Charles Weber ’17

Harvard College

March 2017



Does the end of racial integration affect the racial demographics of school districts and the test scores of students? I analyze the impact of the 2007 Supreme Court decision in Parents Involved in Community Schools v. Seattle School District, which struck down Seattle Public School’s racial integration plan. The court ruling’s sudden and unexpected nature allows for a quasi-experimental interpretation of demographic changes and student outcomes after the court ruling. Using a variety of difference-in-differences models and placebo testing, I find the court ruling had two main effects: (1) white families were 20 percent more likely to live in Seattle after 2007 and (2) the test scores of black students increased at a significantly higher rate than those of white students causing the racial test score gap in math to decrease by 10 percent. These findings indicate the court ruling both mitigated the historical trend of white flight and reduced the racial test score gap. This paper emphasizes the unintended consequences of racial integration policies and the tradeoffs policymakers face when aiming to improve educational equity.




  1. Introduction

The promise of Brown v. Board of Education (1954), that students of all races will attend integrated and equal schools, is waning for this generation’s students. While evidence shows court-ordered desegregation plans of the 1960s and 70s reduced educational inequity (Guryan 2004; Johnson 2011), school integration is decreasingly considered as a viable policy tool today. The US Supreme Court dealt proponents of racial integration a further blow with their decision in Parents Involved in Community Schools v. Seattle School District in 2007. Chief Justice John Roberts illustrated the Court’s belief in a race-blind constitution in his written opinion:

“The way to stop discrimination on the basis of race is to stop discriminating on the basis of race.” (Parents Involved 2007)

The decision declared that primary and secondary public school enrollment schemes must also be race-blind. In effect, the ruling struck down Seattle Public School’s race-conscious student assignment plan just before the start of the 2007-08 school year. The plan allowed rising high schoolers to choose which school they wanted to attend subject to racial balancing constraints. These constraints aimed to maintain racial diversity across schools and were subsequently eliminated after 2007. No attempt to integrate Seattle schools has occurred since the court ruling (Rosenthal & Mayo 2012; Higgins 2013).

Unfortunately for students, legal scholars argue ideology played the main role in determining the court’s decision (Ryan 2007) rather than evidence evaluating the impact of the integration plan on academic outcomes. Regardless of whether the case was decided considering evidence from the school districts and other integration efforts, the ruling likely had real impacts on students and their families. Some argue that this ruling has led to the resegregation of schools in Seattle (Shaw 2008; Ruiz & Ellis 2011). There has yet to be a rigorous analysis of the ruling’s effects on Seattle. This paper is the first to exploit the sudden change in policy in 2007 to determine the causal effect of the court ruling.

The sudden change in policy in Seattle lends itself naturally to a quasi-experimental analysis of the policy’s effects on the demographics of Seattle and student outcomes. I find the end of racial integration had two major effects. First, using yearly demographic and dwelling data from the American Community Survey (ACS), I find that white parents are 20 percent more likely to live in Seattle after the court ruling. This suggests that white families in particular view Seattle more favorably after the court ruling and are moving from other parts of Washington and beyond to enroll their children in the public schools there. In effect, the ruling in Parents Involved mitigated the historical trend of white flight.

Second, I find the test scores of black students increase at a significantly higher rate than the test scores of white students after the court ruling. This in turn reduces the racial test score gap in math by 10 percent. Furthermore, I show the positive test score effects are at least in part due to a growing number of white students in Seattle schools. This indicates that the influx of white families and their social capital into Seattle had test score benefits for black students. There is one caveat to the positive test score effects: While black math test scores increase on average and reading scores are not significantly affected, the reading test scores of black students in schools with the fewest white students drop rapidly after the court ruling. This suggests that any benefits to schools after the court ruling were not equally distributed across all schools.

  1. Background
  2. Estimating the Causal Effect of School Integration

An extensive amount of research has analyzed the potential of school integration as a tool for increasing educational equity. The most legitimate estimates of the causal effect of integration stems from work leveraging the impact of court-ordered desegregation plans in the 1970s. Guryan (2004) uses a difference-in-differences method to show these court-ordered desegregation plans reduced black high school dropout rates by 2-3 percentage points. The same plans had no effect on white dropout rates indicating integration does not need to be a zero sum game. Reber (2010) and Johnson (2015) find similar benefits from the court-ordered desegregation plans but for a wider array of outcomes including substantial gains in educational attainment, college quality, later family income, and declines in the probability of incarceration. More recently, Bergman (forthcoming) leverages a randomized desegregation program to show the program raised college enrollment by ten percentage points for the mainly black and Hispanic participants. Unfortunately, little research analyzing the causal effect of integration on short-term academic achievement exists, but the work discussed above clearly shows court-ordered desegregation plans largely benefit the medium and long-term outcomes of black and minority students and has no effect on white students. This paper elucidates both the causal effect of ending integration in Seattle and its effects on the short-term academic outcomes of black and white students.

There are three possible channels through which integration may benefit students: peer effects, improved resources, and improved social capital (Guryan 2004; Barnum 2016). Research analyzing the differential impact of each channel is limited, but Johnson attributes much of the benefits he finds to improved resources such as reductions in class size and increases in per pupil funding. This is not surprising as the resource disparities between white and black schools before Brown are well documented (Graham 2007). Bergman’s findings echo Johnson’s as the receiving districts that benefited transfer students spent approximately 62 percent more per pupil than the sending districts. Nonetheless, credible research evaluating the causal effect of integration and its channels is lacking. Meanwhile, research showing the importance of peer effects (Wilson 2012) and social capital (Coleman 1988) may indicate we do not have the clearest explanation of why integration benefits some students. Still, it appears school resources and quality certainly matter, something which even Reardon (2016) suggests might be correlated with the segregation measures most strongly correlated with racial test score gaps. Whether school integration has a causal impact on students in today’s schools and if so how remain unanswered questions. This paper’s results begin to give a clearer answer to these questions but further clarity is still needed.

