Fall 22
lisawatson842
TheGenderGapinCharterSchoolEnrollment.pdf
https://doi.org/10.1177/0895904816673737
Educational Policy 2018, Vol. 32(5) 635 –663
© The Author(s) 2016 Reprints and permissions:
sagepub.com/journalsPermissions.nav DOI: 10.1177/0895904816673737
journals.sagepub.com/home/epx
Article
The Gender Gap in Charter School Enrollment
Sean P. Corcoran1 and Jennifer Jennings1
Abstract Many studies have investigated whether students in charter schools differ systematically from those in traditional public schools with respect to prior achievement, special education, or English Language Learner status. None, however, has examined gender differences in charter school enrollment. Using data for all U.S. public schools over 11 years, we find charters enroll a higher fraction of girls, a gap that has grown steadily over time and is larger in secondary grades and KIPP schools. We then analyze longitudinal student- level data from North Carolina to examine whether differential rates of attrition explain this gap. We find boys are more likely than girls to exit charters once enrolled, and gender differences in attrition are larger than in traditional schools. However, the difference is not large enough to explain the full enrollment gap between charter and traditional schools in North Carolina, suggesting gaps exist from initial matriculation.
Keywords charter schools, gender gap, attrition, mobility
Introduction
Considerable attention has been devoted to the ways in which charter school students in the United States differ from those attending traditional public schools, with respect to their prior achievement, race/ethnicity, socioeconomic
1New York University, New York City, NY, USA
Corresponding Author: Sean P. Corcoran, Institute for Education and Social Policy, New York University, 246 Greene Street, Suite 300, New York City, NY 10003, USA. Email: [email protected]
673737 EPXXXX10.1177/0895904816673737Educational PolicyCorcoran and Jennings research-article2016
636 Educational Policy 32(5)
background, special education, and English language learner status (Abdulkadiroğlu, Angrist, Dynarski, Kane, & Pathak, 2011; Angrist, Dynarski, Kane, Pathak, & Walters, 2010; Bifulco & Ladd, 2006; Buckley & Schneider, 2005; Dobbie & Fryer, 2011; Hoxby & Murarka, 2007; Imberman, 2011; Raymond, 2009; Tuttle et al., 2013; Zimmer et al., 2009). No study, however, has examined gender differences in charter school enrollment. This omission is somewhat surprising, as gender gaps in educational outcomes have long been of interest to academics and policy makers, have grown in recent decades in favor of girls, and are larger in disadvantaged communities where charters tend to locate (Buchmann, DiPrete, & McDaniel, 2008).
Whether there are gender differences in charter enrollment is important for several reasons. First, the distributional effects of charter schools ulti- mately depend on who attends them. To date, research on charter schools has emphasized estimating the causal effect of attending a charter school (Betts & Tang, 2008). Who enrolls—and who remains—in charters has been less studied (Zimmer & Guarino, 2013). Second, student sorting determines the peer composition of charter and non-charter schools alike (Bifulco, Ladd, & Ross, 2009; Booker, Gilpatric, Gronberg, & Jansen, 2008; Dee & Fu, 2004; Weiher & Tedin, 2002), and there is robust evidence that boys and girls per- form better when a larger fraction of their peers are girls (Black, Devereux, & Salvanes, 2013; Hoxby, 2000; Lavy & Schlosser, 2011; Whitmore, 2005). Third, a gender imbalance may reflect sorting into charters on non-cognitive skills, such as behavior, or preferences for distinctive educational environ- ments that differ by gender.
Taken together, the efficacy and distributional consequences of charter schools cannot be fully understood without attention to how students and families sort into schools. In this article, we provide the first analysis of char- ter enrollment and retention by gender. Analyzing enrollment data in a nation- wide panel of charter and non-charter public schools from 1999-2000 through 2010-2011, we estimate the gender gap in enrollment between charter schools and observationally similar schools, and examine how this gap varies across grade levels and over time as the charter sector has expanded. To address potential explanations for this gap, we use our national panel of schools and student-level data from North Carolina to examine whether differential rates of attrition can account for these enrollment differences, and whether gender gaps are larger in charter schools known to have strict behavioral norms.
In the following section, we offer several reasons why the female share of enrollment might vary between the charter and traditional public school sec- tors. We then briefly review the existing literature on charters and school choice for evidence of gender differences in enrollment. Following that, we describe the data and empirical methods used in this article, and present our
Corcoran and Jennings 637
main results. We conclude with a brief discussion of the implications of our findings for policy, practice, and future research on charter schools and school choice more broadly.
Why Might Charter Schools Enroll More Girls?
There are a number of reasons one might expect the gender composition of charter schools to differ from that of traditional public schools.
Student Preferences
Charter schools may have different appeal to boys and girls. Girls may be more likely to apply to charters because of differences in curriculum or pro- gram offerings, particularly in middle and high school, where schools differ in academic focus and extracurricular activities such as athletics. Previous research has found that girls are more likely to choose academically oriented programs (Hastings, Kane, & Staiger, 2006), are more engaged with school in general, and have higher educational aspirations (DiPrete & Buchmann, 2013). This may make them more likely to welcome a more challenging or differentiated curriculum offered at a charter school.1
Parental Preferences
Parents may prefer different environments for their sons and daughters. For example, parents concerned with safety, particularly as students enter ado- lescence, may prefer charters, which tend to be smaller than traditional pub- lic schools (shown later in Table 1). They may also anticipate greater benefits for girls at charter schools than boys. At least one study found school choice has greater academic benefits for girls than for boys (Hastings et al., 2006). Finally, they may specifically be attracted to schools with a non-traditional gender balance, such as single-sex schools (C. Jackson & Bisset, 2005; C. K. Jackson, 2012).
Educational Needs
Boys are more likely to be classified with special education needs (Halpern, 1997; Jennings & Beveridge, 2009), and many charters do not offer a full range of special education services (Wilkens, 2011). It may be that boys are less likely to apply to charter schools as a result, that boys are less likely to enroll upon winning an admissions lottery, or that boys are more likely to exit charter schools once enrolled, if needed services are unavailable.
638 Educational Policy 32(5)
Behavior
Girls have better behavior, on average, than boys (Bertrand & Pan, 2013; DiPrete & Jennings, 2012), and may be less opposed to strict disciplinary standards adopted by some charter schools, including KIPP and others adher- ent to a “no excuses” philosophy (Angrist et al., 2010; Thernstrom & Thernstrom, 2003). Once enrolled, boys may be more likely to exit charters after behavioral infractions, either on their own accord or at the prompting of the school. However, parents may be more inclined to enroll their boys in a charter school if they believe they will benefit from its disciplinary policies.
Academic Achievement
Attrition from charter schools may vary by gender due to academic perfor- mance. For example, if boys perform worse, on average, than girls, they may be more likely to return to a traditional school, either by choice or at the urg- ing of the school. Hanushek, Kain, Rivkin, & Branch (2007) found that char- ter families in Texas were particularly sensitive to their child’s academic performance, and were more likely to exit when their child was under-per- forming (see also Zimmer & Guarino, 2013).
Existing Evidence
Although many studies have examined the impact of charter schools on achievement, none has specifically addressed enrollment gaps by gender— whether they exist, when and where they emerge, how they have changed over time, and factors that explain them. Characteristics of the samples used in these studies, however, provide suggestive evidence about the gender gap in charter enrollment. For example, in an evaluation of Boston charter schools, 53% of students enrolled in charter elementary schools were girls, as compared with 48% in traditional elementary schools (Abdulkadiroğlu et al., 2011). The gap was two points smaller in middle school, but larger in high school, where 60% of charter school students were female, versus 51% of non-charter students. Comparable gaps have been reported in studies of charter schools in Chicago (Booker, Gill, Zimmer, & Sass, 2009; Hoxby & Rockoff, 2005), Baltimore (Baltimore City Public Schools, 2009), New York City (Hoxby & Murarka, 2007, 2009), and in KIPP middle schools (Tuttle et al., 2013).
