Sarah Barfels
Kirkwood Community College
School racial composition has modest effects on test score gaps, but
evidence of a longer-term impact is scarce. Perpetuation theory suggests
that blacks who attend schools with higher proportions of white classmates may have better job outcomes. Multilevel analyses of two national longitudinal surveys reveal no effects of high school racial composition on occupational status, employment, or annual earnings for
blacks or whites. For other minority groups, attending schools with
more whites impedes occupational advancement. For all groups, however, school racial composition predicts workplace racial composition:
Whites who attend high schools with higher proportions of white students have higher proportions of white coworkers, while nonwhites
who attend schools with higher proportions of whites have fewer samerace coworkers. The ndings are modest in size but robust to alternative specications, and sensitivity analyses support a causal interpretation for same-race coworkers. These results support perpetuation
theory for workplace composition but not for stratication outcomes.
Particularly in the case of African-Americans, the history of race relations
in the United States is largely a history of exclusion. Despite the major mile1
This research was supported by a grant from the William T. Grant Foundation while
the rst author was on the faculty of the University of WisconsinMadison. The au-
1116
1117
Drawing on Allports 1954 contact theory and Pettigrews 1965 analysis of barriers to desegregation, Braddock 1980 developed perpetuation
theory to explain how young persons experiences of racial isolation may
1118
Data for this study come from two national longitudinal studies conducted
by the U.S. Department of Education. High School and Beyond HSB
began in 1980 with a sample of more than 20,000 high school sophomores in
more than 1,000 schools. Students were followed up in 1982, 1984, 1986, and
1992. This study draws on the base year and the 1992 follow-up, focusing on
the association between high school characteristics and labor market outcomes 10 years after the scheduled high school graduation.2 In 1984, the
sample was reduced to a random subsample of about 14,000 respondents,
who constitute the sample for this study.
The National Education Longitudinal Survey NELS began in 1988 with
a sample of nearly 25,000 eighth graders in more than 1,000 schools. A representative subsample of about 15,000 students was selected for follow-up
to more than 1,000 high schools in 1990, and subsequent follow-up surveys
2
Students who changed schools or who dropped out of school were retained in the
analytic sample because dropout and mobility may be endogenous to school racial
composition. In both cases, students school characteristics were taken from the schools
they attended in tenth grade.
1126
1127
1128
Student level:
Proportion of coworkers same
Student test scores . . . . . . . .
Missing test scores . . . . . . . .
Student SES . . . . . . . . . . . .
Missing SES . . . . . . . . . . . .
Male . . . . . . . . . . . . . . . . . .
Black . . . . . . . . . . . . . . . . .
Other . . . . . . . . . . . . . . . . .
Native American . . . . . . . . .
Hispanic . . . . . . . . . . . . . . .
Asian . . . . . . . . . . . . . . . . .
Race relations . . . . . . . . . . .
Missing race relations . . . . . .
Number of students . . . . . . .
School level:a
School proportion white . . . .
School SES . . . . . . . . . . . . .
Private . . . . . . . . . . . . . . . .
North . . . . . . . . . . . . . . . . .
Midwest . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
race
....
....
....
....
....
....
....
....
....
....
....
....
....
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.72
2.08
.11
.22
.28
9,300
.72
50.41
.07
2.04
.03
.48
.10
.14
Mean
HSB
.31
.46
.31
.41
.45
.30
8.62
.25
.71
.17
.50
.30
.35
SD
TABLE 1
Descriptive Statistics
NELS 10
.69
.01
.17
.21
.25
10,120
.65
51.39
.06
.02
.05
.49
.11
.14
Mean
.32
.57
.38
.41
.43
.31
9.59
.23
.75
.21
.50
.31
.35
SD
NELS 8
.32
.58
.38
.41
.43
.50
.32
.34
.11
.30
.19
.67
.22
.49
.11
.14
.01
.10
.04
3.09
.05
10,060
.69
.00
.17
.21
.25
.31
9.89
.19
.74
SD
.64
51.53
.04
2.05
Mean
1129
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.......
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
35.27
N 5 9,300
9.95
N 5 8,620
.90
N 5 9,020
870
.30
.20
.24
.28
.48
Mean
.30
.85
12.43
.46
.40
.45
.45
.50
SD
39.53
N 5 10,080
9.96
N 5 9,200
.96
N 5 9,360
1,070
.34
.19
.36
.26
.38
Mean
NELS 10
.20
.78
13.72
.48
.39
.48
.44
.49
SD
39.13
N 5 10,020
9.96
N 5 9,140
.96
N 5 9,300
.34
.19
.36
.27
.38
3.54
2.48
.07
.08
.04
.17
.08
.05
.35
1,070
Mean
NELS 8
.20
.79
13.70
.47
.40
.48
.44
.48
.59
1.28
.25
.27
.16
.38
.27
.17
.48
SD
NOTE.Student-level statistics are weighted as follows: for HSB, FU4WT; for NELS tenth-grade sample, F4QWT; and for
NELS eighth-grade sample, F4BYPNWT. In accordance with NCES security procedures for restricted data, unweighted
sample sizes have have been rounded to the nearest 10.
a
For cases with data on proportion of coworkers same race.