Beyond the apparent benefits of integrated schools for black and minority students, researchers have also documented how white students and their families respond to integration. The findings in this paper specifically address the phenomenon of white flight, an often sensationalized term but one that has a precise meaning. It is the accelerating loss of white students above and beyond what may have been previously predicted (Clotfelter 2001). Multiple sources have concurred in showing the connection between white flight and mandatory desegregation plans (Coleman 1975; Welch & Light 1987; Clotfelter 2001; Lutz 2011). Clotfelter (2001) shows that for a dataset of 238 metropolitan areas, white flight between 1987 and 1996 is most strongly associated with increasing interracial contact. Lutz (2011) shows that the end of court-ordered desegregation plans induced the opposite of white flight—white students and families re-entering districts. The findings of Clotfelter and Lutz alone seem to indicate that Parents Involved, which ended an integration plan aiming to increase interracial contact in Seattle schools, would likely reduce white flight. I test this hypothesis in the following sections.

Other research has looked beyond white flight to measure other reactions of white families to integration plans. Cascio, Gordon, Lewis, and Reber (2010) also leverage the 1970s era of court-ordered desegregation plans. They focus on the amount of desegregation that occurred in predominantly white districts and how it related to the Title I funding a district risked losing if schools were not substantially integrated. They find that districts that moved beyond token desegregation were more likely to receive greater amounts of Title I funding. This relationship indicated that the average district would have to be paid $1,200 per pupil to move beyond token desegregation. This elasticity to the effects of desegregation or a threshold effect is closely related to the marginal changes in white enrollment in Seattle following Parents Involved that I explore in this paper. It may seem irrational that white families respond at all to school integration when the majority of research indicates segregation disproportionately harms students of color rather than white students (Billings, Deming & Rockoff 2012). This paper further adds to this literature examining the link between integration plans and white reactions. It also provides new estimates of the causal effect of removing integration plans similar to research analyzing the end of busing in Charlotte-Mecklenburg Schools and other districts under court order (Billings et al. 2012; Deming 2011; Lutz 2011). Because of the legal infeasibility of implementing new school integration plans, these are the most recent estimates of the causal effect of changing school integration plans.

  1. Student Assignment in Seattle Before and After Parents Involved

Seattle schools are a unique example within the larger history of segregation and integration in US schools. Seattle diverges from the common narrative around urban US schools because the city itself has historically been much whiter than cities of similar size (Brown & Morrill 2011). Nonetheless, the city’s schools, being majority-minority since 1985, look strikingly different than the rest of the city (Seattle Times Staff 2007a). Today white students are much less likely to attend the public schools than students of color, and white enrollment has been roughly stable around 20,000 students between 1985 and 2005 down from nearly 80,000 in 1970 (Ruiz & Ellis 2011). This illustration of white flight from the Seattle schools is very similar to the national narrative of white flight discussed in Clotfelter (2001). In this setting, Seattle school leaders developed the racial integration plan in question in Parents Involved.

Prior to the court’s ruling in the summer of 2007, Seattle allowed incoming high school students to choose which school they would attend among the district’s 13 high schools. Jefferson County Public Schools in Louisville, KY used a similar policy that was struck down by the court ruling, but I focus on the case in Seattle where there was less resistance to the court ruling in the following years allowing for a clearer evaluation of the end of the policy. In Seattle, the policy entailed that if a school was oversubscribed―more students wanted to attend the school than capacity permitted―the district assigned students using a system of tiebreakers. The second most important tiebreaker (after whether or not the student had a sibling at the school) was student race. If an oversubscribed school was more white than the rest of the district as a whole, nonwhite students would “win” this tiebreaker and gain admission into their first choice school.[1] If the school was more nonwhite than the rest of the district as a whole, white students would win the tiebreaker. This system had been in place since 1998 making Seattle perhaps the only district in the country to pursue an integration plan when it had never faced a court-ordered plan (Kiel, 2010). To a degree, Seattle voluntarily used this policy to racially integrate its schools.

Seattle used this student assignment plan until the Supreme Court struck it down in 2007. In hearing the case, the court recognized that seeking diversity and avoiding racial isolation are compelling state interests, but struck down both school districts’ assignment plans because the plans were not sufficiently “narrowly tailored.” This legal term refers to past Supreme Court cases on race and education, Grutter v. Bollinger (2003) and Gratz v. Bollinger (2003), which set the precedent that race-based classifications must be directed toward a “compelling government interest” and be “narrowly tailored” to that interest. The case was decided by a 4-1-4 decision with Justice Kennedy writing the swing vote opinion that led to the case being decided against the district. Kennedy agreed that the student assignment plan in use was unconstitutional, but also held that avoiding racial isolation and promoting diversity were compelling interests. Ultimately, the student assignment plan was struck down by a split court who likely based their decision around ideology and ambiguous and complicated evidence. The new student assignment plans developed after 2007 did not seriously consider mechanisms to integrate Seattle schools (Rosenthal & Mayo 2012; Higgins 2013).

There are multiple reasons affirming the validity of using a quasi-experimental design to assess the impact of the policy change in Seattle. Quasi-experimental design relies on the assumption that the policy change that determines treatment was both unexpected and sudden. The court decision came during the summer before the 2007-08 school year, about one month before classes began. Furthermore, when questioned about the close timing, attorney for the plaintiff Teddy Gordon stated that all the District had to do was “push a button” to change things over to a plan compliant with the court’s ruling (Parents Involved 2007). And finally, as the 4-1-4 vote indicates, it was not clear to the public that the court would vote for a certain position leaving the original student assignment plan’s fate up to chance. In fact, multiple lower courts had upheld the constitutionality of the district’s plan leading up to the Supreme Court’s decision (Seattle Times staff 2007b). This evidence combined illustrates the sudden and unexpected nature of the court decision, and the consequently valid quasi-experimental environment in which the student assignment plan was changed.