Two lottery studies of charter schools offer some indication as to when gender gaps in enrollment emerge. In Hoxby and Murarka’s (2007) evalua- tion of charter schools in New York City, girls were slightly underrepresented among applicants to charter elementary schools but were more likely to attend when accepted. Across all years in their study, 49% of students who
Corcoran and Jennings 639
applied and were admitted (mostly in kindergarten and first grade) were girls, while 52% of those who accepted and enrolled were female. In a similar study of Boston charter schools, the female share of applicants to charter middle schools was 1.8 percentage points higher than in traditional schools (48.8 vs. 47.0), and the female share of applicants to charter high schools was 8.5 percentage points higher (59.3 vs. 50.8; Abdulkadiroğlu et al., 2011). There was little difference observed between applicants and enrollees.
Charter school attrition has been less studied, and most of these studies do not specifically examine differential attrition by gender. In their study of attrition from 19 KIPP middle schools, Nichols, Gill, Gleason, and Tuttle (2012) found that attrition from KIPP was similar to traditional public schools, but students who replaced leavers were more likely to be girls and higher achievers. Although 48% of students in the entering grade (fifth) of KIPP middle schools were boys, only 40% of students in the final grade (eighth) were. A report on Baltimore charters showed that boys were slightly more likely to leave elementary and middle charter schools (Baltimore City Public Schools, 2009). In that report, 52% and 54% of charter elementary and middle school leavers were male, respectively. In high school, however, leavers were disproportionately female (54%). As there was no comparison group, it was unclear whether these gaps in attrition were different from those in traditional public schools.
The gender composition of charter schools is important in light of research on peer effects, which finds the gender balance has a meaningful effect on student performance. Hoxby (2000), for example, showed that a 10 percent- age point change in the female share increases girls’ and boys’ math achieve- ment by 0.06 and 0.08 standard deviations, respectively. This suggests that exposure to a two-point difference in the female share (the average difference we find in this study) can increase math achievement by 0.01 to 0.02 standard deviations for every year of exposure. The cumulative impact would be greater in settings where the gender gap is larger.
Data and Method
We rely on two data sources for this study. The first is a panel of all schools in operation nationwide between 1999-2000 and 2010-2011, including charter schools, constructed from the Common Core of Data (CCD) from the National Center for Education Statistics.2 We use this panel to estimate the average dif- ference between charter and non-charter schools in female enrollment, and to provide descriptive evidence of how this gender gap varies by grade, state, and over time. The second is a longitudinal database of all students in Grades 3 to 12 who attended North Carolina public and charter schools between 2005- 2006 and 2010-2011. These data follow students across grades and between schools, allowing us to more closely examine attrition patterns.
640 Educational Policy 32(5)
We begin by using the national panel to estimate the average difference between charter and non-charter schools in the female share, controlling for observed characteristics of the school and their student populations. That is, we estimate the coefficient β1 in regressions such as the following:
pfemale charter X uit it it it= + + +( ) β β γ0 1 . (1)
In Equation 1, pfemaleit is the percent female in school i in year t, charterit is an indicator variable equal to one if school i is a charter school in year t, and X is a vector of time-varying school characteristics plausibly related to the school’s female share, described below. The error term uit includes factors other than X and the school’s charter status associated with the female enroll- ment share in year t.
Unlike achievement or socioeconomic status, for example, there are few geographic or school-level variables that are systematically associated with gender composition. The female share of live births in each year is about 48.8% in the United States (Martin, Hamilton, Ventura, Osterman, & Mathews, 2013) and the vast majority of schools fall within a few percentage points of this average. An obvious exception is single-sex schools, which in practice are rare among public schools.3 That said, there are reasons other than random variation why a public school’s female share may deviate from the natural rate. First, girls may be more or less likely to enroll in private or home school. Where private or homeschooled students constitute a meaning- ful share of the school-aged population, a higher proportion of girls in these schools would result in a correspondingly higher male share in public schools, and vice versa.4 Second, boys and girls may be retained at different rates, particularly in key transition grades (Hauser, 2004). Third, dropout rates vary by gender. Fourth, there are minor differences in sex ratios across demo- graphics groups. For example, the female share of births is slightly higher among Black mothers in the population (49.1%) and lower among Asian/ Pacific Islander mothers (48.5%; Chahnazarian, 1988; Martin et al., 2013).
We cannot explicitly account for all these factors in our analysis, but we attempt to account for systematic differences in the female share of enroll- ment across schools and communities in several ways. We limit our panel to schools operating in the 41 states with charter laws during this period, includ- ing D.C. (The 10 states without a charter law during this period were Alabama, Kentucky, Maine, Montana, North Dakota, Nebraska, South Dakota, Vermont, Washington, and West Virginia.) We exclude vocational and special education schools, which tend to be gender imbalanced, and schools that appear to be single sex, defined as having greater than 95% or fewer than 5% girls. We do include schools classified in the CCD as “other/alternative.” As explained in the online appendix, many charters appear to be miscoded in the
Corcoran and Jennings 641
CCD as “other/alternative,” especially in early years. As a robustness check, we report estimates excluding these schools as well as estimates that include single-sex schools. The number of charter schools that meet the above condi- tions range from about 1,500 in 1999 to 5,100 in 2010. The panel also includes approximately 80,000 non-charter schools annually that meet these criteria.
Charters are more likely to have non-traditional grade configurations, which may complicate comparisons with traditional schools if the female share varies by grade (from grade retention and dropout, for example). “Start-up” charter schools often build up their schools from successive cohorts such that, for example, the school enrolls sixth grade in Year 1, sixth and seventh grades in Year 2, and so on. We address grade-specific variation in the female share by estimating models with school-by-grade-level observations with grade-level effects and grade-specific state, county, or school district fixed effects.5 Thus, if schools serving 11th grade are disproportionately female due to higher dropout rates among 10th-grade boys, for example, the main grade-level effect will capture these differences. If this imbalance is greater in a particular urban dis- trict, the grade-specific district effect will account for this difference.
Taken together, our main empirical model for grade g in school i in year t is the following:
pfemale charter X uigjt it igt gj t ijt= + ( ) + + + +β β γ α λ0 1 . (2)
In Equation 2, the αgj are alternately grade-specific state, county, or school district effects, while the λt are year dummies. One would not expect the mean female share of schools to vary much systematically across geographic areas or time, but these will capture the effects of differential public school attendance (vs. private or home schools) and dropout incidence across locales, as well as any cohort variation. Finally, Xit includes observed charac- teristics of schools, including the share of enrollment by race/ethnicity, and the percent eligible for free lunch and for reduced price lunch (proxies for family income). While the CCD reports data on race by grade, free and reduced price lunch eligibility is only reported at the school level. Given multiple observations per school per year, we allow for clustered standard errors by school. All regressions are weighted by student enrollment in grade g in school i in year t such that results generalize to the average student in grade g. We experimented with alternative weights, and estimated all models without weights; the results were not substantively different.
To assess how the gender gap in charter enrollment has evolved over time, we estimate a version of Equation 2 separately by year, 1999-2010. To pro- duce grade-specific estimates of the gender gap, we estimate models in which charter is interacted with grade, and we summarize variation in the gender gap across states by estimating models separately by state for the most recent
642 Educational Policy 32(5)
year (2010). Finally, we use a subset of charters that converted from tradi- tional public schools to examine changes in the female share within schools over time as the school changes status.
In the second part of the article, we explore several mechanisms behind the gender gap in charter enrollment. First, using the CCD panel, we examine whether the gender gap is larger in charter management organization (CMO)- affiliated schools—and KIPP schools in particular—than in other charter schools. KIPP is known for its “no excuses” philosophy and strict behavioral expectations, and a larger gap in these schools could indicate that behavior is a relevant factor. As noted earlier, Nichols et al. (2012) found an initial gen- der gap in 19 KIPP middle schools that grew over time, as leavers were more likely to be male and late entrants more likely to be female. As an initial look at attrition within schools, we also use the CCD to examine changes in enroll- ment shares by gender across grades and years.
