GENDER INEQUALITY IN COLLEGE ENROLLMENT AND STEM MAJOR
IN U.S. 1980-2013: A FAMILY RESOURCE PERSPECTIVE
By
I-Chien Chen

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
Sociology- Doctor of Philosophy
2017

ABSTRACT
GENDER INEQUALITY IN COLLEGE ENROLLMENT AND STEM MAJOR
IN U.S. 1980-2013: A FAMILY RESOURCE PERSPECTIVE
By
I-Chien Chen
The goal of this dissertation is to understand the gender gap in college enrollment and participation
in STEM majors, considering changes in family size and mothers’ labor force participation.
Different types of family resources and gender-specific family contexts that affect children's
college enrollment and STEM participation were explored. I employ a three-article format to
examine the links between (1) the trend in the number of siblings, family resources, and the
transition into college; (2) the relationship of sibling configuration, sibship-gender composition,
and family resource allocations to college enrollment; and (3) the effect of married mothers’
employment on children’s participation in STEM related majors.
My first article shows that females’ advantage in college enrollment is associated with number of
siblings and family resource allocation. This female advantage in college enrollment strengthens
over time and widens between smaller and larger families.
My second article draws from the resources dilution model and gender development literature to
identify new advantages in college enrollment emerged from smaller families, daughters in the
sibling gender majority position, and families with greater socio-cultural resources. The third
article examines the effect of married mothers’ employment, found that mothers with full-time or
high prestige jobs had a positive effect on children’s participation in STEM in the most recent
cohort. School achievement, math self-efficacy, and parents’ expectations are linked to mothers’
role and schooling processes. Taken together, these results demonstrate that the presence of
specific family contexts, family resources, and mothers’ roles may exert a net positive effect on

children’s college opportunities and STEM interests. In investigating these three linkages I show
how the number of siblings, sibling composition, maternal employment, and family resources
affect female advantage in college enrollment and STEM participation.

Copyright by
I-CHIEN CHEN
2017

ACKNOWLEDGEMENTS

I would like to start by thanking my dissertation chair, Barbara Schneider. Her insightful and
supportive mentorship blossomed into a great professional training that has been central to my
development as a researcher. I am extremely grateful to Ken Frank, who has been very supportive
and encouraging since the days I took his significant course of social network analysis. I would
also like to thank the other members of my committee, Cathy Liu and Steve Gold, for reading
drafts of dissertation and providing insightful perspectives on my research work and career
development.
I have been blessed to have met and learned from Yeu-Sheng Hsieh, Chyi-In Wu, and
Tse-Chuan Yang. They encouraged and inspired me to pursue an academic career as a social
scientist. I also thank the Department of Sociology, The Great Lakes Integrated Sciences and
Assessments Center, Middle-school Mathematics and the Institutional Setting of Teaching at
Vanderbilt University for the support of research assistantship during my graduate studies.
My friends and colleagues are as important as comrade-in-arms in my graduate student life.
There have been times when I questioned my decision to pursue this degree. At that time, it was
them who kept me going. I thank Emily, Kyoko, Yun, Qinyun, Jihyun, Annie, Lysann, Janet,
Barbara, Megan, Jessie, Christy, Lin-ying, Jerry, Mengchuan, Tzufeng, Dawn, Bette, You-Kyung,
Seung-won, Tingqiao, Yen-Ling, Michelle, and more.
Finally, I thank my parents, especially my mother and siblings who have really been there
from the beginning. To my mother, thank you for always providing support even when you weren’t
too sure about all this stuff; to my sisters, thank you for being the one to whom I could always turn
to and count on.

v

TABLE OF CONTENTS

LIST OF TABLES ................................................................................................................. viii
LIST OF FIGURES ................................................................................................................... x
CHAPTER ONE THE FAMILY ADVANTAGE IN THE TRENDS OF SIBLING SIZE:
EXAMINING THE RELATIONSHIP BETWEEN FAMILY RESOURCES AND COLLEGE
ENROLLMENT ........................................................................................................................ 1
Introduction ................................................................................................................................ 1
Background ................................................................................................................................ 2
The Shifting of Family Culture and Investment in Children College Education .............. 2
Changes in Family Investment and Family Resources ...................................................... 4
Data and Methods ...................................................................................................................... 6
Data .................................................................................................................................... 6
Measures of Independent and Dependent Variables .......................................................... 8
Dependent Variables .......................................................................................................... 8
Independent Variables ........................................................................................................ 9
Analytic Methods ..................................................................................................................... 11
Results ...................................................................................................................................... 13
Trends in Number of Siblings, Family Resources and College Enrollment .................... 13
Trend in Two-Year and Four-Year College Enrollment with Family Resources ............. 16
Discussion and Conclusion ...................................................................................................... 20
APPENDICES ......................................................................................................................... 24
APPENDIX A FIGURES FOR CHAPTER ONE ........................................................... 25
APPENDIX B TABLES FOR CHAPTER ONE ............................................................. 28
REFERENCES ........................................................................................................................ 37