III. Data, Summary Statistics, and Trends

  1. Data Description

This paper uses predominantly two data sets. The first data set is yearly cross-sectional demographic data from the ACS (Ruggles et al., 2015).  I limit the sample to individuals from Washington state. The data spans the years 2000-2014 providing seven years of data on either side of the 2007 ruling. The data contains individual-level demographic, schooling, and migration information. Specifically, the data contains information on all individuals in families including their race, what type of school a child attends (public or private), as well as where a family lives and how recently they moved there.

The second data set is a combination of school-level enrollment data from the Common Core of Data (CCD) and yearly test score data from the Washington Office of the Superintendent of Public Instruction (OSPI). The CCD data is generated from nationwide surveys which track enrollment by race, gender, and free or reduced lunch status at every public and private school. The OSPI data is broken down within the school level by grade, subject (reading or math), and racial subgroup. For example, an observation might be: “Adams Elementary, 2001-02, grade 4, 50 percent meeting reading standard, Hispanic.” The one limitation of this data is that the percentages are not reported if there is less than 10 students in a subgroup in order to protect the privacy of the students. These missing observations account for about 50% of the 12,000 observations in Seattle from 2001-2013. This is not ideal, but it is unlikely that there is differential selection into and out of subgroups with less than 10 students. Consequently, this privacy requirement may lower the precision of estimates but should not bias estimates.

  1. Summary Statistics

Table 1 presents summary statistics for the ACS data.  Columns 1 and 2 present the sample means and standard deviations for the data before and after the 2007 ruling respectively,

and column 3 presents the same statistics for the whole sample. The data is restricted to adults ages 25 and older. While the subsamples before and after 2007 are similar, a limited amount of differences are worth noting. First, the post-2007 sample is roughly fifty percent larger than the pre-2007 sample. This is a result of the ACS becoming a more robust survey overtime. In addition, the ACS did not survey individuals about their current city before 2004 so those observations are dropped from analysis. Beyond the size of the two samples, the post-2007 subsample has a greater percentage of adults living in Seattle (11 versus 4 percent) and a greater percentage of adults who send their child to public schools (78 versus 58 percent). These two differences are exactly what is expected if the 2007 court ruling made the Seattle Public Schools more desirable. For the remaining variables, the two samples before and after 2007 are similar. This relative balance across time periods indicates it is unlikely that other major demographic changes occurred in Seattle across these time periods. This only further validates the experimental design explained in the next section.

Table 2 presents summary statistics for the Seattle enrollment and test score data and is formatted similarly to Table 1. While the relatively large standard deviations imply there is variation across schools in Seattle and over the time period, it appears that the average school had a higher percentage of white students and a lower percentage of nonwhite students after the court ruling. The average percent meeting the math and reading standards improved across this time period as well. In analysis not shown, I disaggregate the data by racial subgroup. In Seattle, the average percentage meeting the math and reading standards is highest for white students and lowest for black students. Both groups show similar growth over time with black students

Table 1. ACS Summary Statistics

(1) (2) (3)
Pre-2007 Post-2007 Full Sample
Lives in Seattle (%) 0.04 0.11 0.07
(0.19) (0.31) (0.26)
White (%) 0.84 0.82 0.83
(0.37) (0.39) (0.38)
Has kids (%) 0.40 0.38 0.39
(0.49) (0.48) (0.49)
Child attends public school (%) 0.58 0.78 0.68
(0.49) (0.42) (0.47)
Age 45.53 47.06 46.29
(17.75) (18.24) (18.01)
Number of own family members in household 2.44 2.40 2.42
(1.43) (1.48) (1.46)
Number of own children in the household 0.72 0.69 0.71
(1.08) (1.08) (1.08)
Observations 207,303 348,093 555,396

Table 1 reports sample means (and standard deviations) for the demographic and dwelling variables in the ACS data.


Table 2. Enrollment and Test Score Summary Statistics

(1) (2) (3)
Pre-2007 Post-2007 Full Sample
Total students 491.80 498.96 496.87
(316.13) (270.32) (284.45)
Percent white 41.16 41.93 41.71
(26.65) (26.81) (26.76)
Percent nonwhite 58.84 58.07 58.29
(26.65) (26.81) (26.76)
Percent free/reduced lunch 43.42 44.27 44.03
(26.18) (27.36) (27.02)
Total tested (math) 23.04 18.61 19.80
(35.21) (28.71) (30.64)
Percent meeting math standard 50.77 62.94 59.58
(26.67) (23.02) (24.68)
Total tested (reading) 23.03 19.43 20.37
(35.29) (30.28) (31.70)
Percent meeting reading standard 64.30 70.66 69.00
(23.20) (18.52) (20.04)
Observations 3805 9051 12856

Table 2 reports sample means (and standard deviations) for the CCD enrollment variables and the WA OSPI test score variables.

improving their average percent meeting the math standard by over 10 percentage points across

time periods.

  1. Seattle Enrollment and Test Score Trends

            After the court ruling struck down the racial-balancing student assignment plan in use in Seattle, it is possible that this policy change affected how families viewed the district. Furthermore, it is possible that the change in student assignment plan had heterogeneous effects on families’ views of the district. In order to assess this possibility, I graph Seattle schools enrollment by race in the years before and after the court ruling.

