ArticlePDF Available

Race, Poverty and SAT Scores: Modeling the Influences of Family Income on Black and White High School Students' SAT Performance

Authors:

Abstract and Figures

Background Educational policy makers and test critics often assert that standardized test scores are strongly influenced by factors beyond individual differences in academic achievement such as family income and wealth. Unfortunately, few empirical studies consider the simultaneous and related influences of family income, parental education, and high school achievement on college admissions test scores. Focus Of Study This research was animated by the nagging question of the association of family income with SAT performance. For example, is the relationship between family income and SAT performance non-linear? Does the relationship differ markedly by race? More importantly, how strong are the effects of poverty on SAT performance? Research Design This study is a secondary analysis of a large national sample of Black and White college-bound high school students who took the SAT in 2003 (N = 781,437). Data Collection and Analysis Employing data from the College Board's Student Descriptive Questionnaire, this study used structural equation modeling (SEM) to estimate the effects of family income on SAT scores for Black and White examinees accounting for the simultaneous effects of parental education and high school achievement. Findings/Results Results suggest the effects of family income on SAT scores, though relatively modest in contrasts to high school achievement, are substantial, non-linear, and nearly twice as large for Black students. Moreover, the unstandardized direct effect of high school achievement on SAT performance is not enough to address the substantial effects of poverty for Black students. Conclusions/Recommendations The findings are discussed with respect to social inequality and educational opportunity in college admissions.
Content may be subject to copyright.
Teachers College Record Volume 115, 040306, April 2013, 33 pages
Copyright © by Teachers College, Columbia University
0161-4681
1
Race, Poverty and SAT Scores: Modeling
the Influences of Family Income on Black
and White High School Students’ SAT
Performance
Ezekiel J. Dixon-Román
University of Pennsylvania
Howard T. Everson
City University of New York
John J. McArdle
University of Southern California
Background: Educational policy makers and test critics often assert that standardized test
scores are strongly influenced by factors beyond individual differences in academic achieve-
ment such as family income and wealth. Unfortunately, few empirical studies consider the
simultaneous and related influences of family income, parental education, and high school
achievement on college admissions test scores.
Focus Of Study: This research was animated by the nagging question of the association of
family income with SAT performance. For example, is the relationship between family
income and SAT performance non-linear? Does the relationship differ markedly by race?
More importantly, how strong are the effects of poverty on SAT performance?
Research Design: This study is a secondar y analysis of a large national sample
of Black and White college-bound high school students who took the SAT in 2003 (N =
781,437).
Data Collection And Analysis: Employing data from the College Board’s Student Descriptive
Questionnaire, this study used structural equation modeling (SEM) to estimate the effects of
27918a_TCR_April2013_text_Layout 1 3/26/13 3:05 PM Page 1
Teachers College Record, 115, 040306 (2013)
family income on SAT scores for Black and White examinees accounting for the simultane-
ous effects of parental education and high school achievement.
Findings/Results: Results suggest the effects of family income on SAT scores, though rela-
tively modest in contrasts to high school achievement, are substantial, non-linear, and
nearly twice as large for Black students. Moreover, the unstandardized direct effect of high
school achievement on SAT performance is not enough to address the substantial effects of
poverty for Black students.
Conclusions/Recommendations: The findings are discussed with respect to social inequal-
ity and educational opportunity in college admissions.
INTRODUCTION
Over the past half century many criticisms have been leveled at standard-
ized tests in education and, in particular, college admissions tests such as
the SAT or ACT (see, for example, Crouse & Trusheim, 1988; Elert, 2008;
National Association of College Admissions Counselors, 2008). One of
the more vocal critics has been Harvard Law Professor Lani Guinier
(Guinier & Torres, 2002). Guinier and others have suggested that the
SAT is not useful in college admissions because, at its core, it is a “wealth
test”—more a measure of family socioeconomic status than academic
potential. While interpretations of these criticisms have been oversimpli-
fied,1such claims have been challenged by others who have examined the
relationship between family wealth and academic achievement, and
report small or negligible direct effects on achievement test scores
(Dixon-Román, 2007; Orr, 2003; Phillips, Brooks-Gunn, Duncan,
Klebanov, & Crane, 1998; Yeung & Conley, 2008).
Although the relationship between wealth and achievement has been
small, the relationship between family income and academic achieve-
ment appears to be more substantial (Anyon, 1997; Bowen & Bok, 1998;
Dixon-Román, 2007; Duncan & Brooks-Gunn, 1997a; Duncan, Huston, &
Weisner, 2007; Everson & Millsap, 2004; Hedges & Nowell, 1998,1999;
Jencks & Phillips, 1998; Miller, 1997; Rothstein, 2004; Wightman, 1995).
Indeed, the consistent relationship between family income and achieve-
ment has also been found with SAT performance (Benners & Everson,
2009; Bowen & Bok, 1998; Camara & Schmidt, 1999; Everson & Millsap,
2004; Zwick, 2004). For the most part, these studies have examined the
relationship between family income and SAT performance by (1) assum-
ing income is linearly related to SAT performance, (2) assuming no
differential associations of family income by race, and (3) not modeling
explicitly the direct and indirect effects of family income after control-
ling for the combined effects of parental education and high school
2
27918a_TCR_April2013_text_Layout 1 3/26/13 3:05 PM Page 2
TCR, 115, 040306 Influence of Family Income
achievement (Camara & Schmidt, 1999; Everson & Millsap, 2005; Marsh
& Kleitman, 2002; Zwick, 2004).
These assumptions have underestimated and simplified the association
between family income and SAT performance. The assumed linear asso-
ciation between family income and SAT performance, we argue and
empirically demonstrate, has led to the underestimation of the associa-
tion. By conducting multiple group analyses, we examined to what extent
the association of family income (as well as other estimates of associa-
tion) substantially varied between Black and White test-takers. Moreover,
the analyses of the direct and indirect effects of family income enabled
us to decompose the unique and independent contribution of family
income with SAT performance, controlling for the effects of parental
education and high school achievement as well as how family income’s
association is additionally working indirectly via these other variables.
These analyses provide a better empirical understanding of the dynamic
influence that family income has on the performance of the SAT for
Black and White test-takers.
In addition, our search of the literature uncovered no studies that
examined the direct and indirect effects of extremely low levels of family
income (i.e., poverty) on SAT performance. The sociological and eco-
nomics literature has long established that family income has an increas-
ing concave relationship with child outcomes, thus indicating a
non-linear relationship and larger marginal return for lower income lev-
els (e.g., poverty) (Becker & Tomes, 1979, 1986; Blau, 1999; Dixon-
Román, 2007; Duncan & Brooks-Gunn, 1997a; Conley, 1999; Mazumder,
2005). This suggests that the previous literature that has assumed and
estimated a linear relationship between family income and SAT perfor-
mance has likely underestimated the association. Thus, our research was
animated by the nagging question of the association of family income
with SAT performance. For example, is the relationship between family
income and SAT performance non-linear? Does the relationship differ
markedly by race? More importantly, how strong are the effects of poverty
on SAT performance? And similarly, to what extent does poverty account
for the problematic Black-White performance differences on the SAT?
The current literature, as we show, is incomplete and leaves us with
incomplete answers about the effects of poverty on SAT performance.
With a renewed added emphasis on the achievement of students from
disadvantaged backgrounds, we believe it is important to examine care-
fully the economic and academic factors associated with differences in
performance on standardized tests – particularly, high stakes college
admissions tests. The examination of these questions will reveal whether
the association of family income with SAT performance has been under-
3
27918a_TCR_April2013_text_Layout 1 3/26/13 3:05 PM Page 3
Teachers College Record, 115, 040306 (2013)
estimated, particularly for Black test-takers or those test-takers living in
poverty, and to what extent this may account for Black and White differ-
ences in SAT performance.
We begin by discussing the research on the influences of extremely low
levels of family income (i.e., poverty) on academic achievement. After
setting the context for our study, we then move to a more focused discus-
sion of the literature on extremely low levels of family income and per-
formance on high stakes tests. This is followed by a description of our
sample and data sources, and an outline of the structural and measure-
ment models developed to guide our analyses. The fourth section of the
paper presents the results of our model fitting efforts, and offers inter-
pretations of the model parameters. The paper concludes with a discus-
sion of our findings and places them in the larger context of social
inequality and educational opportunity.
LITERATURE REVIEW
INCOME, POVERTY AND ACHIEVEMENT
Parental income and other indicators of socioeconomic status are related
to various educational outcomes (Bowen & Bok, 1998; Dixon-Román,
2007; Duncan & Brooks-Gunn, 1997a; Jencks & Phillips, 1998; Orr, 2003;
Phillips et al., 1998; Rothstein, 2004; Sexton, 1961; Yeung & Conley, 2008;
White, 1982; White et al., 1993). Recent research suggests that income has
a substantial effect on academic achievement and accounts for a mean-
ingful proportion of the score gap between Black and White test-takers on
most achievement measures (Blau, 1999; Bowen & Bok, 1998; Datcher-
Loury, 1989; Dixon-Román, 2007; Dooley & Stewart, 2004; Duncan &
Brooks-Gunn, 1997a; Jencks & Phillips, 1998; Orr, 2003; Phillips, Brooks-
Gunn, Duncan, Klebanov, & Crane, 1998; Rothstein, 2004; Sirin, 2005).
Using data from the National Longitudinal Survey of Youth (NLSY), Blau
(1999) found meaningful positive effects of permanent income (i.e., a
multi-year average of income) on the PIAT mathematics and reading
comprehension scores, as well as on the PPVT-R. Phillips et al. (1998) not
only found a meaningful positive effect of income on the PPVT-R using
data from the Children of the National Longitudinal Study of Youth (CNLSY),
but levels of income were also related to meaningful reductions in the
Black-White differences in academic achievement. Moreover, Sirin’s
(2005) meta-analysis of the research literature for the decade from 1990
and 2000 on socioeconomic status and achievement indicated that, on
average, there was a modest size effect correlation coefficient (0.29) for
family income on academic achievement.
4
27918a_TCR_April2013_text_Layout 1 3/26/13 3:05 PM Page 4
TCR, 115, 040306 Influence of Family Income
These consistent, meaningful, and positive effects of income on acade-
mic achievement have been found not only with data from the United
States, but also in studies using international achievement data.
Aughinbaugh and Gittleman (2003), for example, used the NLSY data
and data from the National Child Development Study of Great Britain to
examine the comparative differences in the effects of income on achieve-
ment. Their results suggested that the relationship between income and
achievement test scores is similar to what has been reported in the
United States. Similar findings also come from work in Canada, using the
National Longitudinal Survey of Canadian Youth. Dooley and Stewart (2004)
examined the magnitude of the effect of income on three measures of
cognition—including the PPVT and modified versions of the Canadian
Achievement Tests for mathematics and reading comprehension. Their
results indicated that the effect of income was meaningful and positive,
but they suggested it was likely smaller than conventional estimates
reported in the literature.
