Center on Urban & Metropolitan Policy Racial Segregation in the 2000 Census: Promising News By Edward L. Glaeser, Harvard University and the Brookings Institution, and Jacob L. Vigdor, Terry Sanford Institute of Public Policy, Duke University “While segrega- Findings tion remains An analysis of racial segregation in roughly 300 Metropolitan Statistical Areas using 2000 Census redistricting files indicates that: high in many ■ Overall black/non-black segregation American) remained steady between levels are currently at their lowest 1990 and 2000, although the number point since roughly 1920. There are of African-Americans living in those metropolitan still a large number of “hypersegre- tracts dropped. gated” metropolitan areas, but the 1990s continued a three-decade trend ■ The West is the most integrated areas, it has towards decreasing segregation region of the country, followed by throughout the U.S. the South. The Northeast and Midwest are still quite segregated. generally ■ Out of 291 MSAs analyzed, all but 19 are more integrated than they ■ Segregation declined most sharply were in 1990. The average decline in in places that were growing quickly, decreased segregation was 5.5 percent. in places where the percentage of blacks in the population was ■ The decline in segregation comes changing (growing or shrinking), across the about primarily from the integra- and in places where blacks made up tion of formerly entirely white a small portion of the population in census tracts. The number of over- 1990. Segregation remains extreme country.” whelmingly African-American census in the largest metropolitan areas. tracts (80 percent or more African- I. Introduction the same time period, the segregation levels of only 19 MSAs rose.2 Across metropolitan he 2000 Census documents that, for areas the average decline (without adjusting the third straight decade, segregation for population differences) was 5.5 between blacks and non-blacks percentage points. across American metropolitan areas The purpose of this survey is to examine Thas declined dramatically. Between 1990 and the change in the levels of segregation across 2000 the segregation levels of 272 Metropol- metropolitan areas since 1990, and over a itan Statistical Areas (MSAs) declined.1 Over longer period of time. While segregation April 2001 • The Brookings Institution • Survey Series 1 remains high in many large metropol- tion that looks at pre-1940 time tion measures is to define what it itan areas, especially in the Northeast periods is forced to use political wards, means to be African-American. In and Midwest, it has generally which are in a sense arbitrary and previous censuses, “black” was an decreased across the country and over surely too large (some contain tens of exclusive category, but the 2000 time. The West and South are the thousands of people). For the post- Census allows respondents to identify fastest growing and least segregated 1940 time period, there is the option themselves with multiple races, and regions of the U.S. of using blocks (equivalent to a city therefore the measurement has Regional differences may help to block in most urban areas), block become more complicated. On one explain why previous authors have groups (areas with roughly 1,000 hand, over 95 percent of all respon- generally not focused on the declines inhabitants), and tracts (which are dents in our sample who identified in segregation across the U.S. Massey larger units of roughly 4,000 people). themselves as at least partly black and Denton (1993), in their seminal The advantage of blocks or block identified themselves as only black. work on segregation in America, groups is that they are smaller and On the other hand, the remaining 5 focused primarily on larger cities in allow us to better understand the percent are not distributed evenly the Northeast and Midwest, and micro-geography of urban residence. across MSAs and their presence could therefore may have downplayed the The advantage of tracts is the relative potentially skew segregation indices. importance of recent overall declines ease of comparability over time: for We will present segregation indices in segregation. While the continuing most large cities, tract data are avail- that make use of two basic definitions. segregation of the so-called “rust belt” able beginning in 1940. Primarily for First, the most inclusive definition cities is important, it is also important consistency with our previous work, counts as African-American anyone to document and understand the we have decided to use census tracts who checked “black” as one of his or changes in segregation in more vibrant as the relevant sub-areas. her racial identities. Second, we define and developing areas of the country. African-Americans as those who The relevant definition of city checked only black as their racial iden- II. Methodology The question of city definition tends tity. For the implementation of to come down to two choices. First, segregation indices this distinction here are two basic measures segregation indices can be defined for makes little