Economic and Racial Integration Through School Choice in New York City

Introduction

The decades since the Supreme Court’s landmark 1954 ruling in Brown v. Board of Education, which declared that separate schools were “inherently unequal,” have brought on a series of policies aimed at integrating schools across the country. Although many states and school districts resisted change for several years, Congressional passage of civil rights legislation and subsequent court decisions expanding district-level desegregation policies led to peak levels of within-district integration by the mid-1970s (Reardon & Owens, 2014).

By some measures, New York State has the most segregated schools in the country (Kucsera & Orfield, 2014). While much school segregation in New York occurs between school district boundaries, a significant portion occurs within individual districts. This is particularly true within the state’s largest school district—New York City—and the 32 sub-districts (known as Community School Districts) that comprise it. While the New York City public schools enroll more than 1 million racially, ethnically, and socioeconomically diverse students, few schools reflect the diversity of the city (Mader & Costa, 2017). As in many other districts, the distribution of students in New York City Department of Education (NYCDOE) schools has been largely influenced by housing patterns. The high level of educational segregation in New York City reflects the high level of residential segregation.

Research has found that New York City’s school-choice policies may exacerbate segregation across the city (Mader et al., 2018). Nearly one out of five NYCDOE middle school students attend an academically screened school that considers factors such as attendance, behavior, grades, and test scores for admissions (Hemphill et al., 2019). The result is that the top screened middle schools, which often feed the city’s top high schools, admit a higher proportion of White, Asian, and high-income students, creating what has been referred to as a “segregation pipeline” (The Hechinger Report, 2018).

In recent years, the city has initiated efforts to reduce segregation in public schools. The NYCDOE started the Diversity in Admissions pilot project in 2016, released a citywide diversity plan in 2017, and launched the School Diversity Advisory Group (SDAG), which released two sets of recommendations in 2019. Two school districts at the vanguard of integration efforts have been New York City’s Community School District 3 in Manhattan and Community School District 15 in Brooklyn. These districts are among the most racially and economically diverse of New York City’s 32 community school districts. They are also among the most segregated, according to various measures described below. Through local efforts—supported by the city and the state—Districts 3 and 15 developed two of the state’s first district-wide integration plans. Both districts used a “controlled choice” admissions process to integrate middle schools beginning with students entering sixth grade in the 2019–2020 school year, and both districts chose to focus primarily on economic integration. We use difference-in-differences and placebo analysis to quantify the magnitude of changes in segregation and the likelihood that these changes occurred by random chance following these policy changes.

School Choice in District 3 and District 15

New York City’s District 3 is located in Manhattan and includes the neighborhoods of the Upper West Side, Morningside Heights, and a portion of Harlem below 122nd Street. District 15 is in Brooklyn and includes the neighborhoods of Park Slope, Winsor Terrace, Red Hook, and Sunset Park, among others. Both districts are racially and economically diverse, reflecting the diversity of both New York City and New York State. Both districts have a sizable share of students in each major racial or ethnic group. District 3 is composed of 0.6% American Indian, 6.7% Asian or Pacific Islander, 27.9% Black or African American, 30% Hispanic or Latino students, 4.4% Multiracial, and 30.3% White students. By contrast, District 15 has 0.4% American Indian, 17.3% Asian or Pacific Islander, 9.0% Black or African American, 35.8% Hispanic or Latino, 3.9% Multiracial, and 33.6% White students. With roughly 48.8% in District 3 and 52.6% in District 15 economically disadvantaged, Districts 3 and 15 are more affluent than New York City overall, where nearly 74% of students are considered economically disadvantaged. For more about how these districts compared with New York City and New York State as a whole, see Supplementary Appendix B in the online version of the journal.

Middle schools in District 3 and District 15 use the deferred acceptance algorithm to allocate first-round offers to students where demand exceeds available enrollment space (for a detailed description of the deferred acceptance algorithm, see Pathak, 2017). Two important features of the algorithm are that students and parents rank up to 12 school choices for which they are eligible in the fall semester prior to their enrollment year, and students’ optimal ranking strategy is to rank schools in their true preference order.

