It feels like there’s been an increase recently in the number of pieces that take a nationwide look at some aspect of school segregation. Of course, Ed Build has done several prominent national scans of school district secession and school funding inequity. Sean Reardon and colleagues just came out with an exploration of segregation and achievement gaps across the country (more about this in an upcoming post). Vox.com also has several interactive tools on housing and workplace segregation as well as attendance zone boundaries. There are certainly others that I’m missing.

This volume of activity speaks both to the expanding public interest in contemporary school segregation and the wide variety of methods for measuring/visualizing a complex topic that surfaces in many different ways. It’s part of a debate about segregation itself as well as debate about how to measure it. And, the new series and interactive map from the Washington Post contributes in new and interesting ways to both questions. The series just released its first set of articles, though there is more that is forthcoming.

If you aren’t familiar with this series, I definitely recommend checking it out (especially this interactive map) and then coming back here. I don’t have the space to sufficiently summarize the key findings, though most important thing to know is that the series sought to identity districts that “have enough diversity to integrate” and then measured whether or not they were “integrated” within their boundaries. Below, I have a few thoughts about district-level analyses of segregation. But, first, I want to dive into some of the details about how the series measured segregation. 

I’d like to acknowledge the Boston Indicators project, especially director Luc Schuster, for the email and phone conversations that form the basis of what follows, which include email clarifications from one of the reporters on the piece. This exchange went a long way towards helping me understand some of the less-obvious ways that the numbers and ratings were developed for this very important project. Where relevant, I quote from that email exchange. In addition, shortly after publishing this post, one of the lead authors, Kate Rabinowitz, replied with clarifications in a twitter thread. I’ve updated the post accordingly.

Overall, I thought that readers might be interested in learning what I learned. Here’s my best attempt to put everything together, organized by the questions that I had after reading. 

How did they identify districts that “have enough diversity to integrate”?

This part is straightforward: 

• “any district where one race is over 75% of the student population does not receive an integration score because integration is not feasible.”

The 75% threshold was based on a review of the literature and discussion with academics in the field. It is important to note that this includes districts where Black and Latinx students together compose more than 75% of the population. Chicago is a good example – neither sub-group on its own is greater than 75% of the overall population; so, the district is labeled “diverse.” But, Black and Latinx students combine to account for 82.2% of the total district enrollment (based on 2017 data). 

I don’t have space to fully address the question of whether this should or should not count as diverse, but I appreciate that pieces like this bringing these important questions to a wide audience. I’d of course be curious to hear readers’ thoughts on this. 

What, exactly, is the “variance ratio” and how was it used in this series?

Once they identified “diverse” districts, they used a “variance ratio” to measure whether schools were integrated within the district borders. Essentially, the variance ratio calculates “how isolated black and Hispanic students are, given the overall racial makeup of the district.”

They link to this page from the US census which includes more details about measures of housing segregation. The variance ratio is also referred to as the correlation ratio, or eta-squared, from the census page. Unlike measures of exposure, the variance ratio accounts for the overall demographic makeup of a school district. So, a school with 80% Black students might get a low exposure index (because there isn’t much exposure to students of different races), but it might be scored as integrated according to the variance ratio if the district at large was composed of 80% Black students. On twitter, Kate also explained that the authors decided to use the variance ratio because “exposure indices don’t capture or control for demographic change over time,” which was obviously a key goal of this project.

Meanwhile, although measures of evenness account for overall demographic makeup of a district, these measures (i.e., dissimilarity and entrophy) can only account for one racial subgroup at a time – dissimilarity compares one group to another, while entrophy compares multiple groups to each other. Under these measures, a district might get credit for mixing white and Asian students, even if these groups were segregated from Black and Latinx students. 

Instead, the series uses the variance ratio and combines Black and Latinx students into one group. It compares their combined enrollment to a group composed of all other students. As described in the series, “The variance ratio was computed for black and Hispanic students because of the history of exclusion and achievement gaps faced by these groups.” This means that both the following statements could be true:

• A totally integrated school would have a black-hispanic isolation measure equal to the district-level isolation measure.
• Schools where the population are equally black and Hispanic were not classified as integrated.

Note: Kate clarified that the truth of the second statement “depends on other factors, like the proportion of students of other races and district demographics.”

Taken together, this means that an individual school with 40% Black students and 40% Latinx students would not be considered “integrated” even though Black and Latinx students would have a lot of exposure to each other. Instead, the measure of integration turns on whether the combined percentage of Black and Latinx students matches the district demography. If that same school was in a district that is 80% Black and Latinx students, it would be considered “integrated.” It’s a perhaps complicated/in-the-weeds distinction, but I think it’s important. 