Figure 1 graphs both white and nonwhite student enrollment for the school years 2000-01 through 2013-14. While there is no clear trend in nonwhite enrollment, it appears that there is a sudden increase in white enrollment following the 2007 ruling after white enrollment was constant in the decade leading up to the ruling. During the seven years after the court ruling, white enrollment in the district increased by about 20 percent, an increase that amounts to an annual 3 percent increase in white enrollment during those years. While it is possible that this increase in white enrollment could be caused by a number of forces, the timing of it seems to be highly connected with the 2007 court ruling.

The increase in the number of white students enrolled in Seattle after 2007 is mirrored by the trends in Figure 2. Figure 2 shows that the average Seattle school was 40 percent white and

60 percent nonwhite before the court ruling. In the years following the court ruling, these percentages converged until the average Seattle school was about 44 percent white and 56 percent nonwhite in 2011. Figure 3 illustrates how this increase led to an increasing number of predominantly white schools. While before the court ruling, there was a roughly equal number of predominantly white schools (greater than 65 percent white) and well integrated schools (35-65 percent white), the number of heavily white schools increases after 2007 along with a large drop in the number of well integrated schools. Figure 4 reveals the change in the distribution of schools with similar student body racial compositions. Between the 2003-04 and 2010-11 school years, the number of schools comprised of less than five percent white students increases dramatically. This is represented by the growth of the leftmost bar in the 2010-11 histogram. Further, we see a rightward shift of the distribution as the number of schools in the middle of the distribution decreases and the right tail of the distribution increases in magnitude. This reveals that not only did Parents Involved lead to more white students in Seattle schools but that those students tended to cluster in predominantly white schools likely exacerbating segregation in Seattle schools.

Figure 5 plots mean percentages meeting math and reading standards throughout Seattle schools over the same time period of increased white enrollment. The mean percentages appear to be on a steady trend of growth up until 2007. After 2007, there is a drop in the district-wide percentage meeting the reading standard, which appears to be driven by nonwhite students. The effect of the court ruling on test scores will be evaluated in the following sections.

  1. Empirical Framework
  2. Modeling Changing Racial Composition of Seattle Families

If the increase in white enrollment in Seattle is caused by white families relocating from the Washington suburbs to Seattle, white parents of school-aged children will be more likely to live in Seattle than nonwhite parents. This hypothesis will be tested using the linear probability model:

Equation 1 uses a triple differences design to model the probability that a given white adult who has kids lives in Seattle before and after the 2007 court ruling. The coefficient of interest is which represents the increase in probability an adult lives in Seattle given they have kids, the court ruling has occurred, and the adult is white. If the coefficient is positive, this confirms the above hypothesis that the uptick in white enrollment in Seattle is driven by white suburban families choosing to relocate to Seattle after the 2007 court ruling.

It will also prove useful to distill the above triple differences models into two difference-in-differences models, one for each racial subsample. This can be done by removing the indicator variable from the above triple differences models.

  • By restricting the equation 3 model to either racial subsample, measures the increase in likelihood within that subsample that a public school parent lives in Seattle following the 2007 court ruling.

  1. Modeling the Effect of School Racial Composition on Test Scores

            There are two driving questions to my analysis of Seattle test scores: (1) How did the court ruling affect student test scores and the racial test score gap after 2007 and (2) what role, if any, did the changing racial composition of schools play in this effect. To address the first question, I estimate the regression,

where represents the percent meeting the standard of the test at school s, in grade g, and of race r. After controlling for school fixed effects (), measures the the effect of the Parents Involved decision on student test scores. I evaluate this regression separately by black and white subgroups and separately for both math and reading test score effects.

To address the second of the two driving questions―what role did the changing racial composition of schools play in affecting test scores―I estimate the regression,

where equation 5 is no different from equation 4 except for the addition of  and the interaction term.  measures the change in the percentage of a school’s students who are white between the 2003-04 and 2010-11 school years. I choose these two years to account for any leading or lagging changes in the racial composition of Seattle schools.  measures the post-Parents Involved effect of increasing the percentage of a school’s students who are white on student test scores. It is possible that this estimate is biased due to the endogeneity of the variable; changes in racial demographics during this time could be correlated with factors that determine student test scores. Nonetheless, a positive and significant estimate indicates that the racial subgroup’s test scores benefited from a growing white population in the school.

  1. Results
  2. Mitigating White Flight

Baseline Models The first column of Table 3 presents the regression results for the baseline triple-differences model represented by equation 1. The coefficient in the first row of column 1, which represents our estimated , is positive and significant. This indicates that after the court ruling, white parents were nearly two percentage points more likely to live in Seattle compared to nonwhite parents. This is a 20 percent increase on top of a mean of 7.76 percent. This finding indicates that a greater share of Seattle parents will be white in the years after the court ruling. This likely explains the increase in white enrollment in the public schools following the court ruling.

The remaining two columns of Table 3 present the regression results for the equation 2 difference-in-differences model for white and nonwhite subgroups respectively. These estimates

Table 3. Mitigating White Flight

(1) (2) (3)
Full White Nonwhite
Parent indicator X Post-PICS indicator X White indicator 0.0157*


Parent indicator X Post-PICS indicator -0.0311*** -0.0154*** -0.0311***
(0.0065) (0.0019) (0.0065)
Parent indicator -0.0191*** -0.0058*** -0.0191***
(0.0035) (0.0009) (0.0035)
Post-PICS indicator 0.1119*** 0.0707*** 0.1119***
(0.0059) (0.0017) (0.0059)
White indicator -0.0404***
Parent indicator X White indicator 0.0132***
Post-PICS indicator X White indicator -0.0412***
Constant 0.0776*** 0.0371*** 0.0776***
(0.0030) (0.0007) (0.0030)
Observations 572664 482309 90355
Adjusted R2 0.0208 0.0148 0.0198

Table 3 presents estimated regression coefficients (and robust standard errors) for the given independent variable in the baseline models of the probability of living in Seattle using the ACS data. The final two columns use white and nonwhite subsamples and the double-difference model respectively whereas the first column uses the triple-differences model.