Among students who have chosen a school first—or chosen a school second through 12th and not gotten into a higher choice—student enrollment offers are based on prioritization groups and then school rankings. In the 2018–2019 enrollment year, 7 of 20 schools in District 3 gave priority to students who were enrolled in the elementary program in their school, and five of those gave second priority to any students living within the school’s elementary zone. Only one school in District 15 included a priority group in 2018–2019, with 40% of their seats going to students who were eligible for free or reduced-price lunch.

Within these priority groups, there are two types of school ranking methods: screening and lottery. Under screening, schools can order each student applicant from highest to lowest admission rank based on their grades, test scores, attendance, and possibly other factors that are not well known. Under lottery assignment, each person is assigned a random number for the whole admissions process across all schools. Students with the lowest lottery numbers are given the highest preference. In the 2018–2019 enrollment year, 16 of 19 District 3 schools and 10 of 11 District 15 middle schools used screening. Each school used several criteria to screen. Among them, in District 3, 11 schools used course grades, 10 used state math and reading test scores, and 13 used attendance to screen. In District 15, 5 used course grades, 3 used state math and reading test scores, and 5 used attendance to screen.

For its 2019–2020 integration plan, District 3 revised its middle school admissions system to prioritize 25% of sixth-grade seats for students from low-income families with low academic performance. Students were considered low-income if they met the income threshold to qualify for free or reduced-price lunch (FRPL). Low performance was defined based on a performance index that gave 30% weight to English language arts (ELA) course grades, 30% weight to math course grades, 20% weight to NYS ELA test scores, and 20% weight to NYS math test scores. Low-income students were divided into two groups based on the performance index. Ten percent of seats were reserved for the lowest performing FRPL students and 15% of seats were reserved for the next lowest performing group of FRPL students. The remaining 75% of seats were available to all students subject to the same priority grouping from the 2018–2019 enrollment year. In addition, all schools with screens retained them.

District 15’s plan differed from District 3’s plan in several important respects. First, rather than targeting students with low academic performance, District 15 prioritized seats for low-income students, English Language Learners, or students in temporary housing. Second, District 15 set significantly more ambitious targets, prioritizing economically disadvantaged students for 52% of sixth-grade seats. Finally, District 15 removed academic screens from all non-charter middle schools that had them in the district.

Measuring Segregation

To measure within-district segregation, we use the mean absolute percentage point difference between school and district. This measure of segregation, described below, is easy to calculate and can be applied identically to multiple levels of segregation (e.g., between-district, within-district, within-school) and to measures with two or more groupings (e.g., multiple races). For simplicity, we refer to this measure as the segregation index. Specifically, a district’s segregation index for a particular group of students, m, is calculated as:

S m = ∑ i = 1 n τ i T | p i , m − P m | × 100 ,

where n is the number of schools in the district, τ i is the total number of students in school i, T is the total number of students in the district, p_i,m is the proportion of students in group m in school i, and P_m is the proportion of students in group m in the district. Conceptually, the segregation index can be interpreted as how far (in percentage points) a typical school is from the district proportion of students in a particular group.

The index can also be adapted as a population-weighted average of group-level measures to handle multiple groups simultaneously (e.g., multiple race and ethnicity groups):

S ¯ = ∑ m = 1 M τ m T S m ,

where M is the number of groups, τ m is the total number of students in group m, and S_m is the segregation score for each group. For six racial/ethnic groups, for example, S ¯ would be the mean segregation score for each race/ethnicity, weighted by the proportion of students of that race/ethnicity in the district.

Impact on Economic Segregation

To assess changes in segregation in Districts 3 and 15, we estimate the segregation index for each year in each district in sixth grade. As a first approximation of the change, we take the difference between 2020 and 2019. To control for possible district-wide changes in demographics or school enrollment patterns that may be unrelated to the integration plans, we also estimate this change in segregation for seventh and eighth grade in the same districts, average those differences, and difference this average difference from the sixth-grade difference. As a placebo check, we compare these difference-in-difference estimates to those in 40 of the 50 largest districts. From the 50 largest districts, we exclude District 1 which included a desegregation plan outside of sixth grade and districts that have less than 80% of sixth-grade students in the same schools as seventh- and eighth-grade students. In the Supplementary Appendix B in the online version of the journal, we show that our results are not sensitive to these exclusions.