There are a lot of resources for anyone who wants to dig into this in more detail:

• The series itself has a short article that compares its approach to other common methods of measuring segregation.
• Tomas Monarrez – whose research led to the Vox article on attendance zone boundaries – has a very useful write-up on segregation and isolation measures here, including equations.
• The reporters on this story posted the code and output data here.  

What is captured in a district-level analysis of school segregation?

First, as Kate explained on twitter, this is part of an ongoing series, which will explore different aspects of segregation. So, although the overall series isn’t limited to district-level analyses, I thought it provided a good opportunity to look at what this kind of analysis can and cannot tell us about contemporary segregation.

Focusing within districts is certainly useful. As outlined in the series, this approach led to the following important finding: “Millions more American children are attending school with students of other races, even as many urban schools remain deeply segregated.” They repeated this analysis for elementary schools and found the same results, which is interesting. Elementary schools are typically more segregated than secondary schools, in part because there’s likely to be more of them in small districts, whereas there might be just one middle school or high school. So, this speaks to the strength of this finding: if children are attending schools with students of other races – even at the elementary level – then this is clearly a major demographic change.

Of course, these kinds of demographic changes are happening in many places across the country. Boston Indicators, for example, did a report on the diversification of Greater Boston, finding that, between 1990-2017, every single one of Greater Boston’s 147 cities and towns saw an increase in their non-white population shares.

The diversification of suburbs is a major, major development in the contemporary landscape of school segregation. It comes with many opportunities for school diversity as well as many important social and educational challenges, especially if school level changes occur without deliberate attention to inequities in things like school discipline or in curricular exposure to different cultural histories. Pieces like this point toward the urgency to address these issues. And, since they focus on within-district forms of segregation, they point to solutions that are perhaps more politically feasible, given the restrictions on cross-district integration outlined in the 1974 Milliken decision. (See this great recent podcast for more info on Milliken and its relevance today.)

No measure is perfect. What is left out of this approach?

Despite these strengths, I worry that district-specific measures may inadvertently redefine “integration” in a way that does not match students’ lived realities in schools. Take, for example, a school that is 50% Black and Latinx in a district whose overall Black/Latinx population is 90%. As Luc pointed out, I *think* that, according to the measures used in this series, that would count as a segregated school, even though it’s far less segregated than the overall district. Students in that particular school would be getting a different experience than others in the district. 

Relatedly, the methodology may count homogenous schools as “integrated” if they are in homogenous districts. For example, the piece itself highlights Lexington, MA as a “highly integrated school district.” Lexington, however, is just 7% Black and Latinx – its schools match the overall district demography because it’d be hard to racially segregate such a small number of students in a system that has just two middle schools and one high school.   

This gets to a larger concern that I have about within-district analyses of integration. Focusing too much on this kind of measurement may not only overlook important trends in segregation across districts, but also might overlook the fact that school/housing policies are largely responsible for the racial sorting that created overwhelmingly white suburban rings around majority non-white cities. Following the Milliken decision, municipal boundaries essentially functioned as walls against inter-district integration for families who left the city. This is especially true in Northeastern states (e.g., my home city of Boston, which is near Lexington), where school district boundaries align with municipal boundaries, as opposed to the county-wide school districts that are more common in other parts of the country. 

And, it has major policy implications. This recent piece, for example, highlights differences in how the state of Connecticut has responded to segregation in two very different communities. One school – in Hamden – is disproportionately segregated compared to its overall district demographics. Meanwhile, another school – in New Haven – is also segregated, but compared to schools in the region beyond the district borders. The author, Christopher Peak, notes: 

• “Racial imbalance, like at Church Street [the Hamden school], comes with swift consequences if white suburbanites confine themselves to specific schools within their districts.” 
• “Racial isolation, like at Lincoln-Bassett [the New Haven school], takes decades-long lawsuits to set up a voluntary system for only some of a district’s schools, as New Haven’s magnets have attempted to do.”

The Hamden district is required to submit a corrective action plan to the state. Meanwhile, for the students in New Haven, “absent a court order requiring it to do more, the state steps in only if the black and brown neighbors who are being kept out of schools already live within the town.”

It’s not a new argument, but it’s worth repeating – something is clearly wrong here. That article quotes Gary Orfield who says: “I don’t think it makes any sense to define segregation based on the demographics of the district,” he said. “What’s the situation that students are actually confronting? If you’re isolated in a high-poverty school, it doesn’t matter what district you’re in.”

How did the series rate your school or district? Does it seem right to you? I’d love to hear your thoughts!