Although the majority of these studies have consistently found mean-
ingful, positive effects of income on achievement, Mayer’s (1997) work
provides a more complicated perspective. Using data from the NLSY,
Mayer argued that low-income parents may differ from middle- or high-
income parents with respect to social adjustment, enthusiasm, depend-
ability, academic skills, and motivation, suggesting a spurious income
effect. In other words, she argued it is these social and emotional differ-
ences, rather than differences in income, that account for the differences
in children’s academic achievement outcomes.
While others have argued that what Mayer (1997) is referring to as a
spurious effect, is in fact an indirect or unobserved variable effect. For
example, Duncan, Huston, and Weisner (2007) found that earnings sup-
plements do matter for children’s academic achievement in poor fami-
lies. In a random assignment study, they evaluated the effect of the New
Hope Program that provided poor families with an earnings supplement,
subsidized health insurance, subsidized childcare, and a temporary com-
munity-service job while searching for employment. Their evaluation
found that, on average, children in the program scored higher than chil-
dren in the control group. These findings speak to the non-spurious
effects of income for poor families, and the possible meaningful effect of
poverty on academic achievement.
Moreover, poverty is known to be related to poor nutrition, exposure to
lead poisoning, low-birth weight, attention deficit hyperactive disorder,
learning disabilities, lack of health insurance, poor quality housing, poor
quality schooling, school per-pupil expenditures, parenting practices, high
school equivalent or lower parental education, parental unemployment,
5
27918a_TCR_April2013_text_Layout 1 3/26/13 3:05 PM Page 5
Teachers College Record, 115, 040306 (2013)
single-parent homes, et cetera (Birch & Gussow, 1970; Bowles, Durlauf, &
Hoff, 2006; Lareau, 2003; Lee & Wong, 2004; Fass & Cauthen, 2008). In a
report from the National Center for Children in Poverty, Fass and Cauthen
(2008) reported that 18% of children in the United States live below the
federal poverty level (i.e., a family income of less than $22,050 annually);
34% of Black children live in poor families while, in contrast, 10% of
White children. It is for these reasons that there has been extensive
research on the effect of poverty on children’s academic achievement.
While investigating the effect of childhood poverty on educational
achievement, Payne and Biddle (1999) examined the effect of school
funding and child poverty on mathematics achievement for a number of
school districts across the United States. Using data from the Second
International Mathematics Study and the School District Data Book, they found
that after controlling for a school district’s total annual per-student
school funding, percent of non-White students, and average level of cur-
ricular instruction, that measures of child poverty had a meaningful
effect on mathematics achievement.
In other work, the duration and timing of poverty has been a focus of
interest. Guo (1998) indicated that long-term poverty has substantial
influences on both ability and achievement, but the patterns of these
influences may differ by age. Childhood appears to be a much more cru-
cial period for the development of cognitive ability than early adoles-
cence. In contrast, poverty experienced in adolescence appears to have a
stronger influence on academic achievement than poverty experienced
earlier in life, a relevant finding for the current study on the SAT perfor-
mance of rising high school seniors.
In summarizing the work in the edited volume, Consequences of Growing
Up Poor, Duncan and Brooks-Gunn (1997b) concluded that not only does
poverty have its greatest impact during early and middle childhood, they
also suggested that parental income is a stronger correlate of children’s
academic ability and achievement than maternal education levels and
family structure. Given the reviewed research it is clear that income mat-
ters for childhood academic achievement, especially for the poor.
INCOME, POVERTY AND SAT PERFORMANCE
While there has been extensive research on income and academic
achievement, in general, there has been less research on the relationship
between family income and college admissions tests, and, in particular,
performance on the SAT. Using data from the National Education
Longitudinal Study of 1988, Owings, McMillen, and Burkett (1995) exam-
ined the social demographics of students who would make the cut-off
6
27918a_TCR_April2013_text_Layout 1 3/26/13 3:05 PM Page 6
TCR, 115, 040306 Influence of Family Income
criteria for selective colleges and universities (including total SAT scores
greater than 1100). They found that college-bound seniors from high
socio-economic status (SES) families were more likely to meet the cut-off
criteria than their middle and low SES contemporaries. One corollary of
the differences in the SAT distributions, students from middle SES fami-
lies were also more likely to meet the cut-off criteria than their low SES
counterparts.
Camara and Schmidt (1999) from the College Board investigated the
relationship between parental income and education on SAT perfor-
mance. Their results indicated that parental income and education bear
a strong relationship to performance on a variety of measures, with par-
ent education showing the stronger relationship. However, they argued
that: (1) parental income and education are related to most other pre-
dictors and outcomes of academic performance, such as high school GPA
and rank; and (2) Hispanic and African-American students from compa-
rable socioeconomic families scored lower than their Asian-American
and White peers.
In examining the relationship of SES and SAT performance, Zwick
(2004) also found, referencing descriptive statistics from other studies,
that SES influences a variety of academic outcomes and academic perfor-
mance indicators. The outcomes related to SES included the percentage
of students passing the mathematics and language arts tests for the
California High School Exit Exam, the percentage of students attaining basic
versus proficient levels on the National Assessment of Educational Progress
mathematics assessment, the percentage of students that meet the selec-
tion criteria for admissions into selective universities, high school grade
point average, and the likelihood of children to be read to at home by a
family member. Thus, Zwick argued that the SAT is a “wealth test” only in
the sense that every other measure of educational achievement is a
wealth test. Furthermore, she suggested that in order to improve diversity
on college campuses, additional indicators of diversity ought to be incor-
porated explicitly in admissions policies.
In an article published in The Journal of Blacks in Higher Education
(1998), it was argued that family income differences do not explain the
differences in total SAT scores by race/ethnicity. The article, however,
does report that Black students from families with incomes between
$80,000 and $100,000 perform 141 points lower on the SAT than their
White counterparts. Moreover, the Journal authors claimed that Black
students from families with incomes between $80,000 and $100,000 in
fact scored lower on the SAT than did White students from families with
incomes of less than $10,000. Further, it was argued that Black and White
families with similar incomes tend to have very different social and
7
27918a_TCR_April2013_text_Layout 1 3/26/13 3:05 PM Page 7
Teachers College Record, 115, 040306 (2013)
educational characteristics and experiences. Thus, it was asserted that
there is much more to the Black-White SAT performance gap than what
is captured by indices of a family’s economic background.
In a study using SAT data collected in 1995, Everson and Millsap
(2004) examined both individual and school-level effects on the achieve-
ment gap using multilevel latent variable modeling techniques. Their
results indicated that after accounting for the school-level effects, the
SAT performance gaps were reduced on average by a half of a standard
deviation—approximately 50 points on the SAT. For instance, the gap
between Asian American and African-American males on the SAT
Mathematics Reasoning tests was reduced by 56 points. These findings
suggest dramatic differences in the distribution of resources between
schools, which are also closely tied to a family’s income and socio-eco-
nomic status (Massey & Denton, 1993), and their effects on SAT score dif-
ferences.
The existing research on income and SAT performance has shed light
on the importance and limitations of family income in accounting for
SAT performance differences. However, these studies have not ade-
quately accounted for the possible non-linear relationship between fam-
ily income and achievement (Dixon-Román, 2007; Conley, 1999;
Mazumder, 2005). More importantly, for the most part, prior research
has not examined the differential effects of income on academic achieve-
ment by race/ethnicity. Thus, there remains a need for closer investiga-
tions of the effects of poverty on SAT performance, and the differences
between poor Black and White test-takers. Although we know poverty is
related to other measures of academic achievement, we are much less
certain about the relative effects of poverty on SAT performance.
In this study we set out to develop an appropriate and well-specified
model of the relationships among and between students’ academic
achievements, parental education, and family income, and their unique
and joint influences on the SAT scores of both Black and White students,
with a focus on students from low-income families.
METHOD
The analytic models we present in this paper examined the influences of
parental education, family income, and academic achievement on Black
and White high school students’ verbal and mathematics SAT scores. The
explanatory models were fit and tested for model invariance between
Black and White college-bound students. Below, we describe our sample
and data sources, as well as our model-based analytic approach.
8
27918a_TCR_April2013_text_Layout 1 3/26/13 3:05 PM Page 8
TCR, 115, 040306 Influence of Family Income
PARTICIPANTS
Participants for this study come from the population of the 2003 cohort
of college-bound seniors who took the SAT during their junior or senior
year of high school, and who graduated from high school in 2003 (N=
1,417,374). In general, this cohort represents about 45% of all the high
school seniors in the United States. Females make up about 54% of this
group, and the cohort is largely White (69%), with 11% Black, 8% Asian
American, 4% Mexican American, 4% other Latinas/os, 1% Native
American, and 3% who marked “other” when noting their race or ethnic-
ity. Our analyses, however, focus on the Black (n= 121,722) and White (n
= 659,715) students in the 2003 cohort of the SAT. We focus on White
and Black examinees for interpretational simplicity in this study and
because these are the two groups often explored in the test score gap lit-
erature. Of this sample 59.1% of the Black test-takers were females and
54.5% of the White sample were females. This sample has slightly more
females than in contrast to the US population 15 to 19 years of age in
2003, where Black females were 51% and White females were 49% (U.S.
Census Bureau, 2003). Given that the study sample is the total population
of Black and White test-takers in the 2003 cohort sample weights were not
used nor were they appropriate.
While we acknowledge the ethnic variation within both Black and
White racial categories we were constrained by the racial/ethnic cate-
gories of the Student Descriptive Questionnaire (described next). Thus, we
do not refer to the self-identified Black test-takers as African American
because they may also be West Indian, African, or of Latina/o origin.
Similarly, we do not refer to self-identified White test-takers as European
American because they may also be from an Eastern or Western
European countries, Jewish, Canadian, African, or of Latina/o origin.
DATA SOURCE
When students register with the College Board to sit for the SAT they
complete a lengthy questionnaire called the Student Descriptive
Questionnaire (SDQ; see www.collegeboard.org). The SDQ includes ques-
tions on students’ high school courses, class rank, parental education,
family income, and their race or ethnicity. Responses to these questions
provided the data for this study. Appendix A provides further detail about
the SDQ questions and how they were scored.
In addition, the SDQ asks students to indicate their average grade in
specific subject areas of high school courses and to report their cumula-
tive grade point average (GPA) on a scale of A+ to F. The score range of
9
27918a_TCR_April2013_text_Layout 1 3/26/13 3:05 PM Page 9
Teachers College Record, 115, 040306 (2013)
A+ to F is placed on a 4.0 grade point average scale with a 4.3 for A+.
Students report their most recent class rank on a scale from lowest fifth
to highest tenth. The SDQ also asks students to report their parent(s)
level of education from grade school to graduate or professional degree.
Parental levels of education were scaled in actual years of education. That
is, a high school diploma or equivalent was scored a 12 and a bachelor’s
degree was scored a 16. Students also reported their best estimates of
annual family income, with reporting categories ranging from a mini-
mum of less than $10,000 to a maximum of $100,000 or more per year.
Lastly, each student’s SAT verbal reasoning and mathematics reasoning
scores are included separately in the data set of these 2003 cohort of col-
lege-bound seniors. SAT scores are reported on a 200 to 800) scale. These
background variables and SAT scores, 13 in all, were used in our models.