* p < 0.05, ** p < 0.01, *** p < 0.001


illuminate what is driving the positive coefficient in the first row of the first column of the table. Nonwhite parents within their subsample were more than twice less likely to live in Seattle after 2007 compared to white parents within their subsample. This suggests that both nonwhite and white parents were less likely to live in Seattle after 2007, but the effect was even greater for nonwhite parents. This ultimately would lead to a growing share of white parents in Seattle albeit a declining share of parents overall.

Figure 6 puts the findings from Table 3 into graphical context. The figure graphs the adult population in Seattle broken down by race and parent status. Two trends stand out. First, the largest increases in the adult population in Seattle appear to be driven by the non-parent populations of all races. These populations are less likely to be making decisions related to the public schools than the parent population so these increases are not likely caused by changes in the public schools. Nonetheless, it appears that the slope of the lines for white parents and nonwhite parents differ. While the white parent line slopes upward, albeit not as steeply as the white non-parent line, the nonwhite parent line is mainly flat or decreasing. This difference in slope mirrors the relative difference in the estimated coefficients on the interaction terms in Table 3. Both the regression and graphical results suggest that white families are more likely to view Seattle favorably following the court ruling.

Placebo Tests If the court ruling truly increased the favorability of Seattle schools in the eyes of white families, this effect should not exist prior to the court ruling. Furthermore, it should likely fade overtime after the court ruling. To assess the true causal impact of the court ruling in 2007, I compare the results of the baseline models in Table 3 with various placebo models. These placebo models replace theindicator variable with year indicator variables to track the changes in white parents probability of living in Seattle over time. If the court ruling had a true causal impact on families after 2007, the placebo models should indicate this.

Figure 7 presents this placebo test comparison. Each vertical lines represents an estimated  and its confidence interval for the given year indicator variable. The estimates for the 2005 and 2006 indicator variables are positive and significant casting doubt on the findings of the baseline model. This would indicate that white parents viewed Seattle more favorably even before the court case. Nonetheless, the year indicator variable estimate is not significant from zero in the year the court case was decided and then grows in magnitude for the next four years. This rising trend is what ultimately drives the results in the baseline model and provides significant support that white parents began to view Seattle more favorably after 2007. While the first two placebo estimates cast doubt on the baseline findings, the remaining estimates provide convincing support. In combination with the graphical results in Figure 6, the evidence seems to support the notion that white families viewed Seattle schools more favorably after the court ruling.

Lastly, Figure 8 presents the results of a placebo comparison using ACS data from Oregon and Portland rather than Washington and Seattle. In analysis not shown, I confirm the data from Oregon closely mirror the data from Washington. The major if not only difference between the two locales is the change in integration policy in Seattle. If the Oregon data reveal a similar trend to the one in Washington, that is, the probability of white parents from Oregon living in the city of Portland increases after 2007, then something else may be driving the changing demographics in Seattle other than changes in school policy. Figure 8 shows that this is not the case. There is no clear trend in the estimates after 2007 and few estimates differ significantly from zero. Both Washington and Oregon and Seattle and Portland are similar areas, but the differing results in Figures 7 and 7 indicate that white parents in Seattle were behaving differently from white parents in Portland after 2007. With this evidence, we can likely rule out that the trends in Figure 7 and the baseline results are caused by a nationwide phenomenon like the Great Recession. White parents in Seattle truly were behaving differently after 2007 providing further support for my hypothesis that parents viewed Seattle schools differently after Parents Involved.

  1. The Racial Test Score Gap and School Composition

Baseline Models Table 4 presents baseline estimates of the effect of the court ruling on student test scores. Panel A shows that both white and black test scores increased after the court ruling with black test scores increasing by more than one percentage point greater than white test

Table 4. Seattle Test Scores After Parents Involved

(1) (2)
White Black
Panel A. Math
After PICS 3.6946*** 5.0683***
(0.6928) (1.1791)
Constant 74.7717*** 30.6917***
(1.9848) (6.5214)
Observations 1192 889
Adjusted R2 0.4980 0.2348
Panel B. Reading
After PICS 1.3412* -1.6031
(0.5664) (1.0954)
Constant 85.0080*** 48.6031***
(1.4872) (6.5214)
Observations 1025 934
Adjusted R2 0.4072 0.2037

Table 4 presents estimated regression coefficients (and robust standard errors) for the post-PICS effect on student test scores. School fixed effects are included in all models.

* p < 0.05, ** p < 0.01, *** p < 0.001

scores. This indicates that the test score gap in math decreased in the years following the court decision. Both of these test score effects are highly significant (p<.001) although not significantly different from each other.

Panel B shows that while white reading scores significantly increased during this time, the effect on black reading scores is negative and insignificant. Both of these effects are a fraction of the magnitude of the effect on math scores. While panel B indicates that the reading test score gap may have grew after the court ruling, this growth is not nearly as significant as the narrowing of the math test score gap as shown in panel A.

Placebo Tests The findings in Table 4 face a similar threat to their validity as the baseline findings on the mitigation of white flight in Table 3. It is possible that the mainly positive coefficients reported could be a part of a wider positive trend in the test scores of white and black students. To assess this possibility I run a variety of placebo tests which replace the after PICS indicator variable with school year indicators. Figure 9 presents the results of these placebo tests. If the court ruling has a true effect on the test scores of students, this effect should not be present before 2007. This is exactly the case for the math score placebo tests shown in panels A and B of Figure 9. While the estimates are either not significantly different from zero or close to zero before 2007, these yearly effects show a clear increasing trend after the year of the court decision. This shows that Seattle math scores significantly increased in the years following 2007 for both black and white students.