While the strength of the CCD is its national scope and availability for multiple years, it does not permit tracking of individual students over time. We therefore use longitudinal data from North Carolina to test for differential attrition from charter and traditional schools by gender. North Carolina is not necessarily representative of other states with charter schools, but its gender gap is comparable with the national average. (The most notable difference from other states is that North Carolina charters are more geographically dis- persed, rather than concentrated in urban areas.) It is also one of the few available data sources for observing mobility within and between charter schools. We use the Masterbuild compiled by the North Carolina Education Data Research Center, which includes all students enrolled in Grades 3 to 12 between 2005-2006 and 2010-2011. We first estimate the following model, separately by grade, for students in Grades 3 to 11 in charter schools:
Pr non structural move f female X Ssigt s sigt it t- ( ) ( )= + + + +β β γ θ λ0 1 .. (3)
Pr(non-structural movesigt) is the probability that charter school student s makes a non-structural move from school i between year t and t + 1, the λt are year dummies, Xsigt is a vector of student characteristics, and Sit is a vector of school locale and school type dummies (from CCD data). (We do not include school district effects here, as many school districts in North Carolina have only one charter school.) Following the research on student mobility, we define non-structural moves as an exit from school i when the student’s grade g is not a terminal grade in that school (Hanushek et al., 2004; Machin, Telhaj, & Wilson, 2006; Schwartz, Stiefel, & Cordes, in press; Xu, Hannaway, & D’Souza, 2009; Zimmer & Guarino, 2013). We determine whether grade g is a terminal grade in school i in year t by looking ahead to t + 1 to see whether grade g + 1 is offered in that school. As an example, a fourth grader leaving
Corcoran and Jennings 643
school i after year t would be coded as a non-structural move if i enrolled students through fifth grade in year t + 1. A fifth grader leaving i after year t, however, would not be coded as a non-structural move, as this would be a normal transition.6 We consider a move as any move between schools or an exit from North Carolina public schools altogether.
The coefficient β1 in Equation 3 can be interpreted as the female–male difference in the probability of a non-structural move, controlling for stu- dent and school characteristics, for students enrolled in charter schools in grade g. In X, we include controls for student race/ethnicity and eligibility for free or reduced price meals, their interactions, Limited English Proficiency (LEP), and special education status. Models for Grades 3 to 8 also include controls for reading and math scores (standardized by grade, subject, and year).
Estimates of Equation 3 capture differential attrition within the charter school sector, but cannot say whether gender differences in attrition are larger or smaller than those observed in traditional public schools. To that end, we estimate Equation 4, separately by grade level (3-5, 6-8, and 9-11), for all students in school districts that ever housed a charter school in North Carolina:
Pr non structural move f
female
charter fe
sigt
s
it- ( ) =
+ +
+
β β
β β
0 1
2 3 mmale
charter
X S
s
it
sigt it j t
× +
+ + +
γ θ ϕ λ
. (4)
Equation 4 adds an indicator variable charterit to distinguish enrollment in a charter school, and an interaction between females and charterit. In this case, β1 can be interpreted as the female–male difference in the probability of a non-structural move in traditional schools, β2 is the charter–non-charter difference for boys, and the difference-in-difference coefficient β3 is the extent to which the gender gap in attrition is larger (or smaller) in charter schools than in traditional public schools.
We explore the possibility of within-year attrition by re-estimating Models 3 and 4 using within-year moves as the dependent variable. To identify within- year moves, we use variables in the Masterbuild that report enrollment status at multiple points during the school year. We define a within-year move as being enrolled in school i in the fall but not in the spring. Finally, we look descriptively at the gender composition of new (incoming) students in North Carolina charter and traditional schools to test for differences at the point of intake. Due to data limitations, however, we are only able to do this for Grades 4 to 12. (For a student to be classified as new, we must be able to determine they were not a student in the school in the prior year.)
644 Educational Policy 32(5)
The Gender Gap in Enrollment in National Data
The National Gender Gap in Charter School Enrollment
We begin our look at the gender gap in charter school enrollment with our national panel of public schools. Descriptive statistics for schools in the panel, weighted by enrollment, are reported in Table 1 for the most recent year of the CCD (2010-2011) and for all years of the panel. In 2010-2011, the mean female share of enrollment was 48.8% overall (exactly the female share at birth), and 50.7% in charter schools, a gap of 1.9 percentage points. Charter schools were also more likely to enroll Black and free lunch–eligible students than the average school, to be located in large cities, to be classified as “other/ alternative” schools, and to serve a “mixed” grade span such as K-8.
A closer look at the full distribution of female enrollment shares for char- ter and non-charter schools in 1999 and 2010 (Figure 1) shows a shift toward greater female enrollment in charters over time. Charter schools in general
Figure 1. Percent of enrolled students who are female, charter and non-charter schools in 1999-2000 and 2010-2011. Note. Authors’ calculations using the CCD. Distribution for non-charters in 1999 is virtually identical to that in 2010, and thus is not shown for clarity of presentation. CCD = Common Core of Data.
645
T ab
le 1
. M
ea n
C ha
ra ct
er is
tic s
of A
ll Pu
bl ic
a nd
A ll
C ha
rt er
S ch
oo ls
, N at
io na
l P an
el .
A ll
ye ar
s 20
10 -2
01 1
A
ll sc
ho ol
s C
ha rt
er s
ch oo
ls D
iff er
en ce
A ll
sc ho
ol s
C ha
rt er
s ch
oo ls
D iff
er en
ce
Pe rc
en t
fe m
al e
48 .8
50 .3
1. 4
48 .8
50 .7
2. 0
Pe rc
en t
Bl ac
k 17
.3 30
.5 13
.3 16
.2 28
.6 12
.4 Pe
rc en
t H
is pa
ni c
21 .3
23 .8
2. 5
24 .3
27 .5
3. 2
Pe rc
en t
A si
an /P
ac ifi
c Is
la nd
er 4.
8 3.
6 −
1. 2
5. 1
3. 8
− 1.
4 Pe
rc en
t A
m er
ic an
In di
an /A
la sk
a N
at iv
e 1.
2 1.
3 0.
1 1.
2 1.
0 −
0. 2
Pe rc
en t
fr ee
lu nc
h el
ig ib
le 35
.2 39
.4 4.
2 41
.2 43
.4 2.
2 Pe
rc en
t el
em en
ta ry
48 .7
46 .1
− 2.
7 48
.8 44
.3 −
4. 5
Pe rc
en t
m id
dl e
20 .4
8. 2
− 12
.1 19
.6 8.
1 −
11 .5
Pe rc
en t
hi gh
28 .1
18 .2
− 9.
9 28
.2 17
.4 −
10 .8
Pe rc
en t
m ix
ed g
ra de
s pa
n 2.
8 27
.5 24
.8 3.
4 30
.2 26
.8 En
ro llm
en t
92 9.
9 77
0. 1
− 15
9. 8
90 5.
5 92
6. 2
20 .7
O th
er /a
lte rn
at iv
e sc
ho ol
1. 1
7. 0
5. 9
1. 1
5. 0
4. 0
La rg
e or
m id
-s iz
e ci
ty (
or u
rb an
fr in
ge )
72 .6
85 .6
13 .0
72 .2
86 .5
14 .3
La rg
e ci
ty 16
.4 39
.2 22
.9 15
.6 40
.1 24
.5 M
id -s
iz e
ci ty
14 .5
17 .2
2. 7
14 .6
16 .9
2. 2
U rb
an fr
in ge
o f l
ar ge
c ity
30 .3
22 .7
− 7.
6 30
.0 22
.4 −
7. 6
U rb
an fr
in ge
o f m
id -s
iz e
ci ty
11 .4
6. 5
− 4.