A MODEL-BASED APPROACH
In the absence of random assignment, we do not engage in causal infer-
ence but rather the estimation of the partial direct and indirect associa-
tions of family income and poverty on SAT performance. Thus, we
employed structural equation modeling (SEM) with Mplus, a latent vari-
able modeling software package (Muthén & Muthén, 1998-2010), in
order to model the hypothesized associations. This approach is particu-
larly well suited for our study because of the unobserved latent variables
of the model (i.e., high school achievement and SAT performance) that
are measured by a large number, 10 in all, of observed variables and the
hypothesized direct and indirect effects of family income on SAT perfor-
mance (see Figure 2). In addition, the estimation of the partial associa-
tion of family income with performance on the SAT accounting for other
relevant individual differences such as parental education and high
school achievement can be fit simultaneously and with much more flexi-
bility and power within the SEM framework.
SEM analyses often include three broad stages: specifying the model
that relates the variables one to another; estimating the parameters of the
model; and, finally, estimating how well the model fits the empirical data,
i.e., how well the theoretical model replicates the empirical correlations
between and among the variables included in the model. In addition, we
also conducted a multiple group analysis (Horn & McArdle, 1980, 1992)
in order to test the model invariance between Black and White students
in the 2003 cohort of college-bound seniors. Multiple group analysis in
SEM is a simultaneous and flexible way of testing for group invariance in
model parameters, which is analogous to evaluating for interaction
effects. It is via the multiple group analysis that we evaluate to what extent
10
27918a_TCR_April2013_text_Layout 1 3/26/13 3:05 PM Page 10
TCR, 115, 040306 Influence of Family Income
there is a differential association between family income and SAT perfor-
mance by race.
Given that it is well known in the sociological and economics literature
that income has a non-linear effect on social and behavioral outcomes
(Becker & Tomes, 1979, 1986; Dixon-Román, 2007; Conley, 1999;
Mazumder, 2005) family income was converted into a dummy variable in
order to better estimate the non-linear association of family income with
SAT performance. The hypothesized measurement model and structural
model for this study are depicted in Figures 1 and 2 below.
Using SEM, we explored the direct associations of students’ high
school achievement, and their parents’ education levels and family
income on SAT verbal and mathematical reasoning test scores (see
Figure 2). This model also evaluates the indirect associations of family
income with SAT performance via both parents’ education and the test-
takers high school achievement. Based on the existing literature, this
study hypothesizes that higher levels of test-takers’ high school achieve-
ment and both parents’ education and family income are associated with
higher levels of SAT performance. Similarly, the structural model hypoth-
esizes that higher levels of family income and both parents’ education are
associated with higher levels of high school achievement.
11
Figure 1. The Measurement Model of SAT Performance and High School Achievement
!
!
!
!
!
!
!
!
!
!
!
!
27918a_TCR_April2013_text_Layout 1 3/26/13 3:05 PM Page 11
Teachers College Record, 115, 040306 (2013)
Given the degree of missing data in our sample, we employed full infor-
mation maximum likelihood (FIML) estimation procedures with data
missing at random to account for the incomplete cases on each of the
measured variables (Muthén, Kaplan, & Hollis, 1987). Simulation studies
have demonstrated that FIML performs well with at least 30% covariance
coverage (Peng, Harwell, Liou, & Ehman, 2006. This assumes data are
missing at random (MAR). MAR does not assume that data are missing
completely at random (which is rarely ever the case) but that the proba-
bility of Y being missing does not depend on the missing value itself, but
does depend on observed values of Y or other completely observed vari-
ables (Xs).
RESULTS
Descriptive statistics for each of the analysis variables across both the
Black and White student cohorts are presented in Table 1, below, in
order to characterize the sample.
12
Figure 2. The Structural Model of SAT Performance and High School Achievement
!
!
!
!
!
!
!
!
!
!
!
!
27918a_TCR_April2013_text_Layout 1 3/26/13 3:05 PM Page 12
TCR, 115, 040306 Influence of Family Income
The magnitude of the Black/White differences among the measured
variables is evident in Table 1. The White students in our sample are
higher on all measured variables. Levels of parental education, for exam-
ple, among Whites were reported, on average, to be at the college-level,
while Blacks reported their parents’ education at the high school level.
Measures of academic achievement, here measured as grade point aver-
age (GPA)—both overall, class rank, and within subject areas—were also
different for Black and White students. White students, on average,
reported a higher overall GPA and class rank. There were also differences
in the degree of missing data on each of the variables, particularly for
family income.
Perhaps the greatest differences between the Black and White students
13
Table 1. Descriptive Statistics and Percent Missing for All Measured Variables by Black and White Test-takers
Black Test-takers White Test-takers
Percent Percent
Mean (SD) Missing Mean (SD) Missing
Father Education a* HS Diploma/Equivalent 17.2 AA Degree 13
( HS Diploma – AA Degree) (HS Diploma – Some Graduate)
Mother Education a* Some College 10 AA Degree
(HS Diploma – BA Degree) HS Diploma – BA Degree) 12
Family Income b* $30K to $35K 23.2 $60K to $70K 31.9
( $15K-$20K – $ 50K-$60) ($40K-$50K – $80K-$100K )
Overall HS GPA * 2.97 (0.66) 3.8 3.46 (0.68) 4.7
Class Rank a* Second Fifth 30.9 Second Tenth 34.9
(Middle Fifth – Second Tenth) (Second Fifth – Highest Tenth)
GPA English * 3.00 (0.70) 6.3 3.36 (0.65) 9.5
GPA Foreign Languages * 2.94 (0.87) 11.5 3.29 (0.77) 12.8
GPA Arts & Music * 3.56 (0.67) 23.4 3.80 (0.46) 22.3
GPA Mathematics * 2.71 (0.83) 6.9 3.19 (0.76) 10.1
GPA Natural Sciences * 2.88 (0.75) 9.1 3.31 (0.68) 10.8
GPA History * 3.05 (0.74) 7.6 3.45 (0.65) 10.2
SAT Verbal * 430 (99) 0 529 (100) 0
SAT Math * 424 (98) 0 533 (103) 0
Nof Students 121,722 659,715
aGiven that these are ordinal categorical variables, the estimates presented are modes and interquartile ranges
are presented in the parentheses.
bThese are the median category estimates of family income and interquartile ranges are presented in the paren-
theses.
* Indicates statistically significant mean differences between Black and White test-takers.
27918a_TCR_April2013_text_Layout 1 3/26/13 3:05 PM Page 13
Teachers College Record, 115, 040306 (2013)
were found in reported family income and SAT scores. Black students,
for example, reported annual family income levels about half that of
their White counterparts, approximately a $30,000 to $35,000 difference.
In addition, their Mathematics and Verbal SAT scores are about 100
points lower (roughly one standard deviation) than White students.
Recall, the central focus of our analysis is the influence of family income
on SAT scores. To set the context, Table 2, below, presents the average
SAT scores by family income levels for the entire 2003 cohort of SAT col-
lege-bound seniors and for Black and White families separately.
We clearly see in Table 2 that SAT scores increase monotonically as fam-
ily income levels increase. Moreover, the correlation between family
income and SAT Math and SAT Verbal indicates a moderate positive rela-
tionship. It is this relationship, unconditioned by high school achieve-
ment and parental education levels, which leads many to infer a strong
causal relationship between family income and performance on the SAT.
It is important to note that the amount of missing data on family income
(30.6%), as well as the truncated distribution of income at greater than
$100,000, constrain our analyses. Nevertheless, we attempt to account for
these limitations in the structural equation modeling approaches
described below.
As a preliminary exploratory analysis, we used principal components of
the SAT and high school achievement variables in separate group ordinary
14
Table 2. Mean SAT Mathematics and Verbal Scores by Family Income for the 2003 College Bound Cohort
Black Test-takers White Test-takers Total Sample
Math Verbal Math Verbal Math Verbal
Family Income Score Score Score Score Score Score
Less than $10,000 382 381 478 480 418 418
$10,000 to $15,000 395 398 478 481 435 439
$15,000 to $20,000 400 405 485 488 446 450
$20,000 to $25,000 409 413 493 495 460 463
$25,000 to $30,000 411 419 495 497 466 470
$30,000 to $35,000 419 426 502 504 479 482
$35,000 to $40,000 422 430 504 505 484 487
$40,000 to $50,000 431 438 510 510 496 498
$50,000 to $60,000 441 450 516 514 507 506
$60,000 to $70,000 440 450 521 519 512 512
$70,000 to $80,000 448 457 528 524 521 518
$80,000 to $100,000 461 468 539 534 533 529
More than $100,000 490 495 568 557 564 555
No Response
27918a_TCR_April2013_text_Layout 1 3/26/13 3:05 PM Page 14
TCR, 115, 040306 Influence of Family Income
least squares regressions. Although not reported here these analyses
revealed that not only did family income have a meaningful direct associ-
ation with the SAT, but that its association was non-linear and substan-
tially higher for Black test-takers than for White test-takers, particularly at
lower levels of family income. In addition, the estimated regression coef-
ficients were comparable to the structural estimates reported below in
the multiple group structural equation model. For instance, the unstan-
dardized regression coefficient for the high school achievement princi-
ple component for Black test-takers was 0.53 (SE 0.00) and in the multiple
group structural equation model the standardized direct effect was 0.57
(SE 0.00). While these results are comparable to the below structural
equation model results, SEM enabled us to examine and evaluate the fit
of the larger conceptual model with the analysis of both direct and indi-
rect effects of family income simultaneously. Thus, we turn to the struc-
tural equation modeling.
The fit statistics for the separate group and multiple group analyses
(MGA) of the measurement model (e.g., confirmatory factor analysis (or
CFA) model) and the structural model were evaluated. The estimates of
the root mean square error of approximation (RMSEA, a, the compara-
tive fit index (CFI), and the Tucker-Lewis index (TLI) indicate the close-
ness of model fit to the empirical data. An RMSEA of 0.05 or less and a
CFI or TLI of 0.95 or higher indicate a close fit and plausible model
(Bollen, 1989; Kline, 2005; Loehlin, 2004; Raykov & Marcoulides, 2006).
The separate group analyses began by fitting a series of measurement
models including the hypothesized and alternative models. Each of the
alternative measurement models (e.g., a single achievement factor; quan-
titative/science and language/humanities two-factor model; high school
subject area; high school global achievement; and SAT performance
three-factor model) had a poor fit to the data with the exception of the
three-factor model. Each of these alternative models were nested models
and did not add any additional observed variables to the model. Thus,
the chi-square differences test was used to evaluate the difference
between the models. While the chi-square difference test suggests that
the three-factor model is a closer fit to the data than the hypothesized
two-factor model of high school achievement and SAT performance, the
three-factor model for White test-takers had a positive definite covariance
matrix with a correlation of 1.013 between the high school subject area
and global achievement latent variables. The latent variable correlation
greater than one suggests a high degree of redundancy between the two
measures, which indicates that they are likely measuring the same con-
struct. Thus, we proceeded with the hypothesized two-factor measure-
ment model.