The case for reading scores is not as clear. While there appears to be a slight increasing trend in white reading scores after 2007, these increases are not significantly different from the previous positive yearly indicator effects. No clear trend is evident for black reading scores. Figure 9 ultimately indicates that student math scores were significantly affected after 2007 while reading scores do not appear to be significantly affected. This implies that the most valid results in Table 4 are those for the math scores.

School Racial Composition and Test Scores Table 5 presents regression results for the interaction model used to evaluate the test score effect of increases in the size of a school’s white population. All estimates for the interaction term in Table 5 are positive indicating that any increases in student test scores after 2007 are in part due to increases in size of a school’s white population. For white math and reading scores, this estimate is highly significant (p<.01) and averages about a quarter of a percentage point. While this effect may seem small, the average growth in a the size of a school’s white population during this time is about two percentage points. This implies that due to increases in the size of a school’s white population alone, white test scores increased by a half percentage point or about 15 percent of the total estimated increase in white math scores or 33 percent of the total estimated increase in white reading scores. The changes in the white population of the average Seattle school led to a substantial and significant increase in white students’ test scores.

The estimates for black students’ math and reading scores are not nearly as significant as the estimates for white students’ scores. Nonetheless, they are still positive estimates indicating it

Table 5. Changing Racial Enrollment and Test Score Effects

(1) (2)
White Black
Panel A. Math
After PICS 2.9880*** 5.1615***
(0.7335) (1.2006)
Change in percent white -0.7335*** -0.8127*
(0.1793) (0.3482)
After PICS X Change in percent white 0.2750**




Constant 85.0420*** 43.0304***
(2.3348) (3.4719)
Observations 1141 860
Adjusted R2 0.5121 0.2370
Panel B. Reading
After PICS 0.6717 -1.2622
(0.5823) (1.1136)
Change in percent white -0.5582*** -0.9907**
(0.1490) (0.3375)
After PICS X Change in percent white 0.2378***




Constant 92.5853*** 63.7731***
(2.1204) (3.4263)
Observations 983 905
Adjusted R2 0.4156 0.2040

Table 5 presents estimated regression coefficients (and robust standard errors) for the post-PICS effect on

student test scores interacted with a measure of the change in percentage white between 2003 and 2010. School fixed effects are included in all models.

* p < 0.05, ** p < 0.01, *** p < 0.001

is more likely than not that increases in the size of a school’s white population led to increases in the test scores of black students. At a minimum, this suggests that increases in the size of a school’s white student body does not harm the black students in the school. Most evidence indicates that the whitening of a school in part benefits the test scores of that school’s black students.

School Subgroup Analysis To further understand the nature of the effects presented in Table 4, I split the sample schools in three subgroups. I determine the three subgroups by the percentage of each school’s student body that is white in the 2003-04 school year. I divide this distribution into equal-sized thirds so that there is a least white group (bottom third), a most white group (top third), and a middling group (middle third). This is akin to dividing the distribution shown in the first panel of Figure 4 into thirds equal in number of schools.

Table 7 shows the baseline test score model estimated for each school subgroup. The math score results in panel A reflect the results of panel A in Table 4. All estimates are positive. The largest significant estimate for white students appears in the estimate for the middle third subgroup indicating that the positive effect of Parents Involved on white students’ test scores was most pronounced for students in this subgroup’s schools. Each estimate for the math scores of black students is significant (p<.05). For these students, the top third subgroup is driving the positive test score effect after 2007 with an estimate of more than 11 percentage points. Considering that the mean number of black students in this subgroup’s schools is 32.04, this increase is equivalent to an additional three students meeting the standard. This may seem like a meager effect, but in a school of 400 students, of whom 30 are black, an additional three black students meeting the standard would be good news for any school leader.


Table 7. School Subgroup Analysis

(1) (2) (3)
  Bottom Third Middle Third Top Third
Panel A. Math
White 5.3417 4.4935*** 2.7676***
(5.0924)  (1.1217)  (0.7819)
Black 4.7540** 4.5404* 11.9296**
(1.6297)  (2.0080)  (3.9473)
Panel B. Reading
White 3.4284 2.0815* 0.2527
(4.5259)  (0.8346)  (0.6329)
Black -5.0018** 2.3491 6.5006+
 (1.4408)  (1.9428)  (3.6674)

Table 7 presents regression coefficients (and robust standard errors) for the post-PICS test score effects

broken down by each of the three school types and by race. Each cell represents a separate regression model. School fixed effects are included in all models. Adjusted R2 ranges from 0.17 to 0.63 with a mean of .34 for the 12 models.

+ p < 0.10, * p < 0.05, ** p < 0.01, *** p < 0.001

The reading score results in panel B are less significant than those in panel A. The most intriguing estimates are those for black students in the bottom and top third subgroups. Black students in the bottom third subgroup see a five percentage point decrease in their reading scores after 2007 (p<.01). Quite the opposite, black students in the top third subgroup see a more than six percentage point increase in their reading scores after 2007 (p<.1). While the initial estimates for the post-PICS effect on black students’ reading scores appeared to be close to zero, the subgroup analysis shows that the effect’s magnitude and sign varied quite largely across the three school subgroups. While the net effect on black students’ reading scores appears to be zero, there remains roughly five times more black students in the bottom third subgroup than the top third subgroup indicating that the majority of black students in Seattle saw decreases in their reading scores after Parents Involved.