CHAPTER TWO SIBLING SIZE, GENDER COMPOSITION, AND PARENT INVESTMENT
IN COLLEGE ENROLLMENT .............................................................................................. 42
Introduction .............................................................................................................................. 42
Background .............................................................................................................................. 43
Resources Dilution Model: Gender-Specific Dilution ..................................................... 43
Sibship composition and Gender Socialization ............................................................... 44
Differential Parental Resource and Cultural Supports ..................................................... 45
Data and Methods .................................................................................................................... 47
Data .................................................................................................................................. 47
Measures of Dependent Variables .................................................................................... 48
Measures of Independent Variables ................................................................................. 48
Analytic Methods ..................................................................................................................... 50
Results ...................................................................................................................................... 52
Descriptive Results .......................................................................................................... 52

vi

Multi-nominal Regression Results................................................................................... 53
Conclusion and Discussion ...................................................................................................... 58
APPENDICES ......................................................................................................................... 63
APPENDIX A FIGURES FOR CHAPTER TWO........................................................... 64
APPENDIX B TABLES FOR CHAPTER TWO ............................................................ 67
APPENDIX C SUPPLEMENTAL TABLES ................................................................... 75
REFERENCES ........................................................................................................................ 79

CHAPTER THREE THE EFFECT OF MARRIED MOTHERS’ EMPLOYMENT ON
CHILDREN’S HIGH SCHOOL ACHIEVEMENT, MATHEMATICS SELF-EFFICACY AND
STEM MAJORS ...................................................................................................................... 84
Introduction .............................................................................................................................. 84
Background .............................................................................................................................. 86
Career Mothers and Impacts on Children ........................................................................ 86
Theoretical Perspectives: the Causal Effect of Career Mother on Children’s Cognitive
Development and Interest in a STEM Career .................................................................. 88
Career Mother and Family Supports ................................................................................ 90
Research Questions .......................................................................................................... 91
Data and Methods .................................................................................................................... 92
Data .................................................................................................................................. 92
Measures of Dependent Variables .................................................................................... 93
Measures of Independent Variables ................................................................................. 93
Analytic Methods ..................................................................................................................... 97
Results of Multiple-Group SEM Analyses ............................................................................ 100
Discussion and Conclusions .................................................................................................. 103
APPENDICES ....................................................................................................................... 107
APPENDIX A FIGURES FOR CHAPTER THREE ..................................................... 108
APPENDIX B TABLES FOR CHAPTER THREE....................................................... 113
APPENDIX C SUPPLEMENTAL TABLES ................................................................. 128
REFERENCES ...................................................................................................................... 133

vii

LIST OF TABLES

Table 1: Variable Description, Mean and Standard Deviation ...................................................... 29
Table 2: Gender Difference in College Enrollment and Family Resources across Three Decades
....................................................................................................................................................... 32
Table 3: Logistic Regression Predicting College Enrollment across Gender and Cohort ........... 33
Table 4: Multi-nominal Logistic Regression Model Predicted Two-Year and Four-Year College
Enrollment..................................................................................................................................... 35
Table 5: Sample Means by Cohort and Sex for Variables Uses in Analysis ................................. 68
Table 6: Mean Difference on Family Resources, by Child’s Gender and Sibship Position in NELS
and ADD Health cohort ................................................................................................................ 70
Table 7: Multinominal Logistic Regression Predicting College Enrollment Pattern Using NELS
(1990-1994) and ADD Health (1995-2008) .................................................................................. 72
Table 8: Multinominal Logistic Model on College Enrollment by Sibship Gender Composition in
NELS............................................................................................................................................. 73
Table 9: Multinominal Logistic Model on College Enrollment by Sibship Gender Composition in
ADD Health .................................................................................................................................. 74
Table 10: The Classification of Sibhip-sex Majority, Balance and Minority Context at Home ... 76
Table 11: Family Resources Variables .......................................................................................... 77
Table 12: Descriptive Statistics for the HS&B Teens Living with Married-Two-Biological Parents
families .........................................................................................................................................114
Table 13: Descriptive Statistics for the NELS Teens Living with Married-Two-Biological Parents
families .........................................................................................................................................115
Table 14: Descriptive Statistics for the ELS Teens Living with Married-Two-Biological Parents
Families ........................................................................................................................................116

viii

Table 15: Descriptive Statistics for the HSLS Teens Living with Married-Two-Biological Parents
Families ........................................................................................................................................117
Table 16: The Distribution of STEM Participation across Cohort of Teens ................................118
Table 17: Math self-Efficacy Beliefs across HS&B, NELS, ELS and HSLS data ......................119
Table 18: Cohort Structural Weight Invariance Tests ................................................................. 120
Table 19: Final SEM model by Cohort of the High School & Beyond ...................................... 121
Table 20: Final SEM model by Cohort of the NELS (1988) ...................................................... 122
Table 21: Final SEM model by Cohort of the ELS (2002) ......................................................... 123
Table 22: Final SEM Model by Cohort of the HSLS (2009) ...................................................... 124
Table 23: Final SEM Model by Cohort and Gender (reported structural paths only) in the HS&B
(1980) and NELS (1988) ............................................................................................................ 125
Table 24: Final SEM Model by Cohort and Gender (reported structural paths only) in the ELS
(2002) and HSLS (2009) ............................................................................................................. 126
Table 25: Summary of Positively Significant Indirect Effect between Mother Employment,
Math-Efficacy and School Average GPA by Girls and Boys Group over Cohort ...................... 127
Table 26: Four STEM Categories ............................................................................................... 129
Table 27: Average Demographic Characteristics in Each Survey (Including all types of families)
..................................................................................................................................................... 130
Table 28: Mother Occupational and Occupational Prestige Score ............................................. 132