difference in most cases.4 that are generally used to the metropolitan area as a whole. In capture the degree of residen- this case, the segregation measure will The relevant non-black population tial segregation within an reflect both center city-suburb segre- Finally, in choosing the relevant non- TAmerican city. The “Dissimilarity” gation and the segregation of people black population, there are two basic index is a measure of the proportion of within central cities and suburbs. options. First, one can use non- black people (or non-blacks) that Second, segregation indices can be hispanic whites. Second, one can use would need to move across census defined for sub-units of the metropol- all non-blacks. The essential differ- tracts to get a perfectly even propor- itan area, such as the central city. ence between these options lies in the tion of black residents across the While it is often quite interesting to treatment of Hispanics. Asian, Native entire MSA. The “Isolation” index know about segregation for central American and Pacific Islander popula- indicates the percentage of black resi- cities and suburbs separately, we will tions are generally too small to dents in the census tract where the just focus on segregation at the metro- influence segregation (and their resi- average black resident lives. (These are politan area level. Our past work has dential patterns generally resemble explained in more detail below.)3 convinced us that there is an those of non-hispanic whites). Both To calculate these measures one extremely high correlation between options seem quite reasonable to us, must consider four questions: First, segregation at the central city level but it needs to be understood that what is the appropriate sub-area to and segregation at the metropolitan when different definitions are used, use? Second, how do we define a city? area level across cities. However, our different questions are answered. If Third, what is the appropriate defini- past work has also suggested that the non-black, non-hispanic whites are tion of the black population? Fourth, decline in segregation would tend to used as the comparison groups, then what is the appropriate non-black look steeper if we particularly focused the segregation measures will capture population to consider? on central city level segregation. the extent to which blacks are segre- gated from this group. If all The appropriate geography The appropriate definition of non-blacks are used, then segregation There are generally three sub-areas “African-American” measures will capture the extent to that have been used for segregation Another question that needs to be measures. Academic work on segrega- answered to implement these segrega- continued on next page April 2001 • The Brookings Institution • Survey Series 2 which blacks are segregated from this measures above 0.6 are thought to Later in this essay, we will discuss broader group. For ease of historical represent hypersegregation. current patterns of segregation across comparison, we will focus on black- The isolation index captures the cities. At this point, we will reiterate non-black comparisons. percentage of black residents in the that the cities that are highly segre- This distinction matters because census tract where the average black gated with one measure tend to be blacks are less segregated from resident lives, corrected for the fact highly segregated with all measures. Hispanics than they are from other that this number increases mechani- Thus, we will generally restrict non-hispanic whites. This is an inter- cally with the black share of the ourselves to looking at the dissimilarity esting change from the early 20th overall MSA population. A metropol- measure. century, when blacks were more segre- itan area isolation index of 0.5 gated from ethnic immigrants than indicates that the average black III. Findings they were from native whites. We will resident lives in a census tract in also engage in a somewhat unorthodox which the black share of the popula- A. Overall, segregation levels application of the usual segregation tion exceeds the overall metropolitan between blacks and non-blacks approach. We will look at segregation average by roughly 50 percent. An continued their 30-year decline and for the entire country and for the four index of 0.3 reveals that the average are now at their lowest point since census regions, treating them as if black resident lives in a census tract in roughly 1920. they were cities. In this case, we can which the black share of the popula- Cutler, Glaeser and Vigdor (1999) ask both about integration within tion exceeds the overall metropolitan assembled a comprehensive data set metropolitan areas and integration average by roughly 30 percent. Again, on segregation from 1890 to today across metropolitan areas. We think it can range from something close to (it is posted at www.nber.org and that this is an important way to zero (if each black person lives in an www.pubpol.duke.edu/~jvigdor/ approach to the changing level of inte- integrated census tract) to one (if all segregation).