Figure 1 shows the economic segregation index over time in Districts 3 and 15. In District 3, we see little change in segregation after the policy was initiated, with the sixth-grade segregation index dropping from 28.0 in 2019 to 25.7 in 2020, a decline of 2.3 points or 8%. In District 15, we see a large change in segregation, with the sixth-grade segregation index declining from 23.4 in 2019 to 10.6 in 2020, a decline of 12.8 points or 55%. While we focus on the segregation index because it is simple and policy-relevant, we observe similar changes using other measures of unevenness common in the segregation literature, including the dissimilarity index (Sakoda, 1981) and Theil’s H (Theil & Finizza, 1971). See Supplementary Appendix A in the online version of the journal for details about alternative segregation measures and how they compare with the segregation index.

OPEN IN VIEWER

Figure 1.

Economic segregation in District 3 and District 15 in sixth grade compared with seventh- and eighth-grade average by year.

To control for possible district-wide changes in demographics or school enrollment patterns that may be unrelated to the integration plans, we also measure the segregation index for seventh and eighth grade in the same districts. In District 3, while the sixth-grade segregation index fell by 2.3 points (as shown earlier), the average of the seventh- and eighth-grade segregation indices fell by 1.7 points. The difference in the differences was −0.6 points, as the sixth-grade segregation index fell by 0.6 percentage points more than the average of the seventh and eighth-grade indices. In District 15, the corresponding difference-in-differences result was −12.2, as the average of the seventh- and eighth-grade segregation indices fell by only 0.6 points while the sixth-grade segregation index fell by 12.8 points. In District 15, sixth-grade segregation fell by substantially more than seventh- and eighth-grade segregation, consistent with the change in the sixth-grade admissions system causing the changes. The same was not true in District 3. Compared with the 40 large placebo districts in NYC and the rest of NY state, the decline in District 15 was the single largest absolute value difference-in-differences, while District 3’s change was only larger than 16 of those 40 districts. In fact, the decline in District 15 is the largest single year decline in segregation in the sixth grade relative to seventh and eighth grade among all of the included 40 large districts from 2012 to 2020, as shown in Supplementary Appendix B in the online version of the journal.

Impact on Racial Segregation

Both district plans targeted economic segregation. As race is correlated with economic disadvantage, however, it is plausible to assume the plans may have affected racial segregation as well. Overall, the effects for District 3 are similar to those for economic segregation. As shown in Figure 2, both an enrollment-weighted average of racial segregation and Black, Hispanic, and White segregation did not change much from 2018 to 2019, either in absolute terms or in comparison with seventh- and eighth-grade averages of segregation.

OPEN IN VIEWER

Figure 2.

Impact of integration plans on sixth-grade racial segregation in 2019–2020.

District 15, however, had declines outside the normal for the average, Hispanic and White segregation. Black segregation declined as well but the decline was closer to the middle of the 40 large districts. On average, racial segregation declined by 38% from its 2018–2019 baseline. The fall in Black, White, and Hispanic segregation was 26%, 40%, and 42%, respectively. The difference-in-differences estimates back up these declines, both overall and for White and Hispanic students. The exception is for Black students who make up a relatively small share (9%) of sixth-grade students in District 15 and had relatively low pre-implementation segregation (ranking 22/41 of the largest districts in New York, compared with 2/41 for White segregation and 3/41 for Hispanic segregation). Results on the average race measure are backed up by synthetic control methods which are found in Supplementary Appendix B in the online version of the journal.

Enrollment Before and After the Change

One question often asked of desegregation plans by policymakers and observers is whether students of greater economic advantage leave the district in response to desegregation. This harkens back to prior increases in diversity and integration that appeared to lead to a phenomenon nicknamed “white flight” for the increase in between-district segregation it created (Coleman et al., 1975). While it is early to assess the long-term effects of middle school integration plans on enrollment, we can examine the 1-year changes in enrollment patterns for clues as to whether advantaged students appear to be “fleeing” District 3 or District 15 due to changes in policy.