ix

LIST OF FIGURES

Figure 1: Marginal Prediction of College Enrollment by Gender and Cohort within the Number of
Siblings ......................................................................................................................................... 26
Figure 2: Marginal Prediction of College Enrollment by Gender and Separated Cohort within the
Number of Siblings ....................................................................................................................... 26
Figure 3: College Enrollment Probability in Two-Year College by Gender and Cohort Conditional
on Number of Siblings .................................................................................................................. 27
Figure 4: College Enrollment Probability in Four-Year College by Gender and Cohort Conditional
on Number of Siblings .................................................................................................................. 27
Figure 5: College Enrollment by Three Context of sibship Gender Composition in NELS cohort
....................................................................................................................................................... 65
Figure 6: College Enrollment by Three Context of sibship Gender Composition in ADD Health
cohort ............................................................................................................................................ 65
Figure 7: Four-Year College Enrollment Probability by Sibship Gender Composition in the NELS
Cohort ........................................................................................................................................... 66
Figure 8: Two-Year College Enrollment Probability by Sibship Gender Composition in the ADD
Health Cohort ................................................................................................................................ 66
Figure 9:Theoretical Framework ................................................................................................ 109
Figure 10: Main Direct Effect of Mother with Full-time Job and Mothers’ Occupational Prestige
across Cohorts ..............................................................................................................................110
Figure 11: Indirect Effect of Mother with Full-time Job and Mothers’ Occupational Prestige ... 111
Figure 12: Indirect Effect of Mother with Full-time Job and Mothers’ Occupational Prestige ... 111

x

Figure 13: Indirect Effect of Mother with Full-time Job and Mothers’ Occupational Prestige ...112
Figure 14: Indirect Effect of Mother with Full-time Job and Mothers’ Occupational Prestige ...112

xi

CHAPTER ONE
THE FAMILY ADVANTAGE IN THE TRENDS OF SIBLING SIZE:EXAMINING THE
RELATIONSHIP BETWEEN FAMILY RESOURCES AND COLLEGE ENROLLMENT

Introduction
In the United States, the proportion of women obtaining a college degree or higher has
reached an apex (DiPrete & Buchmann, 2013). This female advantage in college attainment has
increased since the 1980s (NCES, 2014). Yet, the female advantage in college attainment remains
significant, despite a substantial demographic change: the growing importance of mothers in the
labor market and household. Whether or not the demographic changes associated with increased
family resources have contributed to the female advantage remain underexplored.
Several explanations have been suggested for the increasing gender gap in college
enrollment, including the occupational incentive, attitudes towards gender roles, and the growth of
educational returns and family expectations (Cho, 2006; Goldin, Katz, &Kuziemko, 2006; Jacobs,
1996; Sweeney, 2002; Flashman, 2013). However, little is known about whether families preserve
the family advantage by differential fertility (Vogl, 2016) and parental investment in children’s
education (Lareau, 2003; Steelman& Powell, 1989). The increasing number of college-educated
mothers and their decision about the number of children may facilitate the preservation of the
family advantage and daughters’ opportunities to attend a four-year or selective college (Mare &
Maralani 2006). To fill the gap, this study examines the extent to which potential factors influence
the female advantage in college enrollment in the context of demographic changes over two
decades. This study aims to answer the following research questions: (1) has the female advantage
in college enrollment changed over time? (2) has the female advantage in four-year college
enrollment widened compared to two-year college enrollment? (3) has the female advantage in

1

4-year college enrollment widened over time? (4) has change in the female advantage in college
enrollment diverged between small and large families? (5) to what degrees do family resources
explain the female advantage in two-year and four-year college?
The goal of this study is to examine the relationship between sibling size, family advantage
and the gender gap in college enrollment using three cohorts of high school seniors in the High
School and Beyond 1980 (HS&B), the National Education Longitudinal Study 1988 (NELS 88),
and the Education Longitudinal Study of 2002 (ELS02). These three datasets are nationally
representative surveys and are consistent with the birth cohort after 1965-66, which has been
identified as a period of children who grew up during a time period when women began to outpace
men in college enrollment and graduation rates.
Background
The Shifting of Family Culture and Investment in Children College Education
Women’s participation in the labor force and college education has increased since the 1980s.
Increasing numbers of women have earned college degree and economic independence over time
(Oppenheimer, 1994, 1997; Sweeney, 2002). According to Oppenheimer both women and men
tend to postpone marriage and parenthood because of higher education and career uncertainty.
When women and men postpone marriage and parenthood after obtaining a college degree, they
are more likely to enter a marriage and have a stable job, compared to those without a college
degree. According to this argument, families with college educated parents may have more family
resources in terms of better human resources, marital disposition and stable jobs.
College-educated mothers are less likely to have children outside marriage and are more
likely to have smaller number of children at home than less educated married mothers and
cohabiting mothers (McLanahan & Percheski, 2008; Sweeney & Raley, 2014). In contrast,