9 11
.9 7.
2 −
4. 8
La rg
e to
w n
1. 0
0. 5
− 0.
4 0.
9 0.
7 −
0. 2
Sm al
l t ow
n 6.
8 2.
7 −
4. 1
6. 4
2. 5
− 3.
9 N
90 6,
77 2
39 ,1
86 82
,5 16
5, 01
8
N ot
e. A
ut ho
rs ’ c
al cu
la tio
ns u
si ng
t he
C C
D . A
ll th
e ab
ov e
ar e
en ro
llm en
t- w
ei gh
te d
av er
ag es
o ve
r sc
ho ol
s. C
C D
= C
om m
on C
or e
of D
at a.
646 Educational Policy 32(5)
Figure 2. Mean percent female by grade, charter and non-charter schools, 1999- 2000 and 2010-2011. Note. Authors’ calculations using the CCD. All the above are enrollment-weighted averages across schools. CCD = Common Core of Data.
exhibit greater variation in the female share than non-charter schools, with a higher fraction of charters serving disproportionately female or dispropor- tionately male populations. For example, in 2010-2011, the percent female at the 75th and 90th percentiles of charter schools was 53.4% and 57.0%, respectively; in non-charters, these percentiles were 50.4% and 52.5%.
As noted earlier, differences in the female share are due in part to differ- ences in grade span. To account for these differences, Figure 2 shows the (enrollment-weighted) female share of enrollment by grade for charter and non-charter schools in 1999-2000 and 2010-2011. The female share is higher in charter schools in all grades in 2010-2011, ranging from 1.1 to 1.7 percent- age points in elementary school (K-5), 1.5 to 1.9 points in middle school (6-8), and 2.4 to 3.9 points in high school (9-12). The figure also shows the national gender gap in charter school enrollment increased in all grades between 1999 and 2010, particularly in secondary schools.
Because charters and the communities they serve differ from the average, unadjusted comparisons of the female share potentially reflect the role of fac- tors other than a gender-specific gap between charter schools and otherwise
Corcoran and Jennings 647
similar non-charter schools. In the next section, we report regression-adjusted estimates of the gender gap in charter school enrollment. We find that controls for grade, race and poverty composition, and grade-by-area fixed effects have almost no effect on the raw gender gaps shown in the preceding figures.
Regression-Adjusted Estimates of the Gender Gap in Charter School Enrollment
Table 2 and Figures 3 and 4 report regression-adjusted estimates of the gen- der gap in charter school enrollment for the most recent wave of school data (Table 2), for all years in our panel (Figure 3), and by grade (Figure 4). Following Equation 2, the dependent variable in each model is the percent female in grade g within a school i in year t, and observations are weighted by grade-level enrollment. Table 2 reports these results for 2010-2011. Each cell provides results from a separate regression, showing only the estimated coefficient on the charter indicator, interpreted as the average gap in the female share between charter and non-charter schools serving the same grades, similar populations, and in the same locations. The first row comes
Table 2. Regression Estimates of the Gender Gap in Charter School Enrollment, 2010-2011.
Fixed effect
N None State × Grade County × Grade District × Grade
Full sample Charter school 1.950***
(0.083) 1.993***
(0.084) 1.989***
(0.086) 1.888***
(0.100) 449,566
Full sample excluding other/alternative Charter school 1.809***
(0.083) 1.830***
(0.083) 1.847***
(0.085) 1.797***
(0.098) 432,542
Large cities Charter school 1.903***
(0.151) 1.999***
(0.146) 1.971***
(0.146) 1.805***
(0.152) 70,078
Large cities excluding other/alternative Charter school 1.787***
(0.151) 1.894***
(0.143) 1.861***
(0.143) 1.758***
(0.150) 66,680
Full sample excluding other/alternative, and including single-sex schools Charter school 1.725***
(0.095) 1.747***
(0.095) 1.752***
(0.098) 1.682***
(0.117) 433,238
Note. Dependent variable: percent of enrollment that is female. Standard errors in parentheses with clustering at the school level. Each cell is the result from a separate regression model, and reports the estimated coefficient on the charter school indicator. The regressions in columns 2 to 4 include grade- specific location effects and controls for enrollment shares by race/ethnicity and grade, the percent eligible for free lunch and eligible for reduced price lunch at the school level, and an indicator for “other/ alternative” schools. Regressions are weighted using school-by-grade enrollment. *p < .05. **p < .01. ***p < .001.
648 Educational Policy 32(5)
Figure 3. Regression-adjusted estimates of the gender gap in charter school enrollment, by year, 1999-2000 through 2010-2011. Note. Dotted lines indicate a 95% confidence interval estimate. Results are from annual regression models for the percent female using school-by-grade observations. The regressions include grade-specific school district effects and controls for enrollment shares by race/ ethnicity and grade, the percent eligible for free lunch and eligible for reduced price lunch at the school level, and an indicator for “other/alternative” schools. Regressions are weighted using school-by-grade enrollment.
from the full sample, while the second through fourth rows represent increas- ingly homogeneous samples of schools (excluding other/alternative schools, schools in large cities, and both). The final row uses the full sample excluding other/alternative schools, but including single-sex schools. The four columns represent model specifications without fixed effects, and with grade-specific state, county, and school district fixed effects, respectively.
Our regression-adjusted estimate of the gender gap in charter school enrollment for 2010-2011 is a statistically significant 1.8 to 2.0 percentage points (p < .001). In most cases, the inclusion of (grade-specific) state or county fixed effects has little effect on the point estimate, and including school district effects decreases the gap by 0.1 percentage point. Limiting the sample to “regular” public schools (excluding other/alternative schools) or to schools in large cities reduces the gender gap by only a small amount (0.083- 0.13 percentage points). Including single-sex schools in the sample also has a small effect on the gender gap, reducing it to 1.7 to 1.8 points. (The gender gap likely falls in this case because single-sex traditional public schools are more likely to be boys’ schools.)
In the interest of space, we do not report coefficient estimates for the other control variables, although they are reported in the online appendix. As
Corcoran and Jennings 649
Figure 4. Regression-adjusted estimates of the gender gap in charter school enrollment, by grade level, 2010-2011. Note. Dotted lines indicate a 95% confidence interval estimate. Results are from regression models for the percent female using school-by-grade observations. The regressions include grade-specific school district effects and controls for enrollment shares by race/ethnicity and grade, the percent eligible for free lunch and eligible for reduced price lunch at the school level, and an indicator for “other/alternative” schools. Regressions are weighted using school-by-grade enrollment.
expected, these controls do little to explain variation across schools in the female share. To provide one example, grades in schools with a one standard deviation higher Black enrollment share (roughly 24.8%) have a 0.32 per- centage point higher female share, on average. Similarly, a one standard deviation higher Hispanic enrollment share (about 27.4%) is associated with a 0.17 percentage point higher female share. Schools with a larger share of free lunch eligible students tend to have lower female shares, as do other/ alternative schools. All these are modest but meaningful differences in the gender balance. Because the gender gap in charter enrollment is larger with- out these covariates, however, it is important to include these controls.
Figure 3 shows estimates of the gender gap separately by year from 1999 to 2010. A 95% confidence interval is shown as a dotted line around each estimate. The trend over time is clear: The gender gap in charter enrollment increased steadily over this 11-year period. In 1999-2000 and 2000-2001, the gap was statistically indistinguishable from zero, but by 2010 had grown to 1.9 percent- age points. One can reject the hypothesis that the gender gap in later years (e.g., 2007-2010) is the same as that in the earlier years (e.g., 1999-2002).