  1. Discussion
  2. Interpretation

The evidence in this paper is clear: the end of the racial integration plan in Seattle after Parents Involved led to an increase in the number of white families living in Seattle, increased the average school’s proportion of white students, and increased student test scores on average. The most plausible explanation for the change in white families’ preferences for the district before and after 2007 is that they believed the court ruling would benefit their children’s outcomes in Seattle schools. For example, before the case, a white family could move to a certain neighborhood in Seattle, but because of the racial balancing tiebreaker, they still might not be able to have their child enrolled in their first choice school. This is, after all, the exact reason why the plaintiff, Parents Involved in Community Schools, brought the case against Seattle Public Schools (Shaw, 2007). In effect, this integrationist policy likely had the unintended consequence of discouraging white families from living in Seattle who would have otherwise. My results indicate that, after the racial tiebreaker was struck down, white families’ preferences for the district changed pulling new families into the district’s schools. Although this is a difficult explanation to test, it is the most likely explanation for why white families were making up an increasing share of the families served by Seattle schools after the court ruling. Ethnographic research of families who have recently moved to Seattle might provide further clarification on this point. This finding also parallels those of Clotfelter (2001) and Lutz (2011) who show white families are averse to districts using integration plans to increase interracial contact.

As the district’s schools saw the share of its students who are white increase, this change in racial composition had complex effects on student test scores. The effect of Parents Involved on student math scores are clearer than the effect on reading scores. The results shown indicate that both black and white math scores increased after 2007 with the increases of black students outpacing those of white students. This effect is at least in part due to the increase in the number of white students at schools. Figure 10 shows this decrease in the size of the math test score gap after 2007. The more than five percentage point drop led to a 10 percent reduction of the racial test score gap in math. The flat trend in the reading test score gap reflects the less substantial effects of Parents Involved on reading scores. If there is any effect at all, it is a slight increase in the reading test score gap, but the results in the subgroup analysis emphasize that black students in the least white schools and the most white schools fared dramatically different after Parents Involved. This suggests that any benefits to schools after the court ruling were not equally distributed across schools.

  1. Limitations

A small number of limitations threaten the implications of these findings. The burgeoning tech industry in Seattle and the Great Recession are possible confounders of the findings on migration of white families. These two possible confounders actually had an interesting interplay during this time period. While it is likely that the tech industry is disproportionately white and a growing tech industry would lead to additional white families living in Seattle, the growth of the tech industry in Seattle stalled after 2007 and into the Great Recession (Brown & Morrill 2011). In addition, home prices dropped dramatically during the Great Recession after 2008 (Brown & Morrill 2011). These two trends in combination would seem to stall any growth of the white population in Seattle. Consequently, the estimates of the reversal of white flight likely are not biased positively. In combination with the Portland, Oregon placebo test results, this indicates that it is unlikely that either the tech industry or the Great Recession are driving the reversal of white flight during the time period after Parents Involved.

The test score estimates also have some limitations. There is no guarantee that the tests developed by the state remained the exact same each year. Furthermore, there is a long line of work illuminating test score manipulation and cultural bias (Dee et al. 2016; Jencks 1998) which could further complicate my results. I have found no clear evidence of wide scale test changes or manipulation during the years I analyze. The graphs of test scores previously shown should also assuage any concerns. The trends of the scores appear to be stable, and there are no obvious signs of the test changing from year to year in the graphs. All evidence indicates that the test delivered was remarkably unchanging during the years I analyze.

The nature of the causal effect on test scores in Table 4 is worth emphasizing. Because the data is disaggregated by racial subgroups, the estimates in Table 4 do not represent the causal effect of changing school racial composition on individual students but the effect on the percentage meeting the test standard. Individual-level test score data would provide the most convincing estimates of the effect of Parents Involved on student test scores, but this data is not publicly available. The subgroup level data could be affected by student attrition or addition. For example, if white, affluent students from the suburbs of Washington have higher test scores than students in the city, an influx of white students from the suburbs would mechanically lead to test score increases for white students. That being said, the percent meeting standard measure of test scores is likely a strong proxy for the actual test scores of individual students. The test score findings carry substantial importance, but they are not to be interpreted as the direct effect of the court ruling or school racial composition on individual student test scores.

VII. Conclusion

The US Supreme Court ruled on Parents Involved with the understanding that the student assignment policy in Seattle was indirect and ineffective. Previous to this paper, no rigorous research tested this belief. This paper reveals that ending the student assignment plan directly impacted the demographics of the city and the test scores of students. After the court ruling, white student enrollment increased 20 percent as a result of families relocating to Seattle. I argue that this large-scale relocation was, at least in part, due to the court ruling, which incentivized parents to enroll their children in Seattle schools in hopes that they would gain access to their first choice school. By creating this incentive structure, the court appears to have mitigated the trend of white flight. I subsequently argue that the influx of white families into the district led to test score increases for both black and white students. My analysis reveals that the math test score gap between black and white students was narrowed by as much as 10 percent. There is one caveat to this analysis however: While white and black test scores increased on average after the court ruling, black students in the schools that saw the smallest influxes of white students saw decreases in their test scores after the court ruling. It is possible then that these average test score benefits are not reaching the most disadvantaged students. This paper’s findings parallel those of Clotfelter (2001) and Lutz (2011), which indicate interracial contact induces white flight and the end of integration plans mitigates white flight. As I have shown, this reentry of white families and students in the district leads to test score benefits for white and black students in most schools and a narrowing racial test score gap.