Log of income . . . . . . . . . . . . . . . . . . . . .
South . . . . . . . . . . . . . . . . . . . .
West . . . . . . . . . . . . . . . . . . . . .
Urban . . . . . . . . . . . . . . . . . . . .
Rural . . . . . . . . . . . . . . . . . . . .
Suburban . . . . . . . . . . . . . . . . .
Racial conict . . . . . . . . . . . . . .
Proporton white racial conict
Missing for school racial conict .
Busing . . . . . . . . . . . . . . . . . . . .
Proportion white busing . . . . .
Missing for busing . . . . . . . . . . .
Court order . . . . . . . . . . . . . . . .
Proportion white court order . .
Missing for court order . . . . . . . .
Number of schools . . . . . . . . . . .
Other dependent variables:
Occupational status . . . . . . . . . .
HSB
TABLE 1 (Continued )
1130
1131
7
In accordance with NCES procedures for restricted data, unweighted sample sizes are
rounded to the nearest 10.
1133
Studies such as this one face two primary methodological problems: nonindependence of students within schools and nonrandom selection of students to segregated and integrated schools. We address both of these challenges in our analyses.
Nonindependence and Multilevel Models
Ordinary least squares regression assumes that each case is independent of
all others, but when samples are drawn within schools, cases in the same
school are not really independent. This can lead to underestimated standard errors, as students in the same school may appear more alike than
would a totally random collection of students. The problem of nonindependence can be addressed with a multilevel model that distinguishes between effects that occur at the level of the individual student and effects
that occur at the level of the group the school, in our case. In the studentlevel model, a student posthigh school outcome is estimated as a function
of student-level background characteristics, including race. The intercept
from this model is an estimate of the average outcome for students in the
school, which in turn becomes an outcome at the school level. This schoollevel outcome is estimated as a function of school-level variables including
school composition e.g., % white, school policy e.g., the proportion of
students who are bused to achieve racial balance, and school context e.g.,
salience of racial conict.
The individual estimate for an outcome among blacks is another level 1
coefcient that becomes an outcome variable at level 2. Here, we ask whether
schools differ in the extent to which blacks and whites reach similar post
high school outcomes. Differences among schools can be modeled as a function of school characteristics, including composition and policy indicators.
Overall, three elements of the multilevel model are key to our analyses: the
difference in outcomes between blacks and whites within schools, the contribution of school composition and school integration policies to average
school outcomes, and the contribution of school composition and school
policies to the black-white difference in outcomes. Note that composition
and policy effects draw on data for all schools, while results for black-white
gaps within schools actually draw only on schools with at least one black
and at least one white student, even though all the data are used in the
1134
level 1:
yij 5 b0j 1 b1j Black 1 b2j Other 1 b3j SES 1 b4j Male
1 b5j Prior Achievement 1 ij ;
level 2:
b0j 5 g00 1 g01 Proportion White 1 g02 Sector 1 g03 Urbanicity
1 g04 Region 1 g05 Sch SES 1 u0j ;
b1j 5 g10 1 g11 Proportion White 1 u1j ;
b2j 5 g20 1 g21 Proportion White 1 u2j ;
where y represents the linear dependent variables and Urbanicity and Region each
stand for a set of dummy variables. Missing data indicators as described in the text are
also included. For the categorical dependent variable, a multilevel logistic specication
is used. The key parameters are g01, g11, and g21, which represent the racial composition
effect for whites g01 and the difference between the effect for whites and that for blacks
g11 and others g21.
1135
First, we report results from our comparison of the HSB and NELS cohorts.
These analyses begin with samples that represent the national populations
of tenth graders in 1980 and 1990 and followed up in 1992 and 2000, respectively. Next, we consider the policy context and, with HSB, examine the
role of internal resegregation. Then, we revisit NELS by focusing on the base
year 1988 sample of eighth graders captured before high school entry, again
relying on the 2000 follow-up for outcome data. Finally, we report the results
of the propensity score analysis and bounds procedure.
Cross-Cohort Comparisons
Table 2 displays the labor market consequences of high school proportion
white, with comparisons of the HSB class of 1982 and NELS tenth-grade
class of 1992 cohorts. Note rst that the ndings are remarkably stable
across the two cohorts. Few of the coefcients are signicantly different
from one another across cohorts. This not only suggests that social processes
9
The standardized mean difference the difference between the means divided by the
square root of half the sum of the squared SDs is commonly used to assess bias reduction
in matching models Harding 2003; Lee 2013; Brand and Thomas 2014.
10
Stuart 2010 suggests that variance ratios should fall between 0.50 and 2.0 and
standardized mean differences should be less than 0.25.
1138
1140
Rural . . . . . . . . . . . . . . . . . . .
School-level controls:
Private . . . . . . . . . . . . . . . . . .
Other . . . . . . . . . . . . . . . . . .
Black . . . . . . . . . . . . . . . . . .