In Figure 4, we present estimates from a model that fully interacts the char- ter indicator and grade level, allowing us to see how the gender gap varies by
650 Educational Policy 32(5)
grade. We again use district fixed effects, and report estimates for 2010-2011, comparable with those in Table 2. Figure 4 shows the gender gap, along with a 95% confidence interval for each estimate. This figure shows the gender gap exists as early as kindergarten, at 1.65 points. There is no statistically signifi- cant difference between the gap in kindergarten and other elementary/middle grades, although the point estimates suggest a gradual increase into seventh and eighth grades. After eighth grade, the gap is larger, at 2.6 points in ninth and 10th grades, 3.6 points in 11th grade, falling again in 12th (presumably due to dropout patterns). Although the point estimates in high school are sta- tistically indistinguishable, we can reject the hypothesis that the gender gap in high school is the same in early (e.g., first to fifth) grades.
We also estimated gender gaps for each state, again including (grade-spe- cific) district fixed effects. Results are available in the online appendix (Figure A1). All but two states have positive gender gaps, although not all are statistically different from zero. There is notable variation in the gap across states. In 2010, Louisiana had the largest gap at 5.7 percentage points, although others, including Arkansas, Pennsylvania, Massachusetts, and Texas, had gaps of 3.0 points or more. Much of the variation appears to be driven by gender gaps in middle school; there is less variability across states in grades K-5. Importantly for our later analysis, North Carolina has a gap near the national average, at 1.7 percentage points.
Estimates of the Gender Gap in Charter School Enrollment Using Conversion Charters
As an alternative test for whether a school’s charter status is associated with its female share, we examined changes in the female share of conversion charters— those that originated as traditional public schools but later changed status. Although conversions are not representative of the full population of charter schools, they provide a unique opportunity to observe how populations change within a school after it becomes a school of choice and the rules governing it change. For this, we obtained a complete list of conversion charters from the 16 states with the largest number of conversions during this period, including California (76), Ohio (68), Georgia (27), and Texas (25; see the online appendix for details).
Table 3 reports results from regressions using grade-by-school observa- tions in states where conversions occurred after 1999-2000. The first two columns represent a simple difference-in-difference, with one indicator vari- able for schools that converted to charter during this period (2000 or later) and a second for these same schools after conversion. The model in column 1 includes grade-specific district effects, while the model in column 2 uses school fixed effects, where the “post-conversion” effect is identified by
Corcoran and Jennings 651
variation within conversion schools over time. Columns 3 and 4 replace the simple “post” indicator with another indicating the number of years elapsed since conversion, which allows any effect of conversion to vary with time. All models include the full set of grade and school controls in Equation 2.
The results in Panel A show that conversion schools have, on average, a higher female enrollment share than comparable schools prior to conversion, a statistically significant difference of about 0.6 percentage points. In the years following conversion, we find the gender gap in enrollment between conversions and other public schools widens by 0.24 points, although the effect in the school fixed effects model is not statistically significant. The estimated coefficient on “years after conversion” in Panel A, columns 3 and 4, is small and statistically insignificant.
Many of the conversions used to identify effects in Table 3 are relatively recent, however. Roughly a quarter of conversion schools in our sample had been observed three or fewer years after conversion, which may not be suf- ficient to observe any enrollment change that might occur. Thus, Panel B reports results for the same models, defining conversion schools as those observed at least 3 years after conversion. In these cases, our point estimates for the post-conversion years are larger. To highlight one example, in the model with school effects, we find that the gender gap in enrollment between conversion and other public schools rose 0.36 points, on average, after
Table 3. Regression Estimates of the Gender Gap in Charter School Enrollment, Before and After Conversion.
(1) (2) (3) (4)
Fixed effect Fixed effect
District School District School
A. Full sample Conversion school 0.553*** (0.054) — 0.608*** (0.062) — Post-conversion 0.235** (0.080) 0.125 (0.124) — — Years after conversion — — 0.008 (0.009) 0.008 (0.012) B. Observed 3+ years post Conversion school 0.314** (0.117) — 0.454*** (0.088) Post-conversion 0.534*** (0.133) 0.362** (0.140) — — Years after conversion — — 0.088*** (0.021) 0.070** (0.023)
Note. Standard errors in parentheses. Each panel (A/B) and columns 1 to 4 set is the result of a separate regression model. The regressions include grade-specific location (or school) effects and controls for enrollment shares by race/ethnicity and grade, the percent eligible for free lunch and eligible for reduced price lunch at the school level, and an indicator for “other/alternative” schools. Regressions are weighted using school-by-grade enrollment. Only schools in states where charter conversions have occurred are included, and only conversion charters observed for at least 3 years after conversion are identified as conversions in Panel B. *p < .05. **p < .01. ***p < .001.
652 Educational Policy 32(5)
conversion, or 0.07 to 0.08 points per year. Again, although conversions are not necessarily representative of all charter schools, these results show that schools converting to charter status see an increase in their female share sev- eral years after conversion.
Potential Explanations for the Gender Gap in Enrollment
Our analysis of all U.S. schools found that charters enroll a higher fraction of girls than observationally similar traditional schools, and the gap is particu- larly large in the secondary grades. It is unclear, however, whether this gap is due to differences in the propensity to enroll in charter schools, attrition, or other factors. In this section, we begin by examining whether the gender gap is larger in CMO-affiliated charter schools (such as KIPP), which adhere to strict academic and behavioral expectations that may be a greater challenge for boys. We then use the CCD and student-level data from North Carolina to look at differential attrition by gender for charter and traditional public schools.
The Gender Gap in CMO and KIPP Charter Schools
Using data from the National Alliance for Public Charter Schools, we identi- fied all charters in the CCD affiliated with a CMO and specifically those affiliated with the KIPP network. An advantage of looking at CMOs and KIPP is that they are more often associated with a “no excuses” philosophy and its strict academic and behavioral expectations (Angrist, Pathak, & Walters, 2013; Thernstrom & Thernstrom, 2003). To the extent these expec- tations are deterrents to boys (or boys are more likely to exit such settings), we might expect to find a larger gender gap in enrollment between these and other schools. Late entrants to CMO and KIPP schools may also be less likely to be male (Nichols et al., 2012).
Table 4 reports estimates from models analogous to the ones in Table 2, with grade-specific district effects, splitting charter schools into CMO (or KIPP) charters and other charter types. In column 1, we see that the estimated gender gap in enrollment is not significantly larger in CMO- affiliated schools than in other charter schools (2.0 vs. 1.9 points). However, our estimate of the gender gap in KIPP schools is nearly 50% larger than that for other charter schools (2.9 vs. 1.9 points). However, although statistically different from zero, we cannot reject the hypothesis that the gap is comparable with that in other charter schools. (This is due in part to the comparably small sample of KIPP schools, ranging from 1 in 1999 to 82 in 2011.)
Corcoran and Jennings 653
Indirect Evidence on Differential Attrition Within Schools From the CCD
In Table 5, we use our national CCD panel to see whether enrollment patterns across grades and years are consistent with differential attrition by gender.
Table 5. Change in Female Enrollment Share Within Schools, Across Grades and Over Time.
(1) (2)
Charter
schools only Charter schools
Charter–non-charter difference
Grade 2 −0.002 (0.081) −0.005 (0.077) 0.230** (0.078) Grade 3 0.101 (0.082) 0.097 (0.078) 0.321*** (0.079) Grade 4 0.046 (0.083) 0.043 (0.079) 0.316*** (0.080) Grade 5 0.055 (0.088) 0.078 (0.083) 0.470*** (0.085) Grade 6 0.129 (0.083) 0.125 (0.078) 0.324*** (0.079) Grade 7 0.626*** (0.083) 0.618*** (0.078) 0.665*** (0.078) Grade 8 0.839*** (0.110) 0.904*** (0.103) 2.311*** (0.106) Grade 9 1.225*** (0.084) 1.326*** (0.077) 0.662*** (0.078) Grade 10 0.722*** (0.087) 0.824*** (0.080) 0.408*** (0.081) Grade 11 0.067 (0.091) 0.164* (0.082) −0.196* (0.083)
Note. Columns 1 and 2 represent two separate regression models using grade-year observations in the CCD between 1999 and 2009. Column 1 includes only charter schools, while column 2 includes both charter and traditional schools. The dependent variable is the change in percent female enrollment from grade g to grade g + 1 in years t and t + 1, within the same school. All models include year and school district effects. Standard errors reported in parentheses. CCD = Common Core of Data. *p < .05. **p < .01. ***p < .001.