2

less-educated women are more likely to bear children outside of marriage and to have multiple
partners over their children’s life course (Cherlin, 2010; Chiappori, Iyigun, & Weiss, 2009).
Family structure and number of children at home are associated with children's educational
opportunities, and affect the availability of family resources that parents can draw upon to raise
their children.
The family production hypothesis is especially notable in Becker's Human Capital theory,
which emphasized that “the production of family capital is in the form of investing in children’s
human capital” (page 64). Schultz (2007) and Becker and Tomes (1976) also found that parents
may choose to have fewer children in order to increase the quality of family reproduction and
through increasing family investments. From this perspective, families with fewer children are
more likely to make greater investments in their children.
In addition to human capital theory, the number of siblings also has theoretical significance in
the resources dilution model and stratification literature. Previous research has suggested that the
number of sibling s is associated with children’s achievement and post-secondary education
(Powell & Steelman, 1989; Steelman & Powell, 1989, 1991). For example, Powell and Steelman
(1993) found that children living in families with more siblings perform worse in school and have
lower GPAs. Conley and Glauber (2006) also found that increasing number of siblings reduces the
children’s attendance at a private high school and four-year college enrollment. Along with the
demographic changes of the decline in the number of children per family, I expect the children’s
four-year college opportunities may change in response to smaller families. Smaller families refer
to families with one or two children in this study and this tendency in terms of demographic
changes may boost daughters’ educational opportunities compared to sons’ opportunities. I
propose the following hypotheses:

3

(H1) The female advantage in college enrollment is increasing over time.
(H2) The female advantage in four-year college enrollment is larger than in
two-year college enrollment.
However, the gap in this line of research is less concerned with the rise of women’s position in
the household. When mothers contribute to family income, they may have more control over the
allocation of family resources among children. This bargaining position puts more weight on
mothers’ decision-making and preference in how to distribute family resources towards daughters’
or sons’ college education (England 2010). This gender-specific perspective emphasizes that the
family culture and investment have increased simultaneous with increases in mothers’ college
education, labor force participation, and the declining number of children at home (Behrman &
Rosenzweig, 2002; Mare & Maralani, 2006; Suitor, Plikuhn, Gilligan, & Powers, 2008). This
perspective also suggests that the growth of mothers’ resources may have a greater effect on
daughters than on sons, and considers this as evidence of a gender-specific cultural socialization
processes. This leads to the third and fourth hypotheses:
(H3) The female advantage in 4-year college enrollment is widening over time.
(H4) The female advantage in college enrollment is widening over time between
children from small families and those from large families, and this gender
difference is larger in the later cohort.
Changes in Family Investment and Family Resources
Structural advantages of family resources have become more important in shaping gender
parity, especially in respect to those resources that parents are more likely to invest in daughters,
rather than sons. For example, parents’ investment strategies differentiate resource allocation by
children’s achievement and sex (Conley & Glauber, 2006; Lee, 2008). Some scholars have

4

emphasized that parents’ perception about their children’s school performance determines their
resource allocation and willingness to provide financial support for college tuition (p.192)
(Crosnoe, 2001; Eccles& Harold, 1993). As a result, parents tend to favor investment in daughters
because they are more likely to have a higher GPA in high school compared to sons. In addition to
school achievement, changes in gender-role attitudes have also shifted parents’ values about the
importance of college education. Thus, daughters with higher school achievement are more likely
to receive more family resources and parental investment in college education than sons.
American societal norms have arrived at a consensus that parents allocate their resources
equally among children. However, the growth of non-traditional families and multiple partner
fertility may shift this consensus in response to family composition and the availability of family
resources at home. Family resources refer to monetary and non-monetary investments, use of time,
parent-child interactions and social-psychological supports. According to this definition, family
resources can be categorized into four types: 1) social-psychological, 2) financial, 3) cultural, and
4) social resources. Socio-psychological resources refer to parents’ encouragement and
expectations of their children to obtain a college education. The effects of parents’ encouragement
and expectations have been examined in the Status Attainment and Wisconsin models (Andrew &
Hause, 2011; Scritchfield & Picou, 1982; Sewell & Hauser, 1980). Parents’ decisions regarding
resource allocation may reflect their relationship and emotional distance with their children.
Crosnoe (2004) identified that parent-children communications serve as an important
social-psychological process that may facilitate resource allocation towards children’s higher
education. Financial resources refer to the educational items that a family provides for targeted
children. Previous research has shown a positive relationship between educational items and
children’s school achievement, particularly for books, computers and school-related materials

5

(Degraaf 1986). Cultural resources represent how well parents prepare their children for school,
which serve as indications of parents’ values about the importance of schooling (Lareau, 1999;
Lareau & Weininger, 2003). Social resources refer to the family-school connection in regards to
how parents participate, communicate, and interact with school teachers, activities, and
organizations (Kim & Schneider, 2005; Schneider & Coleman, 1993). For example, parents’
involvement in school related activities can enable them to access more information about school
learning opportunities, which may help their children’s development and achievement (Perna,
2006). This line of research leads to the fifth hypothesis:
(H5) The role of family resources in explaining the proportion of female advantage in
college enrollment is increasing over cohorts and this tendency is more
pronounced for four-year college enrollment than two-year college
enrollment.
Data and Methods
Data
In order to answer these questions, I utilize data that includes the number of siblings at home,
family resources, parents' investment strategies and measures of children's non-cognitive skills
and cognitive achievements so that I can demonstrate the effect of inter- and intra-family
gender-specific resources on the decision to enroll in college. In addition, covariates that measure
children's school performance and college preparation in early adolescence are used
contemporaneously, rather than retrospectively. I use follow up data to track how early school
performance and parents’ educational investment affect later college enrollment. My data offers
reasonable comparability in these measures across birth cohorts. I am able to meet these data
requirements by using survey data on three cohorts of 10th grade students: High School and