Slopes:
School proportion white:
Reference group white . . . .
Other . . . . . . . . . . . . . . . . . . .
Black . . . . . . . . . . . . . . . . . . .
Intercepts:
Reference group white . . . . . .
2.33**a
.59
.06
.47
2.63
1.22
.41
1.84
23.89**
1.48
35.07**a
.25
1.77**
.62
1.00*
.42
HSB
.12a
.70
2.75
.52
2.18
.96
.04
2.09
24.24*
1.80
39.61**a
.25
.29
.80
.17
.63
NELS10
OCCUPATIONAL STATUS
Linear
.04
.04
2.04
.04
.09
.11
2.26a
.17
2.11
.14
9.92**a
.02
2.05
.06
2.03
.04
HSB
.04
.05
2.03
.04
.00
.06
.10a
.13
.11
.13
9.99**a
.02
2.07
.04
2.05
.04
NELS10
LOG OF INCOME
Linear
.00a
.16
.12
.15
.22
.38
2.29
.55
2.25
.45
2.20**a
.08
.13
.19
.03
.13
HSB
2.45**a
.16
.05
.13
.33
.23
.35
.29
.04
.37
3.05**a
.06
2.30**
.08
2.11
.12
NELS10
EMPLOYED > 20
WEEKS Binary
TABLE 2
Labor Market Consequences of School Proportion White in HSB and NELS 10
2.02
.01
.05**
.01
.19**
.03
2.31**
.05
.02a
.04
.75**a
.01
2.39**
.02
2.21**a
.01
HSB
.01
.02
.06**
.01
.18**
.03
2.39**
.05
2.29**a
.05
.69**a
.01
2.35**
.02
2.32**a
.02
NELS10
PROPORTION
COWORKERS SAME
RACE Linear
1141
.38
.44
2.87
.45
2.99*
.45
21.97**
.51
1.65**
.48
23.69**a
.33
2.59**
.30
.34**a
.02
1,070
10,080
2.32
.51
2.38
.57
2.54
.54
21.53*
.61
.89
.53
.43**
.02
.14**
.02
.01**
.00
870
8,620
.07
.04
2.10**
.03
2.09*
.04
2.12**
.04
.02
.04
.37**
.02
.06**
.02
.005**
.00
1,070
9,200
.05
.04
2.03
.03
2.06
.03
2.13**
.04
2.03
.04
1.27**a
.09
.20**a
.07
.01*
.01
870
9,020
.07
.13
2.04
.14
2.02
.14
2.27a
.16
.11
.15
LOG OF INCOME
Linear
.67**a
.08
2.14**a
.05
.01*
.00
1,070
9,360
.06
.13
.18
.14
.02
.12
.29*a
.13
2.01
.10
.00
.01
.00a
.00
.002**a
.00
870
9,300
.00
.01
.02
.01
2.02
.01
2.06**
.01
.00
.01
EMPLOYED > 20
WEEKS Binary
2.01
.01
2.01a
.01
.00a
.00
1,070
10,120
.01
.01
.04**
.01
2.01
.01
2.04**
.02
.01
.01
NOTE.In accordance with NCES security procedures for restricted data, unweighted sample sizes have been rounded to the nearest 10. Nos. in
parentheses are SEs.
a
The difference between NELS and HSB is signicant at P < .05.
* P < .05.
** P < .01.
N schools . . . . . . . . . . . . . . . . . .
N students . . . . . . . . . . . . . . . . .
Student test . . . . . . . . . . . . . . .
Student SES . . . . . . . . . . . . . .
Student-level controls:
Male . . . . . . . . . . . . . . . . . . . .
School SES . . . . . . . . . . . . . . .
West . . . . . . . . . . . . . . . . . . . .
South . . . . . . . . . . . . . . . . . . .
Midwest . . . . . . . . . . . . . . . . .
Suburban . . . . . . . . . . . . . . . .
OCCUPATIONAL STATUS
Linear
SAME
RACE
Linear
KERS
COWOR-
TION
PROPOR-
1142
1143
1144
1145
1146
Other . . . . . . . . . . . . . . . . . . . .
Black . . . . . . . . . . . . . . . . . . . .
Bused:
Reference group white . . . . . .
Other . . . . . . . . . . . . . . . . . . . .
Black . . . . . . . . . . . . . . . . . . . .
Slopes:
School proportion white:
Reference group white . . . . . .
Race relations . . . . . . . . . . . . . . .
Other . . . . . . . . . . . . . . . . . . . . .
Black . . . . . . . . . . . . . . . . . . . . .
Intercepts:
Reference group white . . . . . . . .