15
27918a_TCR_April2013_text_Layout 1 3/26/13 3:05 PM Page 15
Teachers College Record, 115, 040306 (2013)
Separate group analyses were also conducted with the structural model
fitting both the hypothesized model and an alternative model. The alter-
native structural model included a continuous observed measure of fam-
ily income assuming a linear relationship between family income and
SAT performance. Although the alternative model was found to have a
closer fit based on the chi-square difference test, the RMSEA, which
accounts for sample size and model complexity, suggests a closer fit of the
hypothesized model (0.04) over the alternative model (0.06). In addi-
tion, given that the model with the income dummy variable adds more
observables, model parameters, and model complexity, we also evaluated
the fit based on the Akaike Information Criteria (AIC). The AIC
accounts for model complexity based on the model degrees of freedom
and an AIC reduction of 10 units or more is considered to be a more
plausible model. The model with the income dummy variable reduced
the AIC 2,928,301 units for the model of White test-takers and 360,251
for Black test-takers. These AIC reductions further indicate that the
model with the dummy variable of family income is closer fit to the data
and more plausible. Moreover, as discussed earlier and demonstrated
below, the assumed linear relationship of the alternative model underes-
timates the family income effect on SAT performance. Thus, we
employed the hypothesized model for the multiple group analyses.
The MGA began with a fully constrained model then we allowed specific
model parameters to be freely estimated, and tested for meaningful group
differences in those specific model parameters. The fit indices for the
MGA indicate that there was a meaningful difference between Black and
White students for the covariance estimate between high school achieve-
ment and SAT performance in the measurement model. There were also
meaningful differences in the regression parameters between family
income and SAT and high school achievement and SAT. Therefore, the
final MGA model with the unconstrained regression estimates between
family income and SAT performance and high school achievement and
SAT performance was the final model reported in this paper.
The measurement model shown in Figure 1 posits both SAT-Verbal and
SAT-Mathematics as measurement indicators of a latent variable of cogni-
tive ability, which we are calling SAT performance. It also includes mea-
sures of GPA, class rank, and GPA for history, science, math, English,
art/music, and foreign language and each are posited, collectively, as
indicators of high school achievement. The SAT latent variable is scaled
to the SAT-Mathematics scale, which ranges from 200 to 800, and the
high school achievement latent variable is scaled to the standard four-
point scaling for GPA (with a maximum of 4.3 points for an A+) (see
Table 3). While it is theoretically problematic to model SAT performance
16
27918a_TCR_April2013_text_Layout 1 3/26/13 3:05 PM Page 16
TCR, 115, 040306 Influence of Family Income
as a total score rather than each subtest, we decided to model what is gen-
erally evaluated in practice by college admissions—SAT total scores.
The final model was fit, indicating the plausibility of the conceptual
model, using full information maximum likelihood (FIML) estimation
with data missing at random. In order to evaluate the model perfor-
mance of incomplete case analysis with FIML we fit the same model using
listwise deletion and found the same model fit and very minor variation
in the parameter estimates. As mentioned earlier, there is a meaningful
difference in the measurement model between the Black and White stu-
dents: The covariance between SAT performance and high school
achievement is greater for White students than for Black students.
17
Table 3. The Unstandardized Measurement Model Estimates (Multiple Group Incomplete Case Analysis)
Latent Variables Black Test-takers White Test-takers
SAT Performance BY
SAT Math 1.00 (0.0)* 1.00 (0.0)*
SAT Verbal 0.92 (0.0)* 0.92 (0.0)*
HS Achievement BY
Overall HS GPA 1.00 (0.0)* 1.00 (0.0)*
Class Rank 1.67 (0.0)* 1.67 (0.0)*
GPA English 0.85 (0.0)* 0.85 (0.0)*
GPA Foreign Languages 0.96 (0.0)* 0.96 (0.0)*
GPA Arts & Music 0.36 (0.0)* 0.36 (0.0)*
GPA Mathematics 0.95 (0.0)* 0.95 (0.0)*
GPA Natural Sciences 0.90 (0.0)* 0.90 (0.0)*
GPA History 0.82 (0.0)* 0.82 (0.0)*
Measurement Error
SAT Math 0.23 (0.0)* 0.26 (0.0)*
SAT Verbal 0.33 (0.0)* 0.32 (0.0)*
Overall HS GPA 0.08 (0.0)* 0.04 (0.0)*
Class Rank 0.78 (0.0)* 0.52 (0.0)*
GPA English 0.26 (0.0)* 0.20 (0.0)*
GPA Foreign Languages 0.46 (0.0)* 0.31 (0.0)*
GPA Arts & Music 0.38 (0.0)* 0.18 (0.0)*
GPA Mathematics 0.40 (0.0)* 0.28 (0.0)*
GPA Natural Sciences 0.30 (0.0)* 0.21 (0.0)*
GPA History 0.31 (0.0)* 0.21 (0.0)*
* p < .05
27918a_TCR_April2013_text_Layout 1 3/26/13 3:05 PM Page 17
Teachers College Record, 115, 040306 (2013)
Moreover, there was a 218-point difference in total SAT performance
between Black and White students. This would be approximately the
summed difference in the observed data of SAT-V and SAT-M.
The structural model in Figure 2 posits various direct and indirect
effects of family income on SAT performance. As before, the model’s
structural or path coefficients were computed using maximum likelihood
estimation assuming data missing at random. The combining of income
categories was conducted due to homogeneity of categories in the bivari-
ate analyses. However, we also did not want to collapse too many cate-
gories as this would have taken away from the sensitivity of capturing the
non-linear relationship between family income and SAT performance. In
addition, the $80,000 to $100,000 category was the contrast group so it
does not appear in the model. Again, there are a number of differences
in the structural model between the Black and White students: (1) the
effect of income on SAT performance is meaningfully larger for the
Black test-takers, and (2) the effect of high school achievement on SAT
performance is meaningfully larger for the White test-takers in our sam-
ple. The effect of income on high school achievement was negligible for
both groups.
Table 4, below, presents the unstandardized structural estimates.
18
Table 4. The Unstandardized Structural Estimates (Multiple Group Incomplete Case Analysis)
Effects Black Test-takers White Test-takers
SAT ON
HS Achievement 138.40 (0.8)* 180.80 (0.40)*
Mother Education 8.20 (0.0)* 8.10 (0.00)*
Father Education 9.40 (0.0)* 9.40 (0.00)*
Family Income:
< $10K -92.20 (2.8)* -48.20 (2.0)*
$10k to $15K -70.20 (2.8)* -48.20 (1.8)*
$15K to $20K -64.80 (2.8)* -38.40 (1.8)*
$20K to $30K -48.00 (2.6)* -29.00 (1.2)*
$30K to $40K -37.40 (2.6)* -24.60 (1.0)*
$40K to $50K -27.20 (2.8)* -18.80 (1.0)*
$50K to $70K -16.40 (2.6)* -13.60 (0.8)*
$70K to $80K -10.80 (3.2)* -7.40 (1.0)*
> $100K 35.20 (3.2)* 29.00 (0.8)*
HS Achievement ON
Mother Education 0.02 (0.0)* 0.02 (0.0)*
Father Education 0.03 (0.0)* 0.03 (0.0)*
Family Income:
< $10K -0.04 (0.0)* -0.04 (0.0)*
27918a_TCR_April2013_text_Layout 1 3/26/13 3:05 PM Page 18
TCR, 115, 040306 Influence of Family Income
Figure 3, below, presents a plot of family income and the direct income
effects by race in order to observe the non-linear differential direct
income effect on SAT performance.
19
$10k to $15K -0.02 (0.0)* -0.02 (0.0)*
$15K to $20K -0.02 (0.0)* -0.02 (0.0)*
$20K to $30K -0.02 (0.0)* -0.02 (0.0)*
$30K to $40K 0.00 (0.0) 0.00 (0.0)
$40K to $50K 0.01 (0.0)* 0.01 (0.0)*
$50K to $70K 0.01 (0.0)* 0.01 (0.0)*
$70K to $80K 0.00 (0.0) 0.00 (0.0)
> $100K 0.00 (0.0) 0.00 (0.0)
Mother Education WITH
Father Education 3.35 (0.0)* 3.72 (0.0)*
Residual Variances
SAT 0.47 (0.0)* 0.41 (0.0)*
HS Achievement 0.33 (0.0)* 0.31 (0.0)*
* p < .05
Notes. Family income estimates are in contrast to a family income between $80,000 and $100,000.
Unstandardized estimates for the SAT endogenous variable were in total SAT score units.
Figure 3. Plot of Direct Income Effects by Race
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
27918a_TCR_April2013_text_Layout 1 3/26/13 3:05 PM Page 19
Teachers College Record, 115, 040306 (2013)
Figure 4 presents a plot of the observed and model-based estimates of
total SAT score differences by race for each category of income. This plot
shows how the estimated total SAT score differences ranged from 65.2
points for test-takers from family incomes greater than $100,000 to 115.4
points for test-takers from a family income less than $10,000, again indi-
cating the non-linear differential effect of income, the narrowing of SAT
performance differences as income increases, and the effect of poverty.
THE EFFECT OF POVERTY
In order to estimate the effect of poverty on SAT performance a separate
model was fit with a dummy variable for family poverty in contrast to fam-
ily income of $80,000 to $100,000. Given that the weighted average
poverty threshold for a four-person family in 2003 was $18,979 (DeNavas-
Walt, Proctor, & Mills, 2004), family poverty was measured by truncating
the first three income categories to measure family incomes less than
$20,000. This created a poverty category that ranged from $0 to $20,000
that was in contrast to the $80,000 to $100,000 income category. The
additional structural equation model fit in order to estimate the effect of
family poverty maintained closeness of fit as well as all of the same para-
meter estimates as the previous model with the exception of the parame-
ter estimates for the poverty dummy variable.
20
Figure 4. Plot of Obser ved and Model-Based Estimates of SAT Total Score Differences by Race
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
27918a_TCR_April2013_text_Layout 1 3/26/13 3:05 PM Page 20
TCR, 115, 040306 Influence of Family Income
The unstandardized effect of the dummy variable for family poverty
was -77 for Black students and -44.2 for White students. Independent of
the effects of both parents education and students high school academic
achievement, these unstandardized effects indicate that in contrast to stu-
dents of middle-income families, White students living in poverty per-
form 44.2 total SAT score points lower and Black students living in
poverty perform 77 total SAT score points lower. Moreover, the unstan-
dardized effect of family poverty for Black students is more than one half
of the unstandardized effect of academic achievement suggesting that a
77 total SAT score point difference continues to remain between those in
poverty and those from middle-income families even when there is a 1-
unit increase in high school academic achievement for Black students. As
discussed below, this dynamic is exacerbated under conditions of
extreme poverty.
