This is the methodology and data write up of the publication “Landscape of Diversity in D.C. Public Schools” found here.
The data can be found here.
In this report, we estimate racial and ethnic diversity as well as economic diversity. We identify the most diverse schools and compare their attributes to other schools. We also look at changes from 2014-15 (the first year that at-risk data are available) to 2016-17.
To measure diversity, we consider the size of the non-plurality share of a student group at a school. We do not focus on how this relates to the school’s neighborhood because there is a high degree of public school choice (just 27 percent of students attend their in-boundary traditional public school). In theory, this permits D.C.’s schools to be more integrated than our neighborhoods. In addition, D.C. is a small city geographically, which allows students to travel to schools in a large proportion of the city. For example, a student traveling the average distance to school for charter school students of 2.1 miles has access to roughly 20 percent of the city’s area.
Measuring racial and ethnic representation
The substantial presence of three student groups in D.C. means that our measure cannot focus only on one majority and one minority group. We considered a few established methods to measure racial and ethnic representation, but none met our needs. The exposure index measures the extent to which students from one race are around students from other race, but this would ignore one of D.C.’s primary student groups. The isolation index measures how much a single race is clustered in one school, but this would highlight only one group and give an idea instead of which schools are the least diverse. The dissimilarity and divergence indices show how well the racial composition of a school relates to the neighborhood, but because of the student body at D.C.’s public schools, these measures would only identify schools with a majority African American enrollment as diverse. Lastly, the Theil index can compare multiple groups but is both complicated and difficult to interpret.
Simply looking at the share of the plurality group (or the group with the highest percentage of students) and non-plurality group(s) at each school will give the clearest idea of which schools have groups represented more equally. This also corresponds to the idea of a threshold of no more than 70 percent representation from one group to enable a diverse learning environment (Potter and Quick 2018). To measure racial and ethnic diversity, the group in the plurality is identified and the percentages of students in the other groups are summed to calculate a measure of racial and ethnicity diversity. Racial and ethnic groups include African American students, Latino students, white students, and others. The diversity score has a maximum value of 75 percent in theory, which would occur if each of the four groups were equally represented, and a minimum value of zero. However, given D.C.’s demographics, the median racial and ethnic diversity score would 32 percent in 2016-17 if all students were distributed equally.
Diversity will be greatest when the score is highest, and the measure treats all groups equally without prioritizing a mix of historically advantaged and disadvantaged groups. For example, a school with a student body that is 50 percent Latino and 50 percent African American would be considered just as diverse as a student body that is 50 percent African American and 50 percent white. And a school that is 40 percent African American, 50 percent Latino, and 10 percent white would have the same diversity score (50 percent) as a school that is 50 percent white, 25 percent African American, and 25 percent Latino.
Methodology Figure 1 highlights a few examples. A school where the majority of students (white students in the figure below) holds 60 percent of the student body would have a diversity score of 40 percent, or the sum of other groups, and be the most diverse out of the examples below. A school that is most representative of D.C.’s students overall would have a diversity score around 30 percent, as most public school students are African American. A school with only one group (likely African American students), which reflects half of D.C.’s public schools, would be the least diverse of these three examples.
Measuring economic representation
To measure economic diversity, the analysis identifies whether students who are at-risk or not at-risk have a plurality, and uses the percentage of students in the other group as a score of economic diversity. In D.C., almost half (47 percent) of pre-kindergarten through grade 12 students are identified as at-risk. The percentage of students who are at-risk1 is a better metric than economically disadvantaged students (or the percentage of students receiving free or reduced price lunch) in D.C. given data complications. In D.C., almost three-quarters of schools meet the requirements for the Community Eligibility Provision that provides all students with free lunches without submitting FARM applications. This means that data on economic disadvantage are limited. The economic diversity score has a maximum value of 47 percent if each group was represented at every school exactly as they are in the student body, and a minimum value of zero. However, individual schools can have a score of up to 50 percent if groups are evenly distributed at the school level.
The greater the economic diversity score, the more economic diversity at a particular school. Methodology Figure 2 shows examples of economic diversity. Schools with an economic diversity score of 50 percent will be the most diverse, as these schools will have the most parity between students who are at-risk and those who are not at-risk. A school with a plurality of at-risk students at 60 percent would be the next most diverse at 40 percent. A school with a high concentration of at-risk would be less diverse with a score of 10 percent.
Identifying diverse schools and their attributes
Once we have measures of racial and ethnic diversity and economic diversity, we will identify which schools are the most diverse and which characteristics they share. The most diverse schools will have distributions that represent student groups more equally, and the highest diversity scores as defined by the 75th percentile. We will then perform statistical tests of significance between the group of the most diverse schools and other schools to see if they are different across various school characteristics (separately in terms of race and ethnicity, and at-risk population). Specifically, we will use Welch’s t-tests for samples with unequal variances and sample sizes. We are interested in school characteristics related to location, sector, enrollment, program offerings, proximity to transit, grade band, boundary participation distribution of students by ward (see Methodology Table 1 for data sources on school characteristics). We will combine data from local education agencies (OSSE, DCPS, PCSB) to conduct this analysis.