2.51
1.05
2.02
2.12
25.62*
2.25
39.34**
.30
.19
.94
.01
.80
.35
2.91
.44
5.16
21.14
4.00
2.50
1.09
1.34
2.15
25.41*
2.29
39.34**
.29
.62
.87
2.03
.80
Model
2
.20
2.90
1.54
4.35
21.00
4.01
24.24
6.73
7.64
13.19
213.30
13.27
.09
.07
.01
.13
2.06
.14
2.05
.18
.10
.30
.58
.31
.09
.08
.02
.14
2.03
.16
Model
1
2.04
.18
2.04
.30
.54
.31
2.11
.40
1.28
.97
21.36
.73
.61
.31
2.34
.36
2.32
.40
2.09
.56
2.48
.75
3.05**
.68
.63*
.32
2.43
.37
2.29
.42
3.13**
.07
2.51**
.09
2.18
.13
Model
2
Model
1
Model
2
Model
3
PROPORTION COWORKERS
SAME RACE Linear
2.12
.55
2.56
.75
3.18**
.66
2.64
1.69
2.69
2.08
21.08
2.22
.08
.05
2.12
.12
2.04
.12
.08
.05
2.18
.12
2.04
.11
.19**
.19**
.33*
.03
.03
.14
2.45** 2.46**
.21
.06
.06
.37
2.24** 2.23** 2.85**
.05
.05
.27
3.13**
.69**
.69**
.69*
.07
.01
.01
.01
2.41** 2.35** 2.34** 2.35**
.02
.09
.02
.02
2.14
2.32** 2.32** 2.32**
.12
.02
.01
.01
2.06
2.01
.06
.01
Model
3
9.97** 3.13**
.02
.07
2.04
2.57**
.05
.09
2.03
2.29
.04
.13
2.04
.02
Model
3
INCOME Linear
Model
3
Model
1
Model
2
Model
1
OF
LOG
TABLE 3
Busing and Racial Conict as Potential Moderating Conditions for Effects of School Proportion White
NELS Eighth-Grade Sample
1147
.06
.11
2.36
.27
.37
.21
1,060
9,140
1.04
1.80
22.01
3.77
2.21
3.51
1,070
10,020
.09
.27
2.26
.43
2.75
.51
2.06
.09
.37*
.18
2.15
.14
.12
.26
2.52
.41
2.86
.52
21.49
1.47
1.72
2.53
2.22
2.05
21.10
4.03
214.69
8.39
3.63
7.23
.34
.47
.12
.58
.24
.61
1,070
9,300
2.34
.37
.33
.44
.41
.46
.92
.91
.88
.85
2.57
2.28
1.12
1.11
22.85** 22.82**
.96
.94
2.07
.07
.14
.20
.03
.21
2.04
.04
2.19
.11
.17*
.08
1,070
10,060
.03
.03
.13
.07
2.03
.05
2.08
.07
.21
.20
.09
.20
NOTE.Models include school- and student-level controls as in table 2. In accordance with NCES security procedures for restricted data, unweighted sample
sizes have been rounded to the nearest 10. Nos. in parentheses are SEs.
* P < .05.
** P < .01.
N schools . . . . . . . . . . . . . . . . . . . .
N students . . . . . . . . . . . . . . . . . . .
Other . . . . . . . . . . . . . . . . . . . .
Black . . . . . . . . . . . . . . . . . . . .
Other . . . . . . . . . . . . . . . . . . . .
Black . . . . . . . . . . . . . . . . . . . .
Racial conict:
Reference group white . . . . . .
Other . . . . . . . . . . . . . . . . . . . .
Black . . . . . . . . . . . . . . . . . . . .
1148
Native American . . . . . . . . . . . . . . .
Latino . . . . . . . . . . . . . . . . . . . . . . .
Asian . . . . . . . . . . . . . . . . . . . . . . . .
Black . . . . . . . . . . . . . . . . . . . . . . . .
Slopes:
School proportion white:
Reference group white . . . . . . . . . .
Racial relations . . . . . . . . . . . . . . . . . .
Native American . . . . . . . . . . . . . . . . .
Latino . . . . . . . . . . . . . . . . . . . . . . . . .
Asian . . . . . . . . . . . . . . . . . . . . . . . . .
Black . . . . . . . . . . . . . . . . . . . . . . . . .
Intercepts:
Reference group white . . . . . . . . . . . .
2.37
1.14
2.02
2.19
24.48
2.90
26.90*
2.94
.47
3.63
39.29**
.31
.29
.95
.82
.88
2.74
1.16
1.38
1.61
Model 1
2.24
1.20
1.38
2.23
23.70
3.00
26.86*
2.97
2.54
3.67
39.29**
.31
.67
.87
1.18
.86
2.73
1.13
2.57
1.89
Model 2
24.56
7.03
9.56
13.40
2.87
13.96
222.61
15.89
12.65
24.13
39.33**
.30
.59
.97
1.09
.86
2.37
.91
3.14*
1.58
2.27
.27
Model 3
OCCUPATIONAL STATUS
.00
.07
.09
.13
.56*
.28
2.18
.15
.58
.32
9.99**
.02
2.10
.05
2.08
.06
2.08
.06
2.25**
.09
Model 1
OF
INCOME
.01
.08
.09
.14
.64*
.31
2.17
.15
.56
.32
9.99**
.02
2.09
.05
2.06
.06
2.08
.06
2.07
.08
Model 2
LOG
2.01
.41
.98
.96
2.29
1.20
21.99*
.80
21.98
1.96
9.99**
.02
2.07
.05
2.06
.06
2.06
.05
2.01
.08
2.04
.02
Model 3
.18**
.03
2.45**
.06
2.11
.08
2.21**
.06
2.43**
.11
.69**
.01
2.35**
.02
2.41**
.02
2.27**
.02
2.31**
.05
Model 1
.19**
.03
2.46**
.07
2.12
.09
2.20**
.06
2.48**
.11
.69**
.01
2.34**
.02
2.41**
.02
2.27**
.02
2.28**
.05
Model 2
.24
.13
.31
.38
2.31
.43
2.88**
.30
1.49
.87
.69**
.01
2.35**
.02
2.40**
.02
2.27**
.02
2.28**
.05
2.01
.01
Model 3
PROPORTION OF COWORKERS
SAME RACE
TABLE 4
Labor Market Consequences of School Proportion White, Disagreggated by Ethnic Group
NELS Eighth-Grade Sample
1149
Native American . . . . . . . . . . . . . . .
