Previous | Front Page Download report |
Next |
To facilitate comparisons between residential segregation and school segregation, we
aggregate schools into metropolitan regions, and then separately into the central city and
suburban portions of those regions. Therefore our results do not measure segregation
within particular school districts, but more broadly the disparities between schools in
the same geographic area regardless of district boundaries. We recognize that schools are segregated not only by race but also by class. To take class segregation into account, we take advantage of the NCES reports of student eligibility for free or reduced-price lunches. To be eligible, a family must fall below 185% of the poverty line - around $32,000 for a family of four in 1999-2000. Though some states still do not provide information on the school lunch program, this is the best available indicator of the income level of students in public schools. How We Measure Segregation The standard measure of segregation is the Index of Dissimilarity (D), which captures the degree to which two groups are evenly spread among schools in a given city. Evenness is defined with respect to the racial composition of the city as a whole. The index ranges from 0 to 100, giving the percentage of children in one group who would have to attend a different school to achieve racial balance - one where every school replicates the group composition of the city. A value of 60 or above is considered very high. For example, a D score of 60 for black-white segregation would mean that 60% of either group must move to a different school for the two groups to become equally distributed. Values of 40 to 50 are usually considered moderate levels of segregation, while values of 30 or less are considered low. This report also refers to information on neighborhood segregation. For this purpose we use data from the U.S. Census of 1990 and 2000 specifically for the under-18 population. Neighborhood segregation is measured at the level of census tracts, areas that usually have 3000-5000 residents. Standards for evaluating change in dissimilarity scores In our analysis, we interpret change either up or down on the following criteria:
Exposure and Isolation Indices |
Previous | Front Page Download report |
Next |