Table 4. Regression Estimates of the Gender Gap in CMO and KIPP School Enrollment, 2010-2011.
(1) (2)
CMO KIPP
CMO or KIPP charter 1.999*** (0.199) 2.883*** (0.592) Other charter 1.871*** (0.109) 1.874*** (0.101)
Note. Dependent variable: percent of enrollment that is female. Standard errors in parentheses with clustering at the school level. See notes to Table 2 for further details about model specification. CMO = charter management organization. *p < .05. **p < .01. ***p < .001.
654 Educational Policy 32(5)
Because the data are at the school-by-grade rather than student level, these results are only suggestive with respect to attrition. For the models in this table, we identified “pseudo-cohorts” within each school, defined as grade g in year t, grade g + 1 in year t + 1, and so forth. Then, for each grade and year, we calculated the change in percent female enrollment within a school from year t to t + 1. A value of 1.5 for fourth grade, for example, would indicate an increase of 1.5 percentage points in the female share between fourth and fifth grades in the same school between year t and t + 1.
Column 1 of Table 5 reports the results of a regression for within-school between-grade changes in the percent female for charter schools. The esti- mated coefficients are positive for every grade except second, indicating the female share increases from grade to grade in the average charter school. The coefficients are only significant and meaningful in size in middle and high school, however. (The latter likely reflects dropout behavior.) For example, from seventh to eighth grade, there is an increase in the percent female of 0.6 percentage points, on average. To test whether this pattern is different from that observed in traditional schools, the regression in column 2 extends the sample to all schools. The rightmost column reports the estimated difference between charter and traditional schools in their grade-to-grade change in female share. In all grades, the difference is positive and statistically signifi- cant, indicating that the change in female share from grade to grade is larger in charter schools than traditional schools. The differences range from 0.2 percentage points in Grade 2 to 0.67 points in Grades 7 and 9. (The estimate for Grade 8 should be interpreted with caution, because a large number of charters terminate in Grade 8.)
These results suggest that charter schools retain girls at a modestly higher rate than boys, and that this retention gap is larger in charters than in tradi- tional public schools, especially in middle and high school (though the latter is likely due to dropouts). Again, these findings are only indirect because they rely on aggregate data from the CCD. We turn to student-level data to inves- tigate this question further.
The Gender Gap in Attrition From Charter and Traditional Schools
Next, we use North Carolina data for students in Grades 3 to 11 to estimate gender differences in attrition—the propensity to make a non-structural school move or a within-year move. All models pool data over the 5 years 2005-2006 to 2009-2010 and include student covariates and year effects; models with charter and traditional schools add school district effects.
Corcoran and Jennings 655
Panel A of Table 6 reports the estimated gender gap by grade in the pro- pensity to make a non-structural move. (The complete set of regression results is reported in the online appendix.) Column 1 limits the sample to students enrolled in charter schools (i.e., Model 3). The baseline probability of switching schools or exiting the state system is high in charter schools, roughly 21% in Grades 3 to 8 during this period (compared with 12% in tra- ditional public schools). Girls are less likely to exit charters than boys at every grade level, by about 1 to 3 percentage points, with the differences larger in the upper grades. (The gender differences are statistically significant at conventional levels.) Consistent with patterns observed in the CCD, the gender gap in attrition is somewhat larger in charter versus traditional schools. Column 2 of Table 6 reports the results for Model 4, which extends the sam- ple to all students. For charter students, the estimated gender gap in the pro- pensity to make a non-structural move is comparable with that in column 1. The rightmost column shows this gap is larger than in traditional schools, by about 1 percentage point. The difference is statistically significant in the ele- mentary and middle grades.
Table 6. Gender Gap in the Probability of a Non-Structural or Within-Year Move, North Carolina Schools.
(1) (2)
Charter schools Charter and traditional public schools
Female–male
gap Baseline
probability Female–male gap
in charters Charter–non-charter
difference in gap
A. Outcome: non-structural move Grades 3-5 −0.012** (0.004) 0.242 −0.014*** (0.004) −0.010** (0.004) Grades 6-8 −0.020*** (0.004) 0.218 −0.018*** (0.004) −0.009* (0.004) Grades 9-11 −0.026*** (0.006) 0.262 −0.028*** (0.006) −0.006 (0.006) B. Outcome: within-year move Grades 3-5 −0.004** (0.001) 0.026 −0.004** (0.001) −0.001 (0.001) Grades 6-8 −0.006*** (0.001) 0.028 −0.006*** (0.002) 0.004** (0.002) Grades 9-12 −0.013*** (0.003) 0.080 −0.014*** (0.003) 0.001 (0.003)
Note. In each columns 1 and 2, the rows (grade levels) represent separate regression models for the probability of making a non-structural (year-to-year) or within-year move. In column 1 the sample consists of students in Grades 3 to 11 in charter schools in 2006 to 2010, while in column 2, the sample is all students in Grades 3 to 11 in any North Carolina district that ever housed a charter school. (Panel B regressions also include Grade 12.) All models include controls for race/ethnicity and free or reduced price lunch status (and their interaction), LEP and special education status, locale, year, and school type dummies. Models for Grades 3 to 8 in Panel A also include reading and math scores, and the models in column 2 include school district fixed effects. Standard errors reported in parentheses. LEP = Limited English Proficiency. *p < .05. **p < .01. ***p < .001.
656 Educational Policy 32(5)
The estimates in Panel A do not capture within-year attrition, as the data provide the school of record at the time of end-of-grade testing.7 The Masterbuild, however, includes indicators of enrollment at earlier points in the year: the first day of fall testing, the first day of March testing, and the 20th day of school (the latter for 2006-2008 only). We repeated our analysis in Panel B, re-defining the outcome as the propensity to make a within-year move: enrolled in school s in the fall but not during end-of- grade testing. Note that Grade 12 can be added to the analysis of within- year movers.
On the whole, within-year attrition is low in North Carolina. For charter students, about 2.6% of students in Grades 3 to 8 enrolled in the fall but not in the spring. This percentage is higher in Grades 9 to 12, but includes stu- dents who formally dropped out during the year. In all grades, girls in charter schools are less likely to make within-year moves than boys, but the differ- ence is only about 0.5 percentage point. For within-year moves, the gender gap is not significantly different than in traditional schools, as shown in col- umn 2. In fact, in Grades 6 to 8, the gender gap in within-year attrition appears to be larger in traditional schools. These point estimates are small in magni- tude but statistically significant. Overall, it does not appear that within-year attrition is an important explanation for the gender gap in charter enrollment in North Carolina.
Finally, as a cursory look at the gender composition of incoming students in North Carolina—both new students and transfers—we report the female share of incoming students in charter and traditional schools in Table 7. The Masterbuild is more limited for this purpose, as we can only classify students as new to a school in Grades 4 to 12, and not at all points of entry. In the early grades, we find incoming students in charter schools—mostly transfers-in— are more likely to be female than incoming students in traditional schools, a
Table 7. Gender Composition of New Students, North Carolina Schools.
Percent female, charter schools
Percent female, traditional public schools Difference
Grades 4-5 50.3 48.4 1.97*** (0.05) Grades 6-8 49.0 48.5 0.6 (0.5) Grades 9-12 50.3 49.1 1.2 (0.6)
Note. Data consist of all students in Grades 4 to 12 who were new to their school in 2006 to 2010 (i.e., data were available for their school and grade in the prior year, but they were not observed in that school). Column 3 shows the difference in the percent female, along with the standard error from a t test for differences in means. ***p < .001.
Corcoran and Jennings 657
difference of 2 percentage points. The gender gap for incoming students in other grade levels is positive, but not statistically significant.