Latino . . . . . . . . . . . . . . . . . . . . . . .
Asian . . . . . . . . . . . . . . . . . . . . . . . .
Black . . . . . . . . . . . . . . . . . . . . . . . .
Racial conict:
Reference group white . . . . . . . . . .
Native American . . . . . . . . . . . . . . .
Latino . . . . . . . . . . . . . . . . . . . . . . .
Asian . . . . . . . . . . . . . . . . . . . . . . . .
Black . . . . . . . . . . . . . . . . . . . . . . . .
Native American . . . . . . . . . . . . . . .
Latino . . . . . . . . . . . . . . . . . . . . . . .
Asian . . . . . . . . . . . . . . . . . . . . . . . .
Black . . . . . . . . . . . . . . . . . . . . . . . .
Bused:
Reference group white . . . . . . . . . .
21.77
4.21
214.89
8.32
217.30
13.84
11.23
7.55
50.02*
21.40
21.45
4.21
213.24
9.32
219.07
14.52
9.88
7.44
61.51*
30.23
21.59
1.55
1.79
2.53
.60
2.59
2.49
2.19
25.93
4.85
.63
3.01
1.83
4.17
12.20
8.73
24.27
4.09
217.54**
5.80
.60
3.02
.97
4.81
13.83
9.16
24.12
4.05
218.83**
7.25
OCCUPATIONAL STATUS
.15
.26
2.41
.43
21.07
.67
2.87
.68
9.77**
.81
2.08
.17
.08
.32
.88
.54
.51
.33
23.02**
.38
OF
INCOME
2.02
.09
.27
.18
2.15
.28
2.21
.12
2.46
.35
.13
.26
2.23
.45
21.02
.66
2.60
.66
11.87**
1.39
2.07
.17
2.03
.33
.86
.53
.39
.33
23.83**
.50
LOG
2.09
.07
.17
.19
.20
.39
.03
.21
1.83**
.57
.09
.05
2.14
.11
2.11
.21
2.07
.12
2.74**
.14
.01
.03
.15*
.07
.04
.09
2.02
.06
.36**
.14
2.10
.07
.21
.19
.15
.37
.11
.21
2.06**
.59
.09
.05
2.19
.11
2.03
.20
2.09
.12
2.77**
.11
PROPORTION OF
COWORKERS
SAME RACE
Model 3
Model 1
OF
INCOME
Model 2
LOG
.01
.11
2.26
.27
.26
.37
.51*
.21
.68
.55
1,060
9,140
Model 3
Model 1
Model 2
2.01
.04
2.22*
.11
.05
.12
.19*
.09
2.53*
.23
1,070
10,060
Model 3
PROPORTION OF COWORKERS
SAME RACE
NOTE.Models include school- and student-level control variables as in table 2. In accordance with NCES security procedures for restricted data,
unweighted sample sizes have been rounded to the nearest 10. Nos. in parentheses are SEs.
* P < .05.
** P < .01.
N schools . . . . . . . . . . . . . . . . . . . . . . . .
N students . . . . . . . . . . . . . . . . . . . . . . .
Native American . . . . . . . . . . . . . . .
Latino . . . . . . . . . . . . . . . . . . . . . . .
Asian . . . . . . . . . . . . . . . . . . . . . . . .
Black . . . . . . . . . . . . . . . . . . . . . . . .
Model 1
OCCUPATIONAL STATUS
TABLE 4 (Continued )
DEPENDENT VARIABLE
Occupational status . . . . . . . . . . . . .
Employed > 20 weeks . . . . . . . . . . .
Log of income . . . . . . . . . . . . . . . . .
Proportion coworkers same race . . . .
AVERAGE
TREATMENT EFFECT
.55
1.14
2.01
.01
2.08
.07
2.10**
.02
NUMBER
OF
CASES
Treated
Control
320
560
300
530
280
490
320
560
NOTE.Data are not weighted. Number of cases varies because of missing data on the dependent variables; analysis with a common case base yields identical results. In accordance with
NCES security procedures for restricted data, unweighted sample sizes have been rounded to
the nearest 10. Nos. in parentheses are SEs.
