concentrated White communities on the exteriors of Norfolk and Essex counties and all throughout Middlesex county. To achieve racial equality across space in the greater Boston area, the majority of the Black and Hispanic residents in Suffolk county and surrounding areas would need to move outward to more suburban towns.
Next, I went even further by restricting the spatial analysis to Boston’s city limits. Compared to the larger region, the level of residential segregation relative to whites is stronger for Black communities but slightly weaker for Hispanic communities; 72% of Black residents and 54% of Hispanic residents would need to move to achieve racial equality across space. As with the larger region, we can see that there are separate areas with concentrated Black and Hispanic communities. The vast majority of tracts in Dorchester and Mattapan have predominantly Black populations while the majority of tracts in East Boston have predominantly Hispanic populations. Conversely, there are dense populations of white residents in South Boston, Charlestown, and Brighton. To achieve equal distribution of racial groups across Boston, the vast majority of residents in Boston’s southeast corner and about half of those living in the northeast corner would need to relocate to other areas, predominantly those surrounding the central business district.
Overall, the story is consistent: there is great residential segregation along racial lines both within and outside the city of Boston. We can also see that segregation occurs at different scales; while Suffolk county appears to be entirely composed of tracts with high concentrations of Black and Hispanic residents when the analysis is conducted at the county-level, the Boston analysis shows that within the city itself there are pockets of segregated Black, Hispanic, and White communities. No doubt, if this exercise was repeated for Middlesex, Essex, and Norfolk counties independently, unique patterns of segregation would emerge within each county. Don’t believe me? Head on over to the Boston Data Portal and check the data out for yourself.
 Percentages calculated using the index of dissimilarity
 Spatial clusters identified using the Getis-Ord Gi* statistic. Only clusters with 95% significance and above are presented.