In order to estimate the effect of extreme poverty we return to the mul-
tiple group structural equation model absent the poverty effect dummy
variable (i.e., Table 4). Table 5 below provides the direct, indirect, and
total effects of extreme poverty as measured by a family income less than
$10,000 on SAT performance. The direct effects have been taken directly
from the 2003 structural model. The indirect effects were computed by
the product of the effect of income on the mediating variable(s) and the
effect of the mediating variable(s) on SAT. The total effect was computed
by adding the direct effect to the indirect effect.
21
Table 5. Unstandardized Direct, Indirect, and Total Effects of Extreme Poverty (Family Income less than
$10,000) on SAT Scores
Income Effects Direct Indirect Total
Black White Black White Black White
Poverty on SAT -92.2 (2.8) -48.2 (2.0)
Via
HS Achievement -5.7 -7.4 -97.9 -55.6
Via
Father Education -1.3 -0.3 -99.2 -55.9
Via
Mother Education -1.4 -0.2 -100.6 -56.1
Via
Father Education & HS Achievement -0.6 -0.2 -101.1 -56.2
Via
Mother Education & HS Achievement -0.5 -0.1 -101.6 -56.3
Notes. In total SAT score units. The indirect and total effect estimates could not be calculated in Mplus
because the indirect estimation option requires each of the mediators to be endogenous. This model
specifies family income to co-vary with both parents education thus the standard errors were not esti-
mated for these indirect and total effects.
27918a_TCR_April2013_text_Layout 1 3/26/13 3:05 PM Page 21
Teachers College Record, 115, 040306 (2013)
As Table 4 indicates, the direct and total effect of extreme poverty for
Black students is almost twice that of their White counterpart. The
unstandardized effects of extreme poverty for White students (B = –48.2,
SE = 2.0) and Black students (B = –92.2, SE = 2.8) indicate that in contrast
to students of middle-income families, White students living in extreme
poverty perform 48.2 total SAT score points lower and Black students liv-
ing in extreme poverty perform 92.2 points lower. Moreover, the unstan-
dardized direct effect of extreme poverty for Black students (B = –92.2,
SE = 2.8) is more than two thirds of the unstandardized direct effect of
high school achievement (B = –138.6, SE = 0.8) suggesting that a 92.2
total SAT score point difference continues to remain between those in
poverty and those from middle-income families even when there is a 1-
unit increase in high school academic achievement for Black students.
These are non-negligible findings given that over 11% of the sample of
Black students that reported family income were in this category, which
was the modal income estimate for Black students.
Tables 4 and 5 illustrate the extent to which family income indepen-
dently effects the SAT performance of Black and White college-bound
students. As reported earlier, White students performed 218 points
higher, on average, than their Black counterparts. After conditioning
SAT performance on high school achievement and both parents’ educa-
tion and income the difference reduced to 71.4 points for students with
family incomes between $80,000 and $100,000 (and 65.2 points for stu-
dents with family incomes greater than $100,000). Moreover, the effect of
poverty was moderate to large and nearly twice as large for Black students
than for White students.
DISCUSSION & CONCLUSIONS
Using multiple group structural equation modeling, this study sought to
examine the independent direct and indirect associations of family
income and poverty on the SAT reasoning test scores for both Black and
White test-takers, while controlling for their academic achievements in
high school, and their parents’ reported educational achievement levels.
Overall, results indicate that family income and, in particular, extremely
low levels of family income (what we refer to as poverty) has a meaning-
ful contribution to the total SAT reasoning test scores for both Black and
White test-takers, and helps to explain the SAT performance differences
between the two social groups of students.
Family income, for example, was found to have a nonlinear, differen-
tial direct effect on total SAT performance for both Black and White stu-
dents. In fact, for some family income levels the effect was nearly twice as
22
27918a_TCR_April2013_text_Layout 1 3/26/13 3:05 PM Page 22
TCR, 115, 040306 Influence of Family Income
large for Black test-takers than it was for the White test-takers in our sam-
ple. The literature to date examining the effect of family income on SAT
performance has been limited to observational data. While there is no
way to introduce true random design elements to this problem, the
resulting models are providing estimates as of the direct and indirect
associations of family income partialing out the co-variation of test-taker
high school achievement and both parents’ education. Moreover, the
existing literature has assumed linear, universal effects of family income,
and as a result reports somewhat smaller effects of family income on SAT
performance (Bowen & Bok, 1998; Camara & Schmidt, 1999; Everson &
Millsap, 2004). This study, in contrast, indicates that the smaller income
effects found in previous research are a result of model mis-specification,
particularly at the lower levels of family income, resulting in underestima-
tions of the effects of family income on SAT performance, particularly
for Black test-takers.
Moreover, the results of this research also challenge findings from the
1999 College Board report Reaching the Top: A Report of the National Task
Force on Minority High Achievement (College Board, 1999). The College
Board’s Reaching the Top report found that Black students with college-
educated parents score lower on standardized tests than White students
whose parents did not graduate from high school. The results from the
models reported here, however, suggest that not only does parental edu-
cation have a small effect relative to high school achievement, but also
that parental income may serve as an equalizing factor (though not
bringing parity to Black and White students’ SAT performance). That is
to say, as the exogenous variables in the models increase, parental
income remains the only variable that appears to narrow the SAT score
differences between Black and White examinees. The meaningfully
larger effect of family income for Black test-takers, and the meaningful
reductions in score differences by social group in total SAT reasoning
performance suggest the relative and substantial influence of family
income for enabling social and educational opportunities. Social and
educational opportunities, therefore, appear to be substantially con-
strained for test-takers—both Black and White—living in poverty.
Indeed, the models described in this study indicate a large meaningful
effect of poverty, especially extreme poverty, on SAT performance for
both Black and White test-takers. In fact, the large and differential effect
of extreme poverty suggests that even if Black test-takers living in extreme
poverty were to boost their high school academic achievement by one
point (clearly a feat on a 4.0 scale!) they would, nevertheless, perform
25.2 points below their White middle-income counterparts who achieved
one point lower in high school achievement. Thus, the focus on high
23
27918a_TCR_April2013_text_Layout 1 3/26/13 3:05 PM Page 23
Teachers College Record, 115, 040306 (2013)
school achievement for students in poverty may not be sufficient for pro-
viding them with an equal opportunity to higher education based on the
measured proxy of total SAT scores.
While the effect of high school achievement may not alleviate the
effect of extreme poverty, it remained relatively larger than the effects of
higher income levels. The strongest influence in our model is overall
high school achievement, which was measured on the traditional 4.0
grade point average scale. For White students, the effect of high school
achievement suggests that for every unit increase in high school achieve-
ment (GPA) there is a 180-point increase on the total SAT scale. For
Blacks, the increase is approximately only 138 points on the total SAT
scale, suggesting, again, the importance of high school achievement for
increasing performance on total SAT reasoning scores. However, high
school achievement does not have as large an effect for Black test-takers
as it does for White test-takers. Thus, as high school achievement
increases for both Black and White test-takers the between-group differ-
ences in total SAT performance increases rather than decreases.
The differential effect for high school achievement, along with the dif-
ferential and large poverty effect, in part, suggests an effect of schooling
where Black test-takers, especially those living in poverty, are likely
attending poorer quality schools. This implication resonates with the
findings reported earlier by Everson and Millsap (2004) and Benners and
Everson (2009). In addition, the differential direct and indirect effects of
income on high school achievement and, in turn, high school achieve-
ment on SAT scores may also be explained in terms of residential racial
and economic segregation which are, through property values and tax
policies, related to the quality of schooling (see Berends & Peñaloza,
2010; Card & Rothstein, 2007; Everson and Millsap, 2004; Massey,
Condron, & Denton, 1987; Mickelson, 2006; and Wilson, 1987).
Moreover, the poverty effects reported here suggest the lack of social and
educational resources in the larger community that are needed to sup-
plement and complement the learning taking place in schools (Gordon,
Bridglall, & Meroe, 2005).
However, given their finding that most of the variation in student
achievement is within schools and not between schools, Konstantopoulos
and Hedges (2008) caution on the focus of school reform based on social
group parity in achievement as the benchmark. They suggest that inter-
preting the magnitude of the school reform effect on social group differ-
ences in standardized test performance will not only be disappointing,
but misleading. Relatedly, Dixon-Román (2010) and Gutierrez and
Dixon-Román (2011) argue that the “gaze” on achievement or test score
gaps are not just misleading but overlook the problematic inequity in
24
27918a_TCR_April2013_text_Layout 1 3/26/13 3:05 PM Page 24
TCR, 115, 040306 Influence of Family Income
social and pedagogical conditions. These cautionary arguments need to
be given serious consideration in any school reform effort strategy to
address differences in SAT performance.
Our models also suggest that parental education levels, though rela-
tively small, are also important influences on SAT scores, independent of
family income and high school achievement. These measures of mothers’
and fathers’ education were scaled by each matriculated year of educa-
tion toward a diploma or degree. The models for both the Black and
White examinees indicated that the joint direct effects of mothers’ and
fathers’ education are approximately 8 to 9 points on the total SAT scale.
Again, this estimate is roughly equivalent for both Black and White stu-
dents.
Moreover, the differential association of family income and high school
achievement with SAT performance by race is theoretically suggestive of
the continued effects of racism and discrimination in the United States.
Although the nation may be “post-intentional” in race relations (Perry,
2011) there are still the produced unintended consequences of what
Jackson (2008) describes as a sense of racial distrust and even paranoia.
This racial paranoia, fear, and social distrust produces unconscious and
unquestioned actions and responses to racial difference in social situa-
tions which, in the cumulative, enable differential “treatments” even
within the same classrooms or income levels. These differential treat-
ments are also related to the linguistic and cultural variation between and
within Black and White cultural communities. Inherently, the SAT makes
strong assumptions of linguistic and cultural universality at the symbolic
cost of all those that do not comport or perform to those aims (Freedle,
2003; Santelices & Wilson, 2010). The results of this study would suggest
that there is a social distribution to these racialized effects where those of
higher incomes are more familiar with the cultural capital assessed on
the SAT (Bourdieu, 1986).
While these theoretical comments are speculations that are not empir-
ically grounded with the current examination of the SAT the results of
this study do indicate how race and class are co-constitutive and inextri-
cably tied. This is an important finding that has both theoretical and pol-
icy implications. On the one hand, there has and continues to be an
agenda in sociological and educational research to empirically provide
support for the William Julius Wilson argument of class increasingly
trumping race (Wilson, 1980) despite his own re-positioning on this
debate (Wilson, 2009). This study suggests that for SAT performance
race and class inform and constitute each other. On the other hand,
there has been a peeling away of race-conscious policies such as affirma-
tive action over the past 30 years. The substantial interaction effect
25
27918a_TCR_April2013_text_Layout 1 3/26/13 3:05 PM Page 25
Teachers College Record, 115, 040306 (2013)
between race and income as well as race and high school achievement
indicates that these policy shifts have been empirically misguided, partic-
ular as it pertains to SAT performance. Both theoretical and policy impli-
cations of the race and income dynamic of this study point toward
important directions for future research in order to further understand
this dynamic with SAT performance.
While the results of this study appear robust, it is important to note the
limitations inherent in our data. For example, with the exception of SAT
scores, all other measures were derived from students’ self-reports.