Taken together, there are notable differences in the gender composition of exiting and entering students in charter and traditional schools in North Carolina. However, on balance, these differences do not appear large enough to explain the full enrollment gap between charter and traditional schools in North Carolina, suggesting gaps also exist at matriculation.
Conclusion and Implications for Policy
In this article, we documented a gender gap in charter enrollment in two data sources: a national panel of schools and administrative data from North Carolina. This gap is comparable with that observed in charter school impact studies, in which gender was not a primary interest. In our most recent year of data, we found charter schools enroll a larger share of girls, with an average gap of about 1.9 percentage points, after controlling for differences in school characteristics and location. This gap was largest in high school, peaking at 3.1 points in 11th grade. Interestingly, this gap has not shrunk as the sector has expanded. On the contrary, we find the gender gap in enrollment grew steadily over time, and is now more than double its size in 1999-2000. There are some obvious limits to this growth, although in some states where charters have flourished, such as Louisiana, the gap is particularly large. Variation in school programs and practices may help illuminate these differences. For example, our look at KIPP schools, known for their demanding curriculum and strict behavioral expectations, reveals a gender gap nearly a full percentage point higher than in other charter schools (though not statistically different). This sug- gests preferences for, or ability to succeed in, particular types of educa- tional environments may be an important driver of gender gaps in enrollment, not only in charters but in all schools of choice.
Data from North Carolina allowed us to explore one potential explana- tion for the gender gap: differential attrition and retention. We find boys are more likely to make non-structural moves from charters than girls, a gap that is larger in charter schools than traditional public schools. This finding is consistent with our look at grade-to-grade changes in national data, where attrition is observed only indirectly. Although boys are more likely to make within-year moves than girls in charter schools, the gap was similar in traditional schools. Taken together, the gender gap in North Carolina appears to be a function of both differential attrition and selec- tion. Whether this finding differs in other states and contexts is an impor- tant question for future research.
658 Educational Policy 32(5)
To conclude, we highlight three key implications of our findings: (a) inter- est in attending—and the ability to persist in—charter schools may differ for boys and girls, (b) the success of charter schools may in part be a function of positive peer effects, and (c) the growth of charter schools may have conse- quences for the gender gap in achievement.
First, our findings suggest that families may be more likely to enroll girls in charter schools than boys. This may be due to differences in preferences as well as the availability of programmatic offerings and extracurricular activi- ties. The finding of larger gaps in middle and high school suggests students’ own interest in attending charter schools may be an important factor. Policy makers and practitioners should be cognizant of differences in school prefer- ences by gender, and their potential sorting effects, when offering more opportunities for choice. Schools should also be aware of the consequences of strict behavioral norms for recruitment and retention of students by gender. Whether some sorting by gender is positive or negative for educational out- comes is a separate but important question—a normative and an empirical one—and depends on what outcomes are valued and the importance of school “fit” for youth development.
Second, given what is known about the positive peer effects of girls for both boys and girls, the success of charter schools may be due at least in part to the gender balance in these schools. Although the gender gaps estimated here are not large enough to explain documented differences in charter and traditional school performance, they are meaningful in size, and only slightly smaller than the gaps in English Language Learner (ELL) (four points) and special education status (three points) that have commanded significant pol- icy attention. Differences in enrollment by gender are frequently overlooked by policy makers and researchers.
Finally, the rapid growth in charter enrollment over time may have impli- cations for the (also-growing) gender gap in achievement. That higher frac- tions of girls attend charter schools means that if charter schools are more or less effective than schools these students would have attended, they may exacerbate gender gaps in achievement and attainment. At present, the direc- tion of these impacts is not clear, as charters in some urban areas (i.e., New York, Boston, and Chicago) have been found to be more effective than their counterparts. In studies of other locales, however, charters have a more mixed record, and thus could increase, decrease, or leave unchanged gender differ- ences in performance.
Authors’ Note
Any remaining errors are solely our own.
Corcoran and Jennings 659
Acknowledgments
We thank the North Carolina Education Research Data Center for providing access to confidential student-level data. Annie Tan, Juli Simon Thomas, and Lila Nazar de Jaucourt lent valuable research assistance to this project. Julian Betts and other par- ticipants at the 2009 National Center on School Choice conference offered particu- larly helpful feedback.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publi- cation of this article.
Notes
1. Long and Conger (2013) investigate gender sorting across schools of all types in 2008 and find male students are underrepresented in charter and private schools, and over-represented in irregular public schools. Using data from Florida, the same authors link gender sorting in high school to the gender gap in college enrollment (Conger & Long, 2013).
2. Although charter schools have existed since 1992, the Common Core of Data (CCD) has only reliably identified these schools since 1998. Moreover, there appear to be a number of miscoded schools in the CCD. See the online appendix for details, and a description of how we identified likely miscodes.
3. In the CCD data, we estimate that 1.3% of schools are single sex (less than 5% or more than 95% female) and that these schools are disproportionately vocational and alternative schools. Among regular schools, less than 0.4% are single sex. Regardless of school type, male single-sex schools are more common than female.
4. Nationally, the female share in private schools is comparable with that in public schools (49.0%), although the share is slightly higher (49.3%) in urban areas (Broughman, Swaim, & Keaton, 2008). Estimates of the female share of home- schooled students vary over time from 51% in 1999 to 58% in 2007 (Snyder & Dillow, 2013), though the overall percent of students homeschooled is small (1.3%) and more prevalent in rural areas.
5. Not all charter schools in the CCD are administratively linked to a local school district. As explained in the online appendix, we linked schools to geographic school districts using their spatial coordinates and geographic information sys- tems (GIS) software.
6. A fourth grader repeating fourth grade in a different school would be a non- structural move in this example.
7. A state report in 2001 found a higher rate of within-year attrition overall among charter versus traditional schools in North Carolina (Noblit & Corbett, 2001).
660 Educational Policy 32(5)
References
Abdulkadiroğlu, A., Angrist, J. D., Dynarski, S. M., Kane, T. J., & Pathak, P. A. (2011). Accountability and flexibility in public schools: Evidence from Boston’s charters and pilots. The Quarterly Journal of Economics, 126, 699-748.
Angrist, J. D., Dynarski, S. M., Kane, T. J., Pathak, P. A., & Walters, C. R. (2010). Inputs and impacts in charter schools: KIPP Lynn. The American Economic Review, 100, 239-243.
Angrist, J. D., Pathak, P. A., & Walters, C. R. (2013). Explaining charter school effec- tiveness. American Economic Journal: Applied Economics, 5(4), 1-27.
Baltimore City Public Schools. (2009). Charter school report: 2005-06 to 2007- 08 school years. Baltimore, MD: Division of Research Assessment and Accountability, Baltimore City Public Schools.
Bertrand, M., & Pan, J. (2013). The trouble with boys: Social influences and the gen- der gap in disruptive behavior. American Economic Journal: Applied Economics, 5(1), 32-64.
Betts, J. R., & Tang, Y. E. (2008). Value-added and experimental studies of the effect of charter schools on student achievement: A literature review. Seattle, WA: Center for Reinventing Public Education.
Bifulco, R., & Ladd, H. F. (2006). The impacts of charter schools on student achieve- ment: Evidence from North Carolina. Education Finance and Policy, 1, 50-90.
Bifulco, R., Ladd, H. F., & Ross, S. L. (2009). The effects of public school choice on those left behind: Evidence from Durham, North Carolina. Peabody Journal of Education, 84, 130-149.
Black, S. E., Devereux, P. J., & Salvanes, K. G. (2013). Under pressure? The effect of peers on outcomes of young adults. Journal of Labor Economics, 31, 119-153.
Booker, K., Gill, B., Zimmer, R., & Sass, T. R. (2009). Achievement and attainment in Chicago charter schools. Santa Monica, CA: RAND.