While the end of racial integration in Seattle mitigated the effects of white flight and had test score benefits for students, this should not be read as an endorsement for eliminating school integration plans. In the case of Seattle, the end of the school integration plan led to a narrowing of the test score gap likely because of an influx of resources and capital from white families. This demonstrates that no matter the intention of school integration plans, they may have the unintended consequence of pushing valuable resources and capital out of urban school districts. The results in this paper are unique to Seattle and Washington, but the tradeoff between integration and white flight is a tradeoff all policymakers should consider. Whether the integration plan is race-based or socioeconomic-based, this tradeoff will likely remain.

In the fallout of Parents Involved, Ryan argues the court ruling eliminated “the hope that black and white students across the country would routinely enter through the same schoolhouse doors to a world of truly equal educational opportunity” (2007, pp. 155).  This paper actually shows the integration plan in Seattle itself prevented white students from attending the same schools as black students and that the end of the plan did not increase educational inequity but reduced it. Though this paper does not explore the net effect of Parents Involved, it does shed light on an important principle that should influence education policy makers across the country: integration is not a panacea for increasing educational equity. As this paper reveals, to be successful, integration plans cannot drive social and monetary capital out of urban school districts. Integration plans that can attract white and affluent families into urban schools are those that will likely have the most positive effects on educational equity.



Angrist, J. D., & Lang, K. (2004). Does school integration generate peer effects? Evidence from Boston’s Metco Program. The American Economic Review, 94(5), 1613-1634.

Barnum, M. (2016, December 11). Is school segregation getting worse? By some measures, no – but it’s not getting any better. The 74.

Bergman, P. (Forthcoming). The effects of school integration: Evidence from a randomized desegregation program.

Billings, S. B., Deming, D. J., & Rockoff, J. E. (2012). School segregation, educational attainment and crime: Evidence from the end of busing in Charlotte-Mecklenburg (No. w18487). National Bureau of Economic Research.

Brown v. Board of Education of Topeka (1), 347 U.S. 483 (1954)

Brown, M., & Morrill, R. L. (Eds.). (2011). Seattle geographies. University of Washington Press.

Cascio, E., Gordon, N., Lewis, E., & Reber, S. (2010). Paying for progress: Conditional grants and the desegregation of the south. Quarterly Journal of Economics, 125(1), 445-482.

Clotfelter, C. T. (2001). Are Whites still fleeing? Racial patterns and enrollment shifts in urban public schools, 1987–1996. Journal of Policy Analysis and Management, 20(2), 199-221.

Coleman, J. S. (1975). Trends in School Segregation, 1968-73.

Coleman, J. S. (1988). Social capital in the creation of human capital. American journal of sociology, S95-S120.

Dee, T. S., Dobbie, W., Jacob, B. A., & Rockoff, J. (2016). The Causes and Consequences of Test Score Manipulation: Evidence from the New York Regents Examinations (No. w22165). National Bureau of Economic Research.

Deming, D. J. (2011). Better Schools, Less Crime?. The Quarterly journal of economics, 126(4), 2063-2115.

Graham, P. A. (2007). Schooling America: How the public schools meet the nation’s changing needs. Oxford University Press.

Gratz v. Bollinger, 539 U.S. 244 (2003)

Grutter v. Bollinger, 539 U.S. 306 (2003)

Guryan, J. (2004). Desegregation and black dropout rates. The American Economic Review, 94(4), 919-943.

Higgins, J. (2013, November 21). Seattle School Board oks boundary plan. The Seattle Times.

Jencks, C. (1998). Racial bias in testing. The Black-White test score gap, 55, 84.

Johnson, R. C. (2015). Long-run impacts of school desegregation & school quality on adult attainments (No. w16664). National Bureau of Economic Research.

Kiel, D. (2010). Accepting Justice Kennedy’s Dare: The Future of Integration in a Post-PICS World. Fordham Law Review, 76.

Lutz, B. (2011). The end of court-ordered desegregation. American Economic Journal: Economic Policy, 3(2), 130-168.

Parents Involved in Community Schools v. Seattle School District No. 1, 551 U.S. 701 (2007)

Reardon, S. F. (2016). School segregation and racial academic achievement gaps. RSF.

Reber, S. J. (2010). School desegregation and educational attainment for blacks. Journal of Human resources, 45(4), 893-914.

Rosenthal, B., & Mayo, J. (2012, August 8). 6 Seattle schools have become whiter as new assignment plan changes racial balance. The Seattle Times.

Ruiz, T., & Ellis, M. (2011). Turning back the clock: The resegregation of Seattle Public Schools. In M. Brown & R. Morrill (Ed.), Seattle Geographies (pp. 136-143). University of Washington Press.

Ruggles, S., Genadek, K., Goeken, R., Grover, J., Sobek, M. Integrated Public Use Microdata Series: Version 6.0 [Machine-readable database]. Minneapolis: University of Minnesota, 2015.

Ryan, J. E. (2007). Supreme Court and Voluntary Integration, The. Harv. L. Rev., 121, 131.

Seattle Times Staff (2007a, January 2). A tale of two Seattles. The Seattle Times.

Seattle Times Staff (2007b, June 29). A timeline of the racial-tiebreaker case. The Seattle Times.

Shaw, L. (2007, June 29). U.S. Supreme Court rejects Seattle’s racial criteria. The Seattle Times.

Shaw, L. (2008, June 1). The resegregation of Seattle’s schools. The Seattle Times.

Welch, F., & Light, A. (1987). New Evidence on School Desegregation.

Wilson, W. J. (2012). The truly disadvantaged: The inner city, the underclass, and public policy. University of Chicago Press.

[1] SPS uses exactly these two race definitions, white and nonwhite, leaving substantial heterogeneity in race across the nonwhite category, which includes African-American, Latino, Asian-American, and Native American students. In 2007, 41% of SPS students were white and 59% were nonwhite.




This site uses Akismet to reduce spam. Learn how your comment data is processed.