6

Beyond 1980, the National Education Longitudinal Study 1988, and the Education Longitudinal
Study of 2002. These data sets allow me to compare three cohorts of 14-to-20 year old youth
interviewed in 1980, 1988, and 2002 and who had at least three follow-up waves of data. These
survey cohorts are consistent with the birth cohort after 1965-66, which has been identified as a
period of children who grew during a time when women began to outpace men in college
enrollment and graduation rates. I describe each of these surveys in more detail below.
The first cohort comes from the sophomore class survey of High School and Beyond (HS&B),
a nationally representative sample of high school sophomores first surveyed in 1980. The sample
includes information on family background, standardized test scores, non-cognitive traits, college
enrollment and achievement for 27,683 sophomores from 988 schools across the US. A total of
14,670 students were included in the sample across all 5 waves.
The second cohort comes from the National Education Longitudinal Study of 1988
(NELS88). NELS88 started with a survey of a nationally representative sample of 8th graders
(aged 14), who also participated in the first follow-up study two years later in 1990, when
respondents were in 10th grade. In this study, I focus on 10th graders who participated in the first
follow-up survey. The follow-up data offers the same variables as the base-year survey (1988),
including standardized test scores, non-cognitive skills, and family characteristics. The later waves
describe college enrollment and achievement. The first follow-up study covers 16,589 high school
sophomores originally interviewed in 1990.
The third cohort comes from the Education Longitudinal Study of 2000 (ELS 2002), a
nationally representative sample of 10th graders in 2002. The study provides information on
cognitive, as well as non-cognitive skills. The sample includes about 12,441 students from 750
schools across the US.

7

Measures of Independent and Dependent Variables
Dependent Variables
My outcomes of interest include whether a student is enrolled in college, and whether the
student is enrolled in two-year or four-year college. College enrollment is defined by enrollment in
a first postsecondary institution for at least 6 months, which distinguishes between (a) students
who enrolled in any postsecondary college, and (b) students who enrolled in a 4-year college. My
college enrollment outcome measures whether 10th graders enrolled in college two years after
high school graduation.
School GPA measures the overall average high school grades in HS&B in the third wave of
follow-up data and average school GPA in NELS base year survey data. In the ELS second wave
follow-up data, GPA was measured for all courses taken in the 9th through 12th grades. To make
comparisons between datasets, I use the quartile coding of GPA composite, ranging from 1 to 4
across the three data sets.
I also measure student high school achievement by generating NaĂŻve Percentile-Ranking with
10th graders’ composite standardized test score (including reading and math). I recognize that
there is a huge difference between different standardized procedures and test composite scores
across datasets. To accomplish this analysis, I applied a simple linear equating procedure to
generate a percentile-ranking based on the standardized test composite score (Friedkin & Thomas,
1997). This percentile-rank represents the relative position of test-score compared to other
students in each survey cohort. This procedure makes a cohort comparison possible. Although
transcript data is another option to link student performance across three data sets, it produces
many missing cases when students drop out of high school. I also include the overall GPA in the
final model and the main findings are still consistent across models.

8

Independent Variables
(1) Number of siblings. All three data sets asked (in various ways) “the number of siblings
that 10th grader had” in the HS&B data. In the NELS data, respondents were asked about the
number of siblings in both the student and parent surveys. In the ELS data, 10th grade respondents
were asked about “the number of siblings living with the 10th grade respondent.”
(2) Family Structure (Arrangement). The measure of family structure in this study can only
distinguish between four types of family arrangements with 10th graders living at home: (1) intact
family: represents children living with both biological parents, (2) step-family: represents children
living with mother and male guardian; father and female guardian, (3) single family: represents
children living with mother only or father only, (4) other guardians/relatives: represents children
living with relatives or guardians. I do recognize the limitation of this measure, but the HS&B
survey did not provide clear and exclusive information about parents’ marital status or the duration
of marriage (Astone & Mclanahan, 1991). To make a reliable comparison across three datasets, I
use this variable to represent the family structure.
(3) Family Resources. There are five types of family resources measured in this study that are
available to compare across cohorts. The first type of family resource indicates parents’ human
resources, which were measured by fathers’, and mothers’ highest educational achievement. To
make a comparison across three data sets, I follow the same scale (6 categories, See Appendix B,
Table 1) to recode fathers’ and mothers’ highest level of education. In HS&B, I used information
reported from students to construct the education composite and substituted parent self-reported
information. In NELS:88 and ELS:2002, parent self-reported information was used to construct
the education composite; student-reported parent education was substituted when parent