Figure 3 shows the economic and racial breakdown of sixth graders in District 3 and District 15 from 2011–2012 enrollment years to 2019–2020. Overall, there are few changes in the economic or racial breakdown of students between 2018–2019 and 2019–2020, during the first year of policy implementation. In both districts, the share of White students increased slightly and the share of economically disadvantaged students changed little. Neither of these changes were large compared with typical year-to-year variation and the racial changes operate in the opposite direction of what you would expect if “white flight” was taking place. Worth noting is that enrollment in sixth grade for 2020 drops relative to 2019. Trends in enrollment by cohort in District 15, however, reveal that the cohort that entered sixth grade in 2018–2019 was unusually large prior to this change, and therefore the decline in enrollment in 2019–2020 is unlikely to have been due to policy changes.

OPEN IN VIEWER

Figure 3.

Enrollment by race/ethnicity and economic disadvantage.

While the enrollment patterns indicate little “flight” from these two districts, they are clearly short-term, and we think continued follow-up is warranted. It could be that as parents acclimate to the new normal, the flight may increase. One factor that might affect future enrollment is how much choice is preserved under the integration plan, and we can begin to get a sense by looking at the share of students admitted to one of their top choices. According to a New York City Department of Education analysis of the school-choice data, in District 15, the share of students getting their first choice increased from 48% in 2018–2019 to 51% in 2019–2020, while the share of students getting one of their first three choices declined from 87% to 78% over the same time period (NYCDOE, 2019).

Conclusion

New York City’s Community School Districts 3 and 15 are, in some ways, microcosms of the city and the state. Both districts are economically and racially diverse, yet highly segregated. In 2019–2020, both districts launched integration initiatives that targeted incoming sixth graders and sought to make their middle schools better reflect the diversity in each district. The results of these efforts differed dramatically and can provide insight to other school systems seeking to improve integration in a school-choice framework.

District 15 saw a substantial decrease in segregation in the first year of its integration plan. In sixth grade, economic segregation in District 15 decreased by 55% and racial segregation decreased by 38%. These results are statistically significant and robust to numerous alternative methods of analysis. By contrast, District 3 saw, at best, a small decrease in segregation. In sixth grade, economic segregation in District 3 decreased by 8% and racial segregation decreased by 5%, changes that are well within the bounds of typical year-to-year fluctuations.

At first, the divergence in these results may be surprising due to the broad similarities in the districts’ integration plans: Both districts used a controlled-choice admissions mechanism to target reduced economic segregation among sixth graders entering middle school. However, the plans differed in two important details. First, District 15 removed academic screens from all middle schools, admitting students by lottery within priority groups. District 3, by contrast, retained all academic screens. Second, District 15 set a significantly more ambitious target for enrolling economically disadvantaged students, prioritizing low-income students for 52% of sixth-grade seats, compared with 25% of seats in District 3. We believe both changes led to less segregation in District 15, but it is difficult to determine how much each is responsible for, as both were changed in tandem in the same district. For further discussion on this point see Supplementary Appendix C in the online version of the journal.

One clear lesson from this study is that the details of an integration plan matter. Simply introducing a “controlled-choice” admissions policy, or prioritizing economically disadvantaged students, is not enough to ensure a meaningful change in segregation. As other districts in New York and across the country look to better integrate their schools, they must carefully consider the specific design decisions they make when developing plans to reduce segregation.

A second—and simpler—lesson is that integration is possible. Through the combined efforts of many people, District 15 developed and implemented a policy that dramatically decreased economic and racial segregation in sixth grade. In addition, initial results suggest there was little to no flight from the district as a result of this desegregation plan.

There are several limitations to this study. First, the results are limited to a single year in a single grade, and various factors may mitigate the effects of District 15’s desegregation policies in future years. Residential patterns may change, families may become better at gaming the assignment process, or some unknown factor may change how effective these policies are in practice. Second, because District 15 both dropped screening and prioritized economically disadvantaged students simultaneously, the relative contribution of these two mechanisms cannot be determined in this study. Finally, the external generalizability of our findings is uncertain. We cannot yet know the impact of similar policy changes in other districts with different characteristics and norms. As districts in New York and elsewhere continue to implement school integration plans, further research can begin to address these limitations.

Supplemental Material