Perhaps more importantly, family income was reported and measured on
an interval scale and was truncated at incomes greater than $100,000
(the upper categorical limit on the SDQ). As a consequence, we were
unable in our modeling efforts to account for the variation on SAT scores
for family incomes greater than $100,000. It is important to point out that
this is also a single-year self-report of family income, and therefore does
not account for the potential transitory shocks to family income that can
occur in volatile economic times. Lastly, the model does not account for
other potentially relevant variables such as parental occupational pres-
tige, family wealth, grandparents’ socioeconomic variables, access to test
preparation services, or variations in parenting practices.
Future research is needed to account for the limitations in our data
(e.g., with Asian Americans and Hispanic/Latinos), as well as to better
understand the racially differential effects of low levels of family income
on access to higher education. The results reported here also suggest the
need to further our understanding of how variations in school resources
influence performance on standardized test scores, and subsequent
access to postsecondary education. Moreover, it would be useful, we sus-
pect, to develop models that include other intervening variables, such as
test-prep or coaching, variations in parenting practices, as well as other
community and neighborhood resources that may serve to mediate the
effects of family income and poverty. With the increasing availability of
latent growth models (see McArdle, 2008), we are now positioned to
examine further the effects of family income and poverty on the growth
and change in total SAT reasoning scores over multiple test administra-
tions. The latent variable models described in this paper, we believe, may
help provide a better understanding of the direct and indirect effects of
family income, poverty, and high school achievement on access to higher
education.
26
27918a_TCR_April2013_text_Layout 1 3/26/13 3:05 PM Page 26
TCR, 115, 040306 Influence of Family Income
Acknowledgements
The authors thank the College Board for its support and for providing the data used in this study. In
addition, the first author is grateful to the U.S. Education Department’s Institute of Education
Sciences, and Northwestern University’s Institute for Policy Research for postdoctoral research support.
The authors also thank Greg Duncan and Larry Hedges for helpful comments on an earlier draft of
this paper. An earlier version of this paper was presented at the annual meeting of the American
Educational Research Association in Montreal, Canada. All correspondence should be sent to Ezekiel
Dixon-Román, Penn School of Social Policy & Practice, University of Pennsylvania, 3701 Locust
Walk, Philadelphia, PA 19104-6214, ezekield@sp2.upenn.edu.
Notes
1. In The Miner’s Canar y, Guinier and Torres (2002) stated, “Of course, this is an over-
simplification. The SAT and other ‘norm-referenced’ aptitude tests do tell us something
about one’s capacity to do analytic thinking. The problem is that such capacity is often
improved by practice; practice comes from coaching (which costs money), from experience
taking the test (which means exposure to the opportunity of learning from previous mis-
takes), and from other kinds of exposure to travel, books, and unusual words. Thus, while
the tests do tell us something about those who do well, they often tell us less about those
who do poorly; that is, they do not tell us what a poor performer is actually capable of
doing, only what that person has already learned or not learned to do” (p. 387).
References
Anyon, J. (1997). Ghetto schooling: A political economy of urban educational reform. New York, NY:
Teachers College Press.
Aughinbaugh, A., & Gittleman, M. (2003). Does money matter? A comparison of the effect
of income on child development in the United States and Great Britain. The Journal of
Human Resources, 38(2), 416–440.
Becker, G. S., & Tomes, N. (1979). An equilibrium theory of the distribution of income and
intergenerational mobility. Journal of Political Economy, 87, 1153–1189.
Becker, G. S., & Tomes, N. (1986). Human capital and the rise and fall of families. Journal
of Labor Economics, 4, 1–39.
Benners, G. A. & Everson, H. T. (April 2009). School effects on gender differences in learning
mathematics during high school: A multiple group multilevel latent growth analysis of
PSAT/NMSQT to SAT performance in mathematical reasoning. Paper presented at the annual
meeting of the National Council of Measurement in Education, San Diego, CA.
Blau, D. M. (1999). The effect of income on child development. The Review of Economics and
Statistics, 81, 261–276.
Berends M., & Peñaloza, R. (2010). Increasing racial isolation and test score gaps in math-
ematics: A 30-year perspective. Teachers College Record, 112(4), 978–1007.
Birch, H. G., & Gussow, J. D. (1970). Disadvantaged children: Health, nutrition, and school fail-
ure. New York, NY: Harcourt Brace Jovanovich and Grune & Stratton.
Blau, D. M. (1999). The effect of income on child development. Review of Economics and
Statistics, 81(2), 261–276.
Bollen, K. A. (1989). Structural equations with latent variables. New York, NY: Wiley.
27
27918a_TCR_April2013_text_Layout 1 3/26/13 3:05 PM Page 27
Teachers College Record, 115, 040306 (2013)
Bourdieu, P. (1986). The forms of capital. In J. G. Richardson (Ed.), Handbook of theor y and
research for the sociology of education. Westport, CT: Greenwood Press.
Bowen, W.G., and Bok, D. (1998). The shape of the river. Princeton, NJ: Princeton University
Press.
Bowles, S., Durlauf, S. N., & Hoff, K. (2006). Pover ty traps. Princeton NJ: Princeton
University Press.
Camara, W. J. & Schmidt, A. E. (1999). Group differences in standardized testing and social
stratification. College Board Report (99)5. New York, NY: College Board.
Card, D., & Rothstein, J. (2007). Racial segregation and the black-white test score gap.
Journal of Public Economics, 91(11-12), 2158–2184.
College Board. (2003). 2003 college bound seniors. New York, NY: College Board.
College Board. (1999). Reaching the top: A repor t of the national task force on minority high
achievement. New York, NY: College Board.
Conley, D. (1999). Being Black, living in the red: Race, wealth, and social policy in America.
Berkeley CA: University of California Press.
Crouse, J., & Trusheim, D. (1988). The case against the SAT. Chicago, IL: University of
Chicago Press.
Datcher-Loury, Linda. (1989). “Family Background and School Achievement Among Low
Income Blacks.” The Journal of Human Resources 24: 528–544.
DeNavas-Walt, C., Proctor, B. D., & Mills, R. J. (2004). Income, poverty, and health insurance cov-
erage in the United States: 2003. U.S. Census Bureau Current Population Reports, P60-226.
Washington, DC: U.S. Government Printing Office.
Dixon-Román, E. (2007). The mystery of inequity: Modeling the influence of intergenerational access
on intellective competence. (Unpublished doctoral dissertation). Fordham University: New
York, NY.
Dixon-Román, E. (2010). Inheritance and an economy of difference: The importance of
supplementary education. in L. Lin, E.W. Gordon, & H. Varenne (Eds.), Educating com-
prehensively: Varieties of educational experiences. Lewiston, NY: The Edwin Mellen Press.
Dooley, M., & Stewart, J. (2004). Family income and child outcomes in Canada. The
Canadian Journal of Economics, 37(4), 898–917.
Duncan, G. J., & Brooks-Gunn, J. (1997a). Consequences of growing up poor. New York: Russell
Sage Foundation.
Duncan, G. J., & Brooks-Gunn, J. (1997b). Income effects across the life span: Integration
and interpretation. In G. J. Duncan & J. Brooks-Gunn (Eds.), Consequences of growing up
poor. New York: Russell Sage Foundation.
Duncan, G. J., Huston, A. C., & Weisner, T. S. (2007). Higher ground: New hope for the working
poor and their children. New York: Russell Sage Foundation.
Elert, G. (2006. The SAT: Aptitude or demographics. Retrieved December 29, 2009 from
http://hypertextbook.com/eworld/sat.shtml
Everson, H., & Millsap, R. (2004). Beyond individual differences: Exploring school effects
on SAT scores. Educational Psychologist, 39(3), 157–172.
Everson, H. T., & Millsap, R. (2005). Everyone gains: Extracurricular activities in high
school and higher SAT scores. In E. W. Gordon, A.E. Meroe, & B. Bridgall (Eds.),
Supplementary education: The hidden curriculum of high academic achievement. Lanham, MD:
Rowman & Littlefield Publishers.
Everson, H., Millsap, R., & Diones, R. (April, 1995). Exploring group differences in SAT perfor-
mance using structural equation modeling with latent means. Paper presented at the annual
meeting of the American Educational Research Association, San Francisco, CA.
Fass, S., & Cauthen, N. K. (2008). Who are America’s poor children? The official stor y. New York:
National Center for Children in Poverty, Mailman School of Public Health, Columbia
University.
28
27918a_TCR_April2013_text_Layout 1 3/26/13 3:05 PM Page 28
TCR, 115, 040306 Influence of Family Income
Freedle, R. (2003). Correcting the SAT’s ethnic and social-class bias: A method for re-esti-
mating SAT scores. Harvard Educational Review,73(1), 1–43.
Geiser S., & Studley, R. (2001). UC and the SAT: Predictive validity and differential impact of the
SAT I and SAT II at the University of California. Retrieved March 1, 2009 from the University
of California Office of the President, http://www.ucop.edu/sas/research/researchand-
planning/pdf/sat_study.pdf
Gordon, E. W., Bridglall, B. L., & Meroe, A. S. (2005). Supplementary education: The hidden
curriculum of high academic achievement. Lanham, MD: Rowman & Littlefield Publishers.
Guinier, L. & Torres, G. (2002). The miner’s canary: Enlisting race, resisting power, transforming
democracy. Cambridge, MA: Harvard University Press.
Guo, G. (1998). The timing of the influences of cumulative poverty on children’s cognitive
ability and achievement, Social Forces, 77(1), 257–287.
Gutiérrez, R., & Dixon-Román, E. (2011). Beyond gap gazing: How can thinking about edu-
cation comprehensively help us (re)envision mathematics Education? In B. Atweh, M.
Graven, W. Secada, and P. Valero (Eds.), Mapping equity and quality in mathematics educa-
tion. New York, NY: Springer.
Hedges, L. & Nowell, A. (1998). Black-White test score convergence since 1965. In C. Jencks
& M. Phillips (Eds.), The Black-White test score gap. Washington, DC: Brookings Institution
Press.
Hedges, L. & Nowell, A. (1999). Changes in the Black-White gap in achievement test scores:
The evidence from nationally representative samples. Sociology of Education, 7, 111–135.
Horn, J. L. & McArdle, J. J. (1980). Perspectives on Mathematical and Statistical Model
Building (MASMOB) in research on aging. In L. Poon (Ed.), Aging in the 1980s:
Psychological issues (pp. 503-541). Washington, DC: American Psychological Association.
Horn, J. L. & McArdle, J. J. (1992). A practical guide to measurement invariance in aging
research. Experimental Aging Research, 18(3), 117–144.
Jackson, J. L. (2008). Racial paranoia: The unintended consequences of political correctness. New
York, NY: Basic Books.
JBHE Foundation. (1998). Why family income differences don’t explain the racial gap in
SAT scores. The Journal of Blacks in Higher Education, 20, 6–8.
Jencks, C. & Phillips, M. (1998). The Black-White test score gap. Washington, DC: Brookings
Institution Press.