9

information was missing. I also include the level of education for both mother and father in the
analysis if the survey provided information.
The second type of family resource indicates (educational) material resources at home. I
measure how many educational items are available at home, which included four harmonizing
items in three datasets: (a) family has a daily newspaper, (b) family has a computer, (3) family has
more than 50 books, and (4) child has his/her own room. I generated this measure by taking the
sum of all four items at home. One hypothesis is that families with sons are more likely to have
more educational materials than those with daughters (Conley & Glauber, 2006; Steelman &
Powell, 1991). This trend may shift rapidly in the later cohort and is more likely to support
daughters’ educational opportunities, as well as sons’ opportunities.
The third type of family resources indicates family socio-psychological resources in terms of
parents’ educational expectations in pursuit of higher education. There is an abundance of
literature to support the effect of parents’ educational expectation (Morgan, 1998, 2006), and
expectation from significant others (Cheng & Starks, 2002; Scritchfield & Picou, 1982). For this
study, I recode the original educational expectation into seven categories across three data sets.
Both the NELS and ELS data also included a “don't know” option on the questionnaire. In this
study, I aim to use students’ perceived parents’ educational expectation as a social-psychological
factor. I assume that the higher the parental expectations students are exposed to, the higher
academic pressure that students perceive from parents. In this setting, if students report that they
don’t know about their parents’ expectations, this implies less perceived pressure from parents.
The forth type of family resource indicates family and cultural resources in relation to school
value. I measure the school preparation for 10th graders at home, which includes three identical
items among three data sets: (1) how often the student goes to class without pencil/paper; (2) how

10

often the student goes to class without books; (3) how often student goes to class without
homework done. I recoded the original scale with 1 indicating “frequently” and 4 as “never”. This
measure also takes the sum of all three items, ranging from 0-12.
The fifth type of family resources indicates social resources for parents participating in school
related organization, activities, and meetings. I measure how many school related activities that
parents ever participated in at their 10th graders’ school, which includes four identical items
among three datasets. This measure takes the sum of four types of activities: (1) how often parents
attend school meetings; (2) how often parents phoned teacher/ counselor; (3) how often parents
attended school event; and (4) how often parents acted as a volunteer at their 10th graders’ school.
(4) Other Covariates. Family SES: I also included the family SES composite score as a
covariate in the model. The family SES composite score is determined by education and
occupation of the head of the family and total family income. School-level characteristics include
region, school type (Public, Private, or Catholic) also included in the model. For region, I
generated three dummy variables and north-east serves as the reference group. For type of high
school, public school serves as a reference group. I recoded all variables with a similar scale across
all three data sets and use those with similar scale to take the sum of each resource variable. I
account for missing cases for all independent variables by generating missing flags and include
this in the final model. For the weighting procedures, I report current results based on the weight of
panel data, which accounted for the potential attrition rate of panel data is in the post-secondary
outcomes.
Analytic Methods
Table 2 shows descriptive statistics of college enrollment, number of siblings and family
resources by gender and cohort. I start with simple logistic regression where college enrollment is

11

regressed on cohorts and gender difference by five dummy variables, explanatory variables and
other covariates. Model-1 provides a baseline estimate of cohort and gender differences in college
enrollment controlling for school percentile ranking, mothers’ education, family structure, family
SES, students’ expectation, high school types, region, race and ethnicity. I then add the key
explanatory variables, such as number of siblings and each of the family resources in a nested
model fashion to identify the degree to which number of siblings and family resources explains
gender and cohort gaps in college enrollment.
A t-test was applied to examine the gender and cohort effects of the linear combinations. The
results of the t-test are reported in the bottom panel among tables. “Yes/ No” indicates whether
there is a significant difference in the logged odds of enrolling in college between cohorts among
females and males. “Female/ Male” indicated whether there is a significant gender difference in
the logged odds of enrolling college within cohorts. I also account for clustering in high schools by
obtaining robust standard errors in all regression analyses.
The next step is to apply a t-test to examine whether the effects of two explanatory variables
are changing for men and women across cohorts. The college enrollment rate for females in the
1980s serves as the reference group in all models. I adopt the approach of the cohort perspective to
understand the relationship between family resources and female advantage in college enrollment
patterns (Flashman, 2013). To visualize the opportunity differences between gender and cohort, I
plot the predicted probability of college enrollment based on Model 3 at Table 3 in Figure 2.
The last step is to examine whether two-year and four-year college enrollment patterns vary
by gender and cohorts, conditional on changes in the number of siblings and family resources. By
applying multi-nominal logistic regression model, I examine the likelihood of attending two-year

12

and four-year colleges, compared to those not enrolled in any post-secondary education two years
after high school graduation.
In this analysis, I focus on the three way-interactions between gender, cohort, number of
siblings and family resources variables. In the table, I assess whether the main effects of college
enrollment varied significantly across cohorts and whether these effects differ between females
and males. In the bottom of the table, I also apply a t-test to examine whether the effects are
significantly different in a cohort-by-gender manner, such as whether the effect of four-year
college enrollment for daughters in 1980 is statistically different from the effect of four-year
college enrollment for daughters in 2002. I present multi-nominal logistic regression coefficients
in Table 4 and plot the predicted probability in Figure 3 and Figure 4.
Results
Trends in Number of Siblings, Family Resources and College Enrollment
The proportion in Table 2 showed increasing rates of college enrollment within and across
cohort-gender subgroups. The results showed significant growth in college enrollment, especially
for women. Among the 1980 cohort, 63% of women attended college and this trend continued to
increase to 71% in the 1988 cohort, and 79% in the 2002 cohort. By contrast, 56% of men attended
college in the 1980 cohort, and this trend continued to increase to 66% in the 1988 cohort, and 71%
in the 2002 cohort. The gender gap in college enrollment was about 7% in 1980, and slightly
decreased in 1988, but increased to 8% in the later cohort of 2002. The gender gap in college
enrollment was observed across all pathways.
Two major explanatory variables in terms of number of siblings and family resources were
considered in the analytical model, along with a cohort-by-gender linear combination of a series of
dummy variables. This study expected that the declining number of siblings and increases in