Booker, K., Gilpatric, S. M., Gronberg, T., & Jansen, D. (2008). The effect of charter schools on traditional public school students in Texas: Are children who stay behind left behind? Journal of Urban Economics, 64, 123-145.
Broughman, S. P., Swaim, N. L., & Keaton, P. W. (2008). Characteristics of private schools in the United States: Results from the 2005–2006 private school universe survey (NCES 2008-315). Washington, DC: National Center for Education Statistics, Institute of Education Sciences, U.S. Department of Education.
Buchmann, C., DiPrete, T. A., & McDaniel, A. (2008). Gender inequalities in educa- tion. Annual Review of Sociology, 34, 319-337.
Buckley, J., & Schneider, M. (2005). Are charter school students harder to educate? Evidence from Washington, D.C. Educational Evaluation and Policy Analysis, 27, 365-380.
Chahnazarian, A. (1988). Determinants of the sex ratio at birth: Review of recent literature. Social Biology, 35, 214-235.
Conger, D., & Long, M. C. (2013). Gender gaps in college enrollment: The role of gender sorting across public high schools. Educational Researcher, 42, 371-380.
Corcoran and Jennings 661
Dee, T. S., & Fu, H. (2004). Do charter schools skim students or drain resources? Economics of Education Review, 23, 259-271.
DiPrete, T. A., & Buchmann, C. (2013). The rise of women: The growing gender gap in education and what it means for America’s schools. New York, NY: Russell Sage Foundation.
DiPrete, T. A., & Jennings, J. L. (2012). Social and behavioral skills and the gender gap in early educational achievement. Social Science Research, 41, 1-15.
Dobbie, W., & Fryer, R. G. (2011). Are high-quality schools enough to increase achievement among the poor? Evidence from the Harlem Children’s Zone. American Economic Journal: Applied Economics, 3(3), 158-187.
Halpern, D. F. (1997). Sex differences in intelligence: Implications for education. American Psychologist, 52, 1091-1102.
Hanushek, E. A., Kain, J. F., & Rivkin, S. G. (2004). Disruption versus Tiebout improvement: The costs and benefits of switching schools. Journal of Public Economics, 88, 1721-1746.
Hanushek, E. A., Kain, J. F., Rivkin, S. G., & Branch, G. F. (2007). Charter school quality and parental decision making with school choice. Journal of Public Economics, 91, 823-848.
Hastings, J. S., Kane, T. J., & Staiger, D. O. (2006). Gender and performance: Evidence from school assignment by randomized lottery. The American Economic Review, 96, 232-236.
Hauser, R. M. (2004). Progress in schooling. In K. M. Neckerman (Ed.), Social inequality (pp. 271-318). New York, NY: Russell Sage Foundation.
Hoxby, C. M. (2000, August). Peer effects in the classroom: Learning from gender and race variation (NBER Working Paper No. 7867). Cambridge, MA: National Bureau of Economic Research. Retrieved from http://www.pugetsound.edu/files/ resources/4244_APA_Color.pdf.
Hoxby, C. M., & Murarka, S. (2007). New York City’s charter schools overall report. Cambridge, MA: New York City Charter Schools Evaluation Project.
Hoxby, C. M., & Murarka, S. (2009, April). Charter schools in New York City: Who enrolls and how they affect their students’ achievement (NBER Working Paper No. 14852). Cambridge, MA: National Bureau of Economic Research.
Hoxby, C. M., & Rockoff, J. E. (2005, May). The impact of charter schools on student achievement (Working paper). Cambridge, MA: Harvard University.
Imberman, S. A. (2011). The effect of charter schools on achievement and behavior of public school students. Journal of Public Economics, 95, 850-863.
Jackson, C., & Bisset, M. (2005). Gender and school choice: Factors influencing par- ents when choosing single-sex or co-educational independent schools for their children. Cambridge Journal of Education, 35, 195-211.
Jackson, C. K. (2012). Single-sex schools, student achievement, and course selection: Evidence from rule-based student assignments in Trinidad and Tobago. Journal of Public Economics, 96, 173-187.
Jennings, J. L., & Beveridge, A. A. (2009). How does test exemption affect schools’ and students’ academic performance? Educational Evaluation and Policy Analysis, 31, 153-175.
662 Educational Policy 32(5)
Lavy, V., & Schlosser, A. (2011). Mechanisms and impacts of gender peer effects at school. American Economic Journal: Applied Economics, 3(2), 1-33.
Long, M. C., & Conger, D. (2013). Gender sorting across K–12 schools in the United States. American Journal of Education, 119, 349-372.
Machin, S., Telhaj, S., & Wilson, J. (2006). The mobility of English school children. Fiscal Studies, 27, 253-280.
Martin, J., Hamilton, B. E., Ventura, S. J., Osterman, M. J. K., & Mathews, T. J. (2013). Births: Final data for 2011. Hyattsville, MD: National Center for Health Statistics.
Nichols-Barrer, I., Gill, B. P., Gleason, P., & Tuttle, C. C. (2012). Student selection, attrition, and replacement in KIPP Middle Schools. Princeton, NJ: Mathematica Policy Research.
Noblit, G. W., & Corbett, D. (2001). North Carolina charter school evaluation report. Raleigh, NC: Evaluation Section, Division of Accountability Services, Instructional and Accountability Services.
Raymond, M. (2009). Multiple choice: Charter school performance in 16 states. Stanford, CA: Center for Research on Education Outcomes.
Schwartz, A. E., Stiefel, L., & Cordes, S. A. (in press). Moving matters: The causal effect of school mobility on student performance. Education Finance and Policy.
Snyder, T. D., & Dillow, S. A. (2013). Digest of Education Statistics 2012 (NCES 2014-015). Washington, DC: National Center for Education Statistics, Institute of Education Sciences, U.S. Department of Education.
Thernstrom, A., & Thernstrom, S. (2003). No excuses: Closing the racial gap in learning. New York, NY: Simon & Schuster.
Tuttle, C. C., Gill, B., Gleason, P., Knechtel, V., Nichols-Barrer, I., & Resch, A. (2013). KIPP middle schools: Impacts on achievement and other outcomes. Princeton, NJ: Mathematica Policy Research.
Weiher, G. R., & Tedin, K. L. (2002). Does choice lead to racially distinctive schools? Charter schools and household preferences. Journal of Policy Analysis and Management, 21, 79-92.
Whitmore, D. (2005). Resource and peer impacts on girls’ academic achievement: Evidence from a randomized experiment. The American Economic Review, 95, 199-203.
Wilkens, C. P. (2011). Students with disabilities in urban Massachusetts charter schools. Teachers College Record. Retrieved from http://www.tcrecord.org/ Content.asp?ContentID=16389
Xu, Z., Hannaway, J., & D’Souza, S. (2009). Student transience in North Carolina: The effect of school mobility on student outcomes using longitudinal data. Washington, DC: Urban Institute/CALDER Center.
Zimmer, R. W., Gill, B., Booker, K., Lavertu, S., Sass, T. R., & Witte, J. (2009). Charter schools in eight states: Effects on achievement, attainment, integration, and competition. Santa Monica, CA: RAND.
Zimmer, R. W., & Guarino, C. M. (2013). Is there empirical evidence that charter schools “push out” low-performing students? Educational Evaluation and Policy Analysis, 35, 461-480.
Corcoran and Jennings 663
Author Biographies
Sean P. Corcoran is an Associate Professor of Economics and Education Policy at New York University. His research focuses on human capital in the teaching profes- sion, school choice, and public finance. His published papers have examined long-run trends in teacher quality, the impact of income inequality and school finance reform on education funding in the United States, the properties of “value-added” measures of teacher effectiveness, and the high school choices of middle school students in New York City.
Jennifer Jennings is an Associate Professor in the Department of Sociology at New York University. Her research focuses on the effects of accountability systems on racial, gender, and socioeconomic inequality in educational outcomes; teacher and school effects on outcomes beyond test scores; and the effects of non-cognitive skills on cognitive outcomes.