* P < .05.
** P < .01.
1152
TABLE 6
Sensitivity Analysis for Proportion of Same-Race Coworkers
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
POINT ESTIMATE
CONFIDENCE INTERVAL
Upper
Lower
Upper
Lower
2.12
2.13
2.15
2.16
2.17
2.18
2.18
2.19
2.20
2.20
2.21
2.12
2.11
2.11
2.09
2.08
2.07
2.06
2.05
2.04
2.03
2.02
2.16
2.17
2.18
2.19
2.20
2.21
2.21
2.22
2.23
2.24
2.25
2.09
2.07
2.06
2.05
2.04
2.03
2.02
2.01
.00
.01
.02
NOTE.Bounds are estimated on the basis of 320 matched pairs. In accordance with NCES
security procedures for restricted data, unweighted sample sizes have been rounded to the
nearest 10.
1153
Our results provide limited support for Braddocks 1980, 2009 perpetuation theory. We nd that African-American and other minority students
who enrolled in high schools with higher proportions of white students
tend to work in environments with more persons from dissimilar racial
backgrounds. At the same time, white students who attended schools with
lower proportions of whites found themselves in less white-dominated
workplacesa perpetuation of interracial contact for whites as well. The
results were largely unaffected by the context in which high school racial
compositions occurred. They were stable over time and impervious to
busing and internal resegregation. The presence of racial conict did not
change the overall outcomes for blacks, although Hispanic students who
attended schools characterized by moderate to serious levels of racial
conict did not obtain the benets for workplace integration that otherwise ensued.
While the effects for workplace integration were statistically signicant
and robust to model specication, they appeared modest in substantive
terms. For example, a policy that increased the proportion of whites in a
school from 20% to 50% would reduce the proportion of same-race coworkers for blacks by only about 3 percentage points. Moreover, blacks
occupational status, employment, and earnings were all unaffected by high
school racial composition, whether that composition was achieved by busing or other means.
Despite their modest size, the effects on workplace composition have
meaning in their own terms. They offer a direct test of perpetuation theory,
which in its simplest form states that racial isolation tends to perpetuate
itself across the life course Braddock 1980. Moreover, racially mixed work
environments are an indicator of social cohesion and, along with schools
and communities, are central to efforts to reduce racial exclusion. With that
said, our larger conclusion about high school racial composition must be tempered by the lack of impact across the range of social stratication outcomes.
Racially mixed high schools may contribute modestly to racially mixed workplaces, but they have little power to promote economic equality.
1154
1156
APPENDIX
TABLE A1
Descriptive Statistics for Internal Resegregation HSB Only
Mean
SD
Student level:
Workplace % same race . . . . . . . . . . . . . . . . . . . . .
Student test scores . . . . . . . . . . . . . . . . . . . . . . . . .
Student SES . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Male . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Black . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Other . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
N students . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
School level:a
School proportion white . . . . . . . . . . . . . . . . . . . . .
Busing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Proportion white busing . . . . . . . . . . . . . . . . . . .
Court order . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Proportion white court order . . . . . . . . . . . . . . . .
Academic track segment . . . . . . . . . . . . . . . . . . . . .
Proportion white academic track segment . . . . . .
Extracurricular activities segment . . . . . . . . . . . . . .
Proportion white extracurricular activities segment
Dissimilarity index . . . . . . . . . . . . . . . . . . . . . . . . .
Proportion white dissimilarity index . . . . . . . . . .
School SES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Private . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Midwest . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
South . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
West . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Rural . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Suburban . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
N schools . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Other dependent variables:
Occupational status N 5 2,660 . . . . . . . . . . . . . . .
Log of income N 5 2,470 . . . . . . . . . . . . . . . . . .
Employed > 20 weeks N 5 2,560 . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.65
49.54
2.07
.47
.23
.13
2,660
.32
8.91
.75
.50
.42
.34
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.70
.14
.08
.33
.20
.29
.21
.20
.14
.06
.04
2.03
.15
.20
.48
.08
.21
.50
240
.23
.34
.22
.47
.31
.24
.20
.16
.13
.06
.04
.45
.36
.40
.50
.28
.41
.50
........
........
........
35.67
9.89
.90
12.64
.89
.30
.
.
.
.
.
.
.
.
.
.
1157
TABLE A2
Results for the Propensity Score Model: OutcomeStudents School
in Grade 10 Was at Least 60% White
Intercept . . . . . . . . . . .
Male . . . . . . . . . . . . . .
Locus of control . . . . .
Self-concept . . . . . . . . .
Student SES . . . . . . . .
Student test . . . . . . . . .
Missing test . . . . . . . . .
Sports . . . . . . . . . . . . .
Art . . . . . . . . . . . . . . .
Academic . . . . . . . . . .
Other . . . . . . . . . . . . .
School resources . . . . .
Private . . . . . . . . . . . .
Rural . . . . . . . . . . . . .
Urban . . . . . . . . . . . . .
North . . . . . . . . . . . . .