13

certain types of family resources may explain the growing gender gaps in college attendance and
the patterns of two-year college enrollment and four-year college enrollment.
The average number of siblings decreased from 1980 to the 2002. This implies that the
growth trend of living in a smaller family was consistent for both families of daughters and
families of sons. The proportion of step-families and single-families also increased over time.
Eleven percent of daughters lived in a step-family in 1980, and this increased to 15% in 2002. In
contrast, only 8% of sons lived in a step-family in 1980, and this increased to 15% in 2002.
Moreover, 17% of daughters lived in a single family in 1980, and this increased to 21% in 2002.
This tendency was also observed for sons.
Family resources in terms of materials, socio-psychological, cultural, and social also
increased over cohorts, except for social resources in the ELS cohort. Family material resources
and socio-psychological resources also showed significant gender differences among cohorts. The
proportion of college educated mothers increased over cohorts. Following this trend, the
proportion of mothers with some college increased more rapidly than mothers who graduated from
college.
Table 3 shows the results for the logistic regression. The top of the table reports the logged
odds of college enrollment estimates compared to those who are not attending college. Beginning
with the main effect of cohort-gender difference, I found that the female advantage in college
enrollment held across cohorts in Model-1, controlling for number of siblings, relative percentile
ranking, and other demographic characteristics. The cohort difference among females’ college
enrollment was significant compared to the 1980 cohort and increased over cohorts compared with
the 1980 cohort. A daughter from the 1988 cohort with the same percentile ranking class rank,
family SES, and other covariates as a daughter in the 1980 cohort had an odds 1.43 times that of a

14

daughter in the 1980 cohort of attending college (𝑒𝑒 .361 = 1.43). That same daughter would have
odds more than 1.61 times those of the 1980 daughter if she were a member of the 2002 cohort.

This finding confirms the first hypothesis: the female advantage in college enrollment increases
over time. On the other hand, the cohort difference among males’ college enrollment was also
significant compared to the 1980 cohort. However, the gender difference among cohorts in the
bottom panel of table 3 showed a significant advantage of college enrollment in favor daughters
over sons.
What role does number of siblings play in this gender gap? Model 2 describes the
cohort-gender difference in college enrollment controlling for differences in number of siblings,
percentile ranking of school achievement, and other covariates. The results show that the
cohort-gender difference decreased after number of sibling was controlled for in the model. As the
results suggest, with the inclusion of number of siblings, the female advantage coefficient
decreased approximately 5% in the 1988 cohort, and 7.5% in the 2002 cohort. The cohort
difference among females and among males was still observed. Moreover, number of siblings was
negatively associated with college enrollment. This confirms the second hypothesis: the presence
of an additional sibling at home decreases the opportunities for college enrollment. Figure 2 shows
that the predicted probabilities of college enrollment for each cohort-gender subgroup at different
number of siblings. In the HS&B cohort, the female advantage in college enrollment dropped
immediately with an additional sibling at home, compared to females in single-child families. In
the NELS cohort, the female advantage in college enrollment remained at a similar level, with up
to three additional siblings. This trend points to the possible harmful effect for those children
living in larger families in the NELS cohort. In the ELS cohort, the pattern was similar to the
NELS cohort, but the decreasing tendency was slower than the NELS cohort in the larger families.

15

Two patterns were observed in Figure 1 and Figure 2. In general, in the HS&B cohort,
daughters and sons decreased their college enrollment rate with an additional sibling at home. In
the NELS and ELS cohort, daughters consistently held the highest college enrollment rate,
particularly for those living in families with one or two children. Although sons received similar
advantages in smaller families, their college enrollment decreased when the number of siblings
was larger than one. In addition, a family with four siblings presented the worst environment for
children’s college enrollment, and this effect held for both daughters and sons. To examine the
divergent gender gap hypothesis, for example, the predicted probability of college enrollment for
sons with four siblings in the 1980 cohort, was 8% lower than that for the 1980 daughters (58.42%
versus. 66.21%), holding other covariates as the average. In contrast, the predicted probability of
college enrollment for sons with four siblings in the 2002 cohort was 1 % lower than for daughters
(77.84% versus 89.02%) in 2002. This divergent tendency confirms the fourth hypothesis of
increasing gender inequality between smaller and larger families.
What role do family resources play in this trend? Model-3 describes the cohort-gender
difference in college enrollment by adding in the family resources variable. The female advantage
in college enrollment remained significant across cohorts. This implies that the role of family
resources in explaining the gender gap in college enrollment is limited. For example, with the
inclusion of family resources measures, the female advantage in college enrollment decreased
approximately 24% in the NELS cohort and 48% in the ELS cohort. In short, these findings
indicate that family resources play significant roles in explaining the female advantage in college
enrollment, particularly in the recent cohort. This result also supports the fifth hypothesis.
Trend in Two-Year and Four-Year College Enrollment with Family Resources

