Measuring Segregation

Segregation can be measured in many ways. Segregation measures shown in the interactive map represent the two-group normalized exposure index, which measures the difference between two groups’ exposure to one of the groups. For example, the White-Hispanic normalized exposure index for school segregation compares the proportions of White (or Hispanic, equivalently) students in the average White and Hispanic students’ schools (the measure for residential segregation compares the average compositions of two groups’ Census tracts). A White-Hispanic normalized exposure value of 0.5 would indicate that the proportion of White students in the average White student’s school is 50 percentage points higher than in the average Hispanic student’s school. (The two-group normalized exposure measure ignores the presence of other groups aside from the racial dyad of interest.) The normalized exposure index ranges from 0 to 1. A value of 0 implies no segregation — the two groups have equal exposure to one group (all schools have identical proportions of the two groups). A value of 1 implies complete segregation— the two groups have no exposure to one another (no Hispanic student attends a school with any White students, and vice versa). For more information on this measure, see our brief.

The downloadable data files (Get the Data), provide additional segregation measures. In addition to normalized exposure (shown in the maps), the files include two other two-group evenness measures: the dissimilarity and information theory indices. The files also include two exposure measures, other-group exposure and same group isolation. Finally, two measures of income segregation across multiple income categories are included: the rank-order information theory and relative diversity indices.

In addition to racial and economic segregation measures, the data include racial-economic segregation—defined as the difference between two racial groups in their schools’ or neighborhoods economic composition. For school segregation, this measure is the difference in the average free lunch eligibility (FLE) rate in each group’s schools. For example, a Black-White racial-economic segregation value of 0.5 implies that the FL eligible rate in the average Black student’s school is 50 percentage points higher (0.5 higher on a 0 to 1 scale) than in the average White student’s school (e.g., the average Black student attends a school where 75% of students are FLE while the average White student attends a school where 25% of students are FLE). For residential segregation, this measure is the difference in the average poverty rate in each group’s neighborhoods.

For more information on how we estimate segregation, see our documentation on the Get the Data page.

Data Sources

School segregation estimates draw on school-level enrollment data from the Longitudinal Imputed School Dataset (LISD) 1.0, created by the Segregation Tracking Project team. The LISD combines, cleans, and imputes missing and erroneous data from the National Center for Education Statistics Common Core of Data (CCD) since 1991. The LISD comprises 10 imputed datasets. Segregation measures are estimated separately in each imputed dataset and then averaged over the 10 datasets; this ensures that the segregation measures reflect the true underlaying variation in school composition. Moreover, because schools enroll students in different age ranges, it is not appropriate to simply compute segregation between schools by treating all students as potentially transferable among schools. Instead, we compute segregation within each grade level in a unit and then average the segregation values across grades (weighting by the number of students in each grade) to produce a single average segregation measure for each geography of interest. We estimate segregation only among “regular” (as opposed to special education, vocational, or alternative), non-virtual schools. Schools exclusively for the deaf and blind are not included in segregation estimates. More information about the LISD is available at the Get the Data page.

Residential segregation estimates are derived from Census tract-level data on residents from the U.S. Census Bureau’s 1970 to 2020 decennial censuses and American Community Survey (ACS) five-year aggregate estimates from 2005-2009 through 2018-2022 . Neighborhoods are defined as census tracts (prior to 1990, the Census did not delineate tracts outside of metropolitan areas). The ACS trades annual frequency for smaller samples, so tract-level estimates require five-year aggregations. Census and ACS data provide tract-level counts of individuals by race, poverty status, and household income. We estimate residential segregation among all people and among school-relevant populations like children (by race) or families with children (by income since 1990).

Racial and Economic Groups

School segregation estimates rely on racial/ethnic categories are defined by the CCD. White, Black, Asian, Native American, and Multiracial (identified since 2010-11 for all schools) refer to non-Hispanic students. Asian includes Asian, Hawaiian, and Pacific Islander students. School economic composition is defined in terms of eligibility for free or reduced-price lunch. Free lunch or reduced-price lunch eligible students are those whose families earn less than 130% or 185%, respectively, of the poverty threshold for their family size.

Residential segregation estimates use categories defined by the Census; for consistency between the school and residential segregation measures, we use the same racial/ethnic categories as in the CCD when possible. In 1970, racial groups are not defined with respect to Hispanic ethnicity; e.g., White residents include both those that identify as non-Hispanic and Hispanic. Multiracial individuals are identified since 2000.

We estimate residential economic segregation in two ways. First, we compute segregation among households by poverty status–whether a household’s income is above or below the federal poverty threshold for their household size; second, we compute rank-order measures of income segregation using the counts of households in each of 15-25 income categories (e.g., $0 to $9999, $10,000 to $14,999, etc).

Geographies

We estimate segregation between schools and neighborhoods within school districts, counties, commuting zones, metropolitan areas, states, and the nation. We also estimate segregation between school districts within all larger geographies. We provide a geographic crosswalk that links schools to larger geographies. Neighborhoods (tracts) are nested within counties, which link to commuting zones, metropolitan areas, states, and the nation. We link neighborhoods to school districts using geographic crosswalks that use population weighting to assign tracts’ populations to the school district(s) in which they are located. See the Get the Data page for data and documentation.