Midwest . . . . . . . . . . .
West . . . . . . . . . . . . . .
School SES . . . . . . . . .
Student test North . .
Student test Midwest
Student test West . . .
Student test private .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Coefcient
SE
Odds Ratio
2.51
2.25
.03
2.35**
2.12
.021
.44
.52**
2.03
2.19
.12
2.041
21.50
2.03
21.53**
.42
.23
2.30
1.65**
.01
2.01
2.00
.03
.72
.17
.12
.11
.12
.01
.44
.19
.18
.18
.18
.02
1.78
.21
.22
1.25
1.35
1.99
.25
.03
.03
.04
.04
.60
.78
1.03
.71
.89
1.02
1.55
1.68
.97
.82
1.13
.96
.22
.97
.22
1.53
1.26
.74
5.23
1.01
.99
1.00
1.03
NOTE.N 5 890 students blacks only. School-level predictors refer to the students school
in eighth grade, and student-level predictors are from the base year eighth-grade survey. In
accordance with NCES security procedures for restricted data, unweighted sample sizes have
been rounded to the nearest 10.
1
P < .10.
* P < .05.
** P < .01.
Other . . . . . . . . . . . . . . . . . . . . . . . . .
Black . . . . . . . . . . . . . . . . . . . . . . . . .
Other . . . . . . . . . . . . . . . . . . . . . . . . .
Black . . . . . . . . . . . . . . . . . . . . . . . . .
Intercepts:
Reference group white . . . . . . . . . . . .
3.61
.46
1.63
2.73
24.70
3.00
36.23**
.46
1.12
.84
.50
.85
3.53
3.11
3.49
3.82
27.95
5.00
36.21**
.46
1.08
.86
.05
.86
.09
.20
2.17
.28
2.34
.30
9.97**
.03
2.10
.07
2.07
.08
.34
.30
2.79
.47
2.39
.51
9.97**
.03
2.09
.06
2.07
.08
LOG OF INCOME
Linear
OCCUPATIONAL
STATUS Linear
.64
.58
2.24
.72
2.97
.94
2.23**
.09
.08
.17
2.09
.22
.76
.89
2.82
1.23
2.96
1.71
2.24**
.09
.12
.17
2.09
.22
EMPLOYED > 20
WEEKS Binary
TABLE A3
Coefcients for Segregation in the Academic Track HSB Only
.19**
.05
2.30**
.08
.02
.11
.75**
.01
2.37**
.03
2.22**
.11
.21**
.07
2.39**
.12
.01
.19
.75**
.01
2.37**
.03
2.22**
.03
PROPORTION
COWORKERS SAME
RACE Linear
1160
23.12
6.54
6.92
8.61
25.27
8.44
2.58
1.69
2.02
2.48
1.02
2.95
2.10
2.19
1.93
3.46
2.48
3.98
240
2,560
2.81
.74
2.12
1.27
.20
1.08
240
2,470
.58
.61
21.10
.92
.01
.82
EMPLOYED > 20
WEEKS Binary
1
2.08
.18
.32
.32
.07
.41
240
2,660
.05
.15
2.30
.23
2.09
.28
PROPORTION
COWORKERS SAME
RACE Linear
NOTE.Models include school- and student-level control variables as in table 2. In accordance with NCES security procedures for restricted data,
unweighted sample sizes have been rounded to the nearest 10. Nos. in parentheses are SEs.
* P < .05.
** P < .01.
N schools . . . . . . . . . . . . . . . . . . . . . . . .
N students . . . . . . . . . . . . . . . . . . . . . . .
Other . . . . . . . . . . . . . . . . . . . . . . . . .
Black . . . . . . . . . . . . . . . . . . . . . . . . .
Other . . . . . . . . . . . . . . . . . . . . . . . . .
Black . . . . . . . . . . . . . . . . . . . . . . . . .
LOG OF INCOME
Linear
OCCUPATIONAL
STATUS Linear
TABLE A3 (Continued )
TABLE A4
Basic Model with Categorical Values for School Proportion White
NELS Eighth-Grade Sample
Outcomes
Intercepts:
Reference group white . . . . .
Black . . . . . . . . . . . . . . . . . .
Other . . . . . . . . . . . . . . . . . .
Slopes:
School 0%20% white:
Reference group white . . .
Black . . . . . . . . . . . . . . . . .
Other . . . . . . . . . . . . . . . . .
School 21%40% white:
Reference group white . . .
Black . . . . . . . . . . . . . . . . .
Other . . . . . . . . . . . . . . . . .
School 61%80% white:
Reference group white . . .
Black . . . . . . . . . . . . . . . . .
Other . . . . . . . . . . . . . . . . .
School 81%100% white:
Reference group white . . .
Black . . . . . . . . . . . . . . . . .
Other . . . . . . . . . . . . . . . . .
School-level controls:
Private . . . . . . . . . . . . . . . . .
Rural . . . . . . . . . . . . . . . . . .
Suburban . . . . . . . . . . . . . . .
Midwest . . . . . . . . . . . . . . . .
South . . . . . . . . . . . . . . . . . .