Kline, R. B. (2005). Principles and practice of structural equation modeling (2nd ed.). New York,
NY: The Guilford Press.
Konstantopoulos, S., & Hedges, L. V. (2008). How large an effect can we expect from school
reforms? Teachers College Record, 110(8), 1613–1640.
Lareau, A. (2003). Unequal Childhoods: Class, Race, and Family Life. Berkeley, CA: University
of California Press.
Lee, J., & Wong, K. K. (2004). The impact of accountability on racial and socioeconomic
equity: Considering both school resources and achievement outcomes. American
Educational Research Journal, 41(4), 797–832.
Loehlin, J. C. (2004). Latent variable models: An introduction to factor, path, and structural equa-
tion analysis (4th ed.). Mahwah, NJ: Lawrence Erlbaum Associates.
Marsh, H. W., & Kleitman, S. (2002). Extracurricular school activities: The good, the bad,
and the nonlinear. Harvard Educational Review, 72(4), 464–514.
Massey, D. S., Condron, G. A., & Denton, N. A. (1987). The effects of residential segrega-
tion on black social and economic well-being. Social Forces, 66, 29–57.
Masset, D. S., & Denton, N. A. (1993). American apar theid: segregation and the making of the
underclass. Cambridge, MA: Harvard University Press.
29
27918a_TCR_April2013_text_Layout 1 3/26/13 3:05 PM Page 29
Teachers College Record, 115, 040306 (2013)
Mayer, S. (1997). What money can’t buy: Family income and children’s life chances. Cambridge,
MA: Harvard University Press.
Mazumder, B. (2005). The apple falls even closer to the tree than we thought: New and
revised estimates of the intergenerational inheritance of earnings. In S. Bowles, H.
Gintis, and M. O. Groves (Eds.), Unequal chances: Family background and economic success.
Princeton, NJ: Princeton University Press.
McArdle, J. J. (2008). Latent variable modeling of longitudinal data. Annual Review of
Psychology, 60, 577–605.
Mickelson, R. A. (2006). Segregation and the SAT. Ohio State Law Journal, 67, 157–199.
Miller, L. S. (1997). An American imperative: Accelerating minority educational advancement. New
Haven, CT: Yale University Press.
Muthén, B., Kaplan, D., & Hollis, M. (1987). On structural equation modeling with data
that are not missing completely at random. Psychometrika, 51, 431–462.
Muthén, L. K., & Muthén, B. O. (1998–2010). Mplus User’s Guide (6th ed.). Los Angeles, CA:
Muthén & Muthén.
National Association for College Admissions Counseling. (2008). Report of the commission on
the use of standardized tests in undergraduate admission. Arlington, VA: National Association
for College Admission Counseling.
Orr, A. J. (2003). Black-white differences in achievement: The importance of wealth.
Sociology of Education, 76(4), 281–304.
Owings, J., McMillen, M., & Burkett, J. (1995). Making the cut: Who meets highly selective college
entrance criteria? (pp. 3–4). Washington, DC: National Center for Educational Statistics.
Payne, K. J., & Biddle, B. J. (1999). Poor school funding, child poverty, and mathematics
achievement, Educational Researcher, 28(6), 4–13.
Peng, C.-Y. J., Harwell, M., Liou, S.-M., & Ehman, L. H. (2006). Advances in missing data
methods and implications for educational research. In S. Sawilowsky (Ed.), Real data
analysis (pp. 31–78). Greenwich, CT: Information Age.
Perry, I. (2011). More Beautiful and more terrible: The embrace and transcendence of racial inequal-
ity in the United States. New York, NY: NYU Press
Phillips, M., Brooks-Gunn, J., Duncan, G. J., Klebanov, P., & Crane, J. (1998). Family back-
ground, parent practices, and the black-white test score gap. In C. Jencks & M. Phillips
(Eds.), The black-white test score gap (pp. 103-145). Washington, DC: Brookings Institution
Press.
Raykov, T., & Marcoulides, G. A. (2006). A first course in structural equation modeling (2nd ed.).
Mahwah, NJ: Lawrence Erlbaum Associates.
Rothstein, R. (2004). Class and schools: Using social, economic, and educational reform to close the
Black-White achievement gap. Washington, DC: Economic Policy Institute.
Santelices, M. V., & Wilson, M. (2010). Unfair treatment? The case of Freedle, the SAT, and
the Standardization Approach to Differential Item Functioning. Harvard Educational
Review 80(1), 106–133.
Sexton, P.C. (1961). Education and income. New York, NY: Viking Press.
Sirin, S. R. (2005). Socioeconomic status and academic achievement: A meta-analytic review
of research. Review of Educational Research, 75(3), 417–453.
U.S. Census Bureau. (2001). Survey of Income and Program Participation Users’ Guide (3rd ed.).
Retrieved December 14, 2009 from the Demographics Survey Division, Survey of Income
and Program Participation branch website: http://www.census.gov/sipp/usrguide/
sipp2001.pdf
30
27918a_TCR_April2013_text_Layout 1 3/26/13 3:05 PM Page 30
TCR, 115, 040306 Influence of Family Income
U.S. Census Bureau. (2003). The Black Alone or in Combination Population in the United States:
2003. Retrieved March 26, 2010 from the Current Population Sur vey of the US Census
Bureau: http://www.census.gov/population/www/socdemo/race/ppl-185_aoic.html
White, K. (1982). The relation between socioeconomic status and academic achievement.
Psychological Bulletin, 91(3), 461–481.
White, S. B., Reynolds, P. D., Thomas, M. M., & Gitzlaff, N. J. (1993). Socioeconomic status
and achievement revisited. Urban Education, 28, 328–343.
Wightman, L. F. (1995). Legal education at the close of the twentieth century: Descriptions and
analyses of students, financing and professional expectations and attitudes. Newtown, PA: Law
School Admission Council.
Wilson, W. J. (1980). The declining significance of race: Blacks and changing American institutions.
Chicago, IL: University of Chicago Press.
Wilson, W. J. (1987). The truly disadvantaged: The inner city, the underclass, and public policy.
Chicago, IL: University of Chicago Press.
Wilson, W. J. (2009). More than just race: being Black and poor in the inner city. New York, NY:
W.W. Norton & Co.
Yeung, W. J., & Conley, D. (2008). Black-White achievement gap and family wealth. Child
Development, 79(2), 303–324.
Zwick, R. (2004). Is the SAT a “Wealth Test”? The link between educational achievement
and socioeconomic status (pp. 203–216). In R. Zwick, (Ed.). Rethinking the SAT in univer-
sity admissions. New York, NY: Routledge Falmer.
Appendix A
THE SAT QUESTIONNAIRE
This questionnaire is completed by the students when they register with the
College Board to sit for the SAT. The questionnaire contains 43 items surveying
students on high school courses taken, participation in a sweep of extracurricu-
lar activities, academic achievement levels (i.e., grades), parental education, com-
bined family income, and their race/ethnicity (see http://www.collegeboard.org
for a copy of the SAT Questionnaire). Responses to these questions formed much
of the data for this study. In particular, we examined responses to parental edu-
cation, combined family income, high school GPA measures, and reported eth-
nicity. A sampling of the used items in this study is provided below. We also
provide the scoring (termed Value below) used for the responses to the SAT
Questionnaire item.
Enter the average grade for all courses you have already taken in each subject:
Arts and Music Mathematics
English Natural Sciences
Foreign and Classical Languages Social Sciences and Histor y
31
27918a_TCR_April2013_text_Layout 1 3/26/13 3:05 PM Page 31
Teachers College Record, 115, 040306 (2013)
Indicate your cumulative grade point average for all academic subjects in high school.
What is your most recent high school class rank?
Indicate the highest level of education completed by your father and mother.
32
Student Response Value
A or excellent (usually 90-100) 4
B or good (usually 80-89) 3
C or fair (usually 70-79) 2
D or passing (usually 60-69) 1
E or F or failing (usually 59 or below) 0
Student Response Value
A+ (97-100) 4.3
A (93-96) 4.0
A- (90-92) 3.7
B+ (87-89) 3.3
B (83-86) 3.0
B- (80-82) 2.7
C+ (77-79) 2.3
C (73-76) 2.0
C- (70-72) 1.7
D+ (67-69) 1.3
D (65-66) 1.0
E or F (below 65) 0
Student Response Value
Highest tenth 6
Second tenth 5
Second fifth 4
Middle fifth
Fourth fifth 2
Lowest fifth 1
Student Response Value
Grade school 1
Some high school 2
High school diploma or equivalent 3
Business or trade school 4
Some college 5
Associates or two-year degree 6
Bachelor’s or four-year degree 7
Some graduate or professional school 8
Graduate or professional degree 9
27918a_TCR_April2013_text_Layout 1 3/26/13 3:05 PM Page 32
TCR, 115, 040306 Influence of Family Income
What was the approximate combined income of your parents before taxes last year?
EZEKIEL J. DIXON-ROMÁN is an Assistant Professor of Social Policy and
Education in the School of Social Policy & Practice at the University of
Pennsylvania. His research is on the intersections of the sociology of edu-
cation, cultural studies, and quantitative methods. In addition to his
edited volume, Thinking Comprehensively About Education (with Edmund
W. Gordon, Routledge, 2012), he is writing a single-authored volume ten-
tatively titled Inheriting [Im]Possibility.
HOWARD T. EVERSON is Professor of Psychology and Senior Research
Fellow at the Center for Advanced Study in Education, Graduate School,
City University of New York. Professor Everson’s research and scholarly
interests focus on the intersection of cognition and assessment. He has
published and contributed to developments in educational psychology,
psychometrics, and quantitative methods.
JOHN J. MCARDLE is Senior Professor of Psychology at the University of
Southern California where he heads the Quantitative Methods training
program. His research has been focused on age-sensitive methods for psy-
chological and educational measurement and longitudinal data analysis
including publications in factor analysis, growth curve analysis, and
dynamic modeling of adult cognitive abilities. He is now writing a book
on longitudinal structural equation modeling with J.R. Nesselroade (APA
Books, 2013).
33
Student Response Value
Less than $10,000 1
About $10,000 to $ 15,000 2
About $15,000 to $ 20,000 3
About $20,000 to $ 25,000 4
About $25,000 to $ 30,000 5
About $30,000 to $ 35,000 6
About $35,000 to $ 40,000 7
About $40,000 to $ 50,000 8
About $50,000 to $ 60,000 9
About $60,000 to $ 70,000 10
About $70,000 to $ 80,000 11
About $80,000 to $ 100,000 12
More than $100,000 13
27918a_TCR_April2013_text_Layout 1 3/26/13 3:05 PM Page 33
... Students may also be likely to become friends with students of similar academic ability. SAT scores may also be associated with other factors, such as family income and race, which also predict friendships (Dixon-Román et al., 2013). ...
Article
Full-text available
When students are aware of the exam grades of their peers, does this information affect their subsequent exam performance? For example, knowing that my friend scored a higher grade on Exam 1 than myself might motivate me to improve my performance on Exam 2, or might frustrate me such that I stop trying to catch up. We analyze whether students’ performance is shaped by the grades of their classmates. To answer this question, we use survey-based data on students’ connections to other students with the grades that students obtained in a class. We find that a peer effect on grades does exist, where students who know that the grades of their friends were higher than their own on the first exam are motivated to improve their score on the following exam.