West . . . . . . . . . . . . . . . . . . .
Occupational Log of
Status
Income
Linear
Linear
39.34**
.28
.12
.89
2.08
.85
9.96**
.02
2.07
.05
2.03
.04
Employed >
20 Weeks
Binary
Proportion
Coworkers
Same Race
Linear
3.14**
.07
2.52**
.08
2.25
.13
.69**
.01
2.34**
.02
2.32**
.02
2.03
.04
.18**
.06
.10
.06
2.45
1.47
21.64
2.88
.90
2.13
2.12
.11
.14
.15
.18
.17
21.23**
.34
1.56**
.46
1.71**
.46
2.76
2.05
23.31
3.58
23.59
2.82
.20
.11
2.26
.21
2.13
.17
.28
.52
.12
.62
2.56
.73
2.03
.03
2.04
.06
.00
.06
21.20
1.29
22.22
3.14
2.96
1.92
.17
.09
2.10
.14
2.09
.14
2.18
.24
.58
.34
.64
.38
.04
.02
2.07
.05
2.05
.05
2.81
1.19
21.78
2.87
24.55
2.33
.10
.08
2.05
.12
2.03
.14
2.20
.21
.83**
.31
.75*
.35
.10**
.02
2.20**
.05
2.08
.05
.39
.78
21.23*
.60
2.67
.58
2.61
.62
2.53
.59
21.64*
.69
.04
.07
2.03
.05
.05
.04
2.06
.04
2.10*
.04
2.15**
.04
2.45**
.17
.10
.15
.15
.14
.22
.14
2.02
.13
.11
.15
.03
.02
.05**
.02
.00
.01
.04
.02
.00
.02
2.04*
.02
1161
Outcomes
School SES . . . . . . . . . . . . . .
Student-level controls:
Male . . . . . . . . . . . . . . . . . . .
Student SES . . . . . . . . . . . . .
Student test . . . . . . . . . . . . . .
N schools . . . . . . . . . . . . . . . . .
N students . . . . . . . . . . . . . . . .
Occupational Log of
Status
Income
Linear
Linear
.76
.61
23.92**
.37
3.17**
.35
.29**
.02
1,070
10,020
2.10*
.04
.38**
.02
.08**
.02
.01**
.00
1,060
9,140
Employed >
20 Weeks
Binary
Proportion
Coworkers
Same Race
Linear
.08
.13
2.01
.02
.57**
.08
2.17**
.06
.02**
.01
1,070
9,300
2.01
.01
2.01
.01
.00
.00
1,070
10,060
NOTE.In accordance with NCES security procedures for restricted data, unweighted
sample sizes have been rounded to the nearest 10. Nos. in parentheses are SEs.
* P < .05.
** P < .01.
REFERENCES
Alba, Richard. 2009. Blurring the Color Line: The New Chance for a More Integrated
America. Cambridge, Mass.: Harvard University Press.
Allison, Paul. 2001. Missing Data. Thousand Oaks, Calif.: Sage.
Allport, Gordon. 1954. The Nature of Prejudice. Reading, Mass.: Addison-Wesley.
An, Brian P., and Adam Gamoran. 2009. Trends in School Racial Composition in the
Era of Unitary Status. Pp. 19 47 in From the Courtroom to the Classroom: The
Shifting Landscape of School Desegregation, edited by C. E. Smrekar and E. B.
Goldring. Cambridge, Mass.: Harvard Education Press.
Anderson, Elijah. 1990. Streetwise: Race, Class, and Change in an Urban Community.
Chicago: University of Chicago Press.
Armor, David. 1972. The Evidence on Busing. Public Interest Summer: 90126.
Ashenfelter, Orley, William J. Collins, and Albert Yoon. 2006. Evaluating the Role of
Brown v. Board of Education in School Equalization, Desegregation, and the Income
of African Americans. American Law and Economics Review 8:213 48.
Austin, Peter. 2011. An Introduction to Propensity Score Methods for Reducing Effects
of Confounding in Observational Studies. Multivariate Behavioral Research
46:399424.
Bailey, Thomas, and Roger Waldinger. 1991. Primary, Secondary, and Enclave Labor
Markets: A Training Systems Approach. American Sociological Review 56:432 45.
Bates, Percy. 1990. Desegregation: Can We Get There from Here? Phi Delta Kappan
72 September: 817.
Becker, Sascha O., and Andrea Ichino. 2002. Estimation of Average Treatment Effects
Based on Propensity Scores. Stata Journal 2:35877.
Billings, Stephen B., David J. Deming, and Jonah E. Rockoff. 2014. School Segregation, Educational Attainment, and Crime: Evidence from the End of Busing in
Charlotte-Mecklenburg. Quarterly Journal of Economics 129 1: 43576.
BLS Bureau of Labor Statistics. 2013. The Employment SituationNovember 2013.
Washington, D.C.: Bureau of Labor Statistics, U.S. Department of Labor. http://www.bls
.gov/news.release/pdf/empsit.pdf.
1162
1163
1164
1165
1166
1167