... For example, the course selection decisions made by first year college students are stratified by social class [2]. One of the most frequently discussed sites of these ideas and arguments is college admissions, particularly through the well understood correlations between household income and test scores [3]. This line of research along with the Covid-19 pandemic have coalesced to push many universities' decisions to drop standardized testing from first year admissions. ...
Preprint
Full-text available
These authors contributed equally to this work. Abstract Stratification in college application materials is a contentious topic in social science and national discourse in the United States. This line of research has also started to use computational methods to consider qualitative materials, such as personal statements and letters of recommendation. Despite the prominence of this topic, fewer studies have considered a fairly common academic pathway: transferring. Approximately 40% of all college students in the United States transfer schools at least once. One quirk of the system is that students from community colleges are applying for the same spots for students already enrolled in four year schools and trying to transfer. How might these different institutional experiences and the transfer application itself reflect social stratification? We leverage a dataset of 20,532 transfer admissions essays submitted to the University of California system to describe how transfer applicants are stratified linguistically, culturally, and narratively with respect to academic pathways and essay prompts. These results show different types of stratification that can emerge in educational processes intended to equalize opportunity and how combining computational and human reading might illuminate them.
... This approach perpetuates problematic binary relationships, such as singular inputs and outputs of students through post-secondary institutions, negative connotations of STEM exit, the establishment of disciplinary hierarchies, and the resulting rejection of re-entry. Contemporary research has not only highlighted the flawed nature of these approaches but also the structural policies and measures of student success (Dixon-Román et al., 2013;Romero, 2016). It has also been attributed to a reductionist philosophical approach and to furthering of the divide between the applied disciplines. ...
Thesis
The purpose of this study was to examine the extent to which motivational and persistence factors predict the occupational career choices of underrepresented students in their pursuit of a STEM career. Data selected from the High School Longitudinal Study beginning with the base year through the fourth wave were employed in a large-scale multinomial regression analysis. Anticipated STEM occupation at the age of 30 was examined across six years of complex survey data using multiple taxonometric definitions. Social Cognitive Career Theory provided the theoretical framework for defining relevant factors affecting this STEM pursuit construct. The findings from the study suggest that by varying student perspectives on their expected STEM careers, the resulting pathway of pursuit is affected by a different set of predictors. Typographic models developed through fitting multinomial logistic regression models also suggest that female students are propelled into specific STEM careers through early mathematics identity, mid-study science utility, and an evolving dynamic between parent and student expectations. The results additionally highlight race and ethnicity differences which more closely, though less significantly, mirror those of female students. The overall results of these findings raise questions about the continued use of a STEM pipeline metaphor in describing student pursuit. Moreover, adjacent policies, theoretical frameworks, and research methods aligned to this construct should be reviewed on how they portray an inaccurate picture of pursuit amongst underrepresented students seeking STEM careers.
... Taking an example of a standardized academic achievement test such as SAT (a standardized test for college admissions in the United States), biases against unprivileged ethnic/racial groups have been reported decades ago. First, socioeconomic status and family background including parental education were considered to substantially affect child development and academic test score via factors such as nutrition and out-of-schooling test preparation (63,64). Inequalities in socioeconomic status and family background across ethnic/racial groups could hence act as a proxy for disparities in academic achievement test. ...
Article
Full-text available
Algorithmic biases that favor majority populations pose a key challenge to the application of machine learning for precision medicine. Here, we assessed such bias in prediction models of behavioral phenotypes from brain functional magnetic resonance imaging. We examined the prediction bias using two independent datasets (preadolescent versus adult) of mixed ethnic/racial composition. When predictive models were trained on data dominated by white Americans (WA), out-of-sample prediction errors were generally higher for African Americans (AA) than for WA. This bias toward WA corresponds to more WA-like brain-behavior association patterns learned by the models. When models were trained on AA only, compared to training only on WA or an equal number of AA and WA participants, AA prediction accuracy improved but stayed below that for WA. Overall, the results point to the need for caution and further research regarding the application of current brain-behavior prediction models in minority populations.
... Consistent with the UC findings, the College Board itself reports gaps in SAT performance as a function of racial and economic demographics, with discrepancies as large as 15% to 20% for Black and Latinx students compared with White and Asian students (College Board, 2014Board, , 2020. When considering the intersection of race and family income, the results are even more striking: The effect of family income on SAT performance is almost twice as large for Black students than for White students (Dixon-Román et al., 2013). Similar performance discrepancies exist across other common tests used for university admissions (e.g., ACT, 2020) such as intelligence quotient and achievement tests commonly used for gifted identification (e.g., Kena et al., 2016;Silverman, 2009). ...
Article
Full-text available
The importance of creativity for learning, an equitable education, and a competitive nation warrants a broader and deeper understanding of this topic, including how creativity is assessed. This review focuses on subjective creativity assessment, a popular assessment approach that uses judges’ subjective definitions of creativity, and examines its reliability and validity evidence collected from 84 empirical studies under the theoretical frameworks of the 2014 Standards for Educational and Psychological Testing and rater-mediated assessments. The main findings include: 1) The reviewed studies vary across domains, characteristics of subjects/objects and raters, and rating instructions and scales; 2) The major reliability evidence was provided by Cronbach’s alpha and correlations of rating scores, and the major validity evidence came from the evidence based on relationships with other variables through the use of correlations; 3) Cronbach’s alpha values differed through an interaction between domains and judges’ expertise level, and correlations of rating scores differed by domain and judges’ expertise level; 4) There was strong convergent validity evidence between creativity and novelty but a weak discriminant validity evidence between creativity and technical goodness and liking. These findings suggest that the subjective creativity assessment approach shows a good level of reliability and validity but has some degrees of unreliability and invalidity that need to be addressed with good research practices and more advanced measurement theories and methods.
Article
This chapter provides a systematic, synthesizing, and critical review of the literature related to assessments of creativity in education from historical, theoretical, empirical, and practical standpoints. We examined the assessments used in the articles focusing on education that are published from January 2010 to May 2021 in eight creativity, psychological, and educational journals. We found that the assessments of creativity in education are split between psychological and education research and have increased international participation. Additionally, these assessments are more general than specific and focus more on cognitive than noncognitive aspects. Like previous reviews of assessments of creativity in general, this review showed that creativity in education is still mainly assessed by divergent thinking or creativity tests, self-report questionnaires, and product-based subjective techniques. We analyzed the benefits and drawbacks of each approach and highlighted many innovations in the assessment. We further discussed how the major assessment approaches address race, ethnicity, class, and gender issues in education. We concluded the review with recommendations for how to better assess creativity in education and how assessments of creativity in education contribute to our understanding of the creative educational experience and democratizing education.
Article
Background/Context To enhance postsecondary completion and minimize equity gaps, researchers have focused on defining, measuring, and developing students’ college readiness, or the preparation required to persist in higher education. While this work has been useful to identify the ingredients of postsecondary success, the emphasis on individual achievement runs the risk of portraying marginalized students as deficient. Culturally relevant studies that highlight institutional accountability for college readiness are needed to inform policy and practice. Purpose/Objective/Research Question/Focus of Study Using Holland and colleagues’ (1998) figured worlds theory, this study examines college readiness through the lens of student identity in an urban magnet high school. We investigate how first-generation, low-income students of color interpreted and negotiated local discourses and artifacts to “figure” college-going identities—that is, who they should become and how they should behave to earn a college degree. The purpose of this empirical approach is to contribute information that can inform college readiness efforts nationwide. Research Design The study utilizes an ethnographic approach that focused on how students conceptualized and developed their identities within the figured world of the magnet school. Data collection took place over the course of one school year and included over 200 hours of participant observation, in-depth interviews with 25 students and school staff, and document analysis. Findings The figured world of Jackson Magnet fostered and reinforced a hierarchy that consisted of magnet students (“scholars”) and their counterparts in the regular school (“ditchers”). A local feedback loop implied that the magnet school provided more rigorous college preparation than the regular school and, by extension, magnet students figured they would be ready for college. However, real-world feedback (standardized test scores) suggested magnet course rigor did not accurately reflect postsecondary standards. The result was that magnet students were underprepared but did not know it, an outcome that positioned them to experience a drastic identity reckoning in college. Conclusions/Recommendations While most college-readiness research focuses on academic skills preparation, our findings reveal the need to consider how high schools prepare students in terms of identity. In particular, our data suggest the threat of well-intentioned achievement discourses on pre-college identity development. Salient questions include whether and how to increase postsecondary opportunity without socializing students either to discriminate against their peers or to figure a false identity that undermines preparation and future cognitive stability.
Book
Full-text available
Much popular belief--and public policy--rests on the idea that those born into poverty have it in their power to escape. But the persistence of poverty and ever-growing economic inequality around the world have led many economists to seriously question the model of individual economic self-determination when it comes to the poor. In Poverty Traps, Samuel Bowles, Steven Durlauf, Karla Hoff, and the book's other contributors argue that there are many conditions that may trap individuals, groups, and whole economies in intractable poverty. For the first time the editors have brought together the perspectives of economics, economic history, and sociology to assess what we know--and don't know--about such traps. Among the sources of the poverty of nations, the authors assign a primary role to social and political institutions, ranging from corruption to seemingly benign social customs such as kin systems. Many of the institutions that keep nations poor have deep roots in colonial history and persist long after their initial causes are gone. Neighborhood effects--influences such as networks, role models, and aspirations--can create hard-to-escape pockets of poverty even in rich countries. Similar individuals in dissimilar socioeconomic environments develop different preferences and beliefs that can transmit poverty or affluence from generation to generation. The book presents evidence of harmful neighborhood effects and discusses policies to overcome them, with attention to the uncertainty that exists in evaluating such policies.
Article
The debate over "aptitude" versus "achievement" tests in college admissions is an old one. Aptitude-type tests, exemplified by the SAT I, are intended to assess students' capacity for future learning, whereas achievement-type tests, exemplified by the SAT II subject tests, are designed to assess students' current mastery of college-preparatory subjects. As one of the few institutions in the nation that requires both the SAT I and SAT II, the University of California (UC) has an extensive database with which to assess their relative utility in predicting student success in college. This study examines the relationship between SAT scores and freshman grades based on the records of 77,893 students who entered UC between Fall 1996 and Fall 1999. The study found that (a) the SAT II achievement tests are consistently better predictors of student success at UC than the SAT I, although the incremental gain in prediction is relatively modest and there is substantial redundancy across the tests; (b) the predictive validity of the SAT II appears to be less conditioned by socioeconomic factors than is the SAT I; and (c) racial/ethnic group performance is substantially similar across the SAT I and SAT II. Contrary to the conventional wisdom, these findings suggest that the benefits of achievement tests for college admissions-greater clarity in admissions standards, closer linkage to the high-school curriculum - can be realized without any sacrifice in the capacity to predict success in college.