16

Table 4 shows the results of the multi-nominal logistic regression for the pattern of two-year
and four-year college enrollment. Beginning with model 1, nets of school percentile ranking,
family SES, family structure and other covariates, the female advantage was present in the 1988
cohort for both two-year and four-year college enrollment. Females in the 1988 cohort were 1.32
times (𝑒𝑒 .279 = 1.32) more likely to enroll in a two year college over no college compared to
females in the 1980 cohort, and almost 1.64 times (𝑒𝑒 .495 = 1.64) more likely to enroll in a

four-year college over no college compared to females in the 1980 cohort. In other words, a
daughter from the 1988 cohort with the same percentile ranking class rank, family SES and other
covariates as a daughter in the 1980 cohort had odds 1.32 times those of the daughter in the 1980
cohort of attending a two-year college, and 1.64 times those of the daughter in the 1980 cohort of
attending a four-year college. Nevertheless, the female advantage was not observed in the 2002
cohort compared to the 1980 cohort, which partially supports the third hypothesis.
Results also showed that males in the 1980 cohort and the 2002 cohort reduced the likelihood
of enrolling in a two-year and four-year college compared to females in the 1980 cohort. Moreover,
step-families and single families were negatively associated with two-year and four-year college
enrollment. As expected, immigrant status, family SES, and students’ educational expectations
were positively associated with two-year college enrollment, except for the effect of mothers’
education. On the contrary, mothers who graduated from college, family SES and students’
educational expectation were positively associated with four-year college enrollment.
The bottom of the table reports the t-test results examining whether there was a statistically
significant difference in two-year and four-year college enrollment across cohorts and whether
there were statistically significant gender differences among cohorts. Importantly, results showed

17

that female-favorable pattern of two-year and four-year college enrollment was present in the 1988
and 2002 cohort.
Model 2 added number of siblings and showed a reduction in the female advantage in both
two-year and four-year college enrollment. Number of siblings was negatively associated with
two-year and four-year college enrollment. The cohort-gender difference decreased after number
of siblings was controlled for in the model. As the results suggest, with the inclusion of number of
siblings, the female advantage coefficient decreased approximately 6% for the two-year college
advantage, and 4% in the four-year college advantage in the 2002 cohort. With respect to the
sibling dilution effect in model 2, number of siblings was negatively associated with both
four-year and two-year college enrollment across cohorts. Students’ four-year college enrollment
was .92 times (e -0.082=.92) and two-year college is .93 times (e-0.07=.93) lower when an additional
sibling was present in a family. In the bottom panel of table 4, the cohort difference among female
members was observed in the 1988 cohort and the cohort difference among male members was
observed in the 1980 cohort and the 2002 cohort compared to the 1988 cohort. Males’ cohort
differences in the decline in two-year and four-year college enrollment remained strong even after
controlling for the effect of number of siblings. The female-favorable pattern of two-year and
four-year college enrollment remained.
In Model 3, variables measuring students’ family resources were introduced. The female
advantage in two-year and four-year college enrollment remained significant in the 1988 cohort.
The Hausman test on the female advantage in two-year versus four-year college enrollment (p
< .05) was statistically significant, which confirms the second hypothesis.
With the inclusion of family resources variables, the female advantage in college enrollment
decreased approximately 20% for two-year college enrollment, and 26% for four-year college

18

enrollment in the 1988 cohort. In contrast, the male disadvantage decreased approximately 4% for
two-year college enrollment and 11% for four-year college enrollment in the 1980 cohort, holding
constant the family resources variables. In short, these findings indicate that family resources
explained the reduction in the female advantage for both two-year and four-year college
enrollment in the 1988 cohort, and that this reduction tendency was more pronounced for four-year
college enrollment than two-year college enrollment. This finding supports the fifth hypothesis. I
also found that most family resources were positively associated with four-year college enrollment,
except for parental involvement in school activities on two-year college enrollment.
Figure 3 shows the predictive margins of two-year college enrollment in model 3 separated by
cohort and gender and conditional on the number of siblings. In general, the female advantage in
two-year college enrollment was significant between the 1980 and the 1988 cohort. But the gender
gap became smaller when the number of sibling was larger than three. For female members, the
probability difference on two-year college enrollment was larger across cohorts, particularly for
smaller numbers of siblings. In contrast, males’ probability of two-year college enrollment was
more condense and smaller over time.
Figure 4 shows the predictive margins of four-year college enrollment for cohort-gender
subgroups at different number of siblings. Notably, female advantage was larger when number of
siblings grew from one to three, but the gender gap became smaller in larger families in the 1988
cohort. However, the divergent tendency between smaller and larger families still held, especially
for the 2002 cohort. Generally, for men, the probability difference on four-year college enrollment
between cohorts was larger across number of siblings. For women, the difference between cohorts
was the highest at families with one and two children, and this tendency converged when the
number of siblings increased. While this study tends to account for potential family resources as

19

much as possible, I also recognize that there are unexplained variations in the female advantage of
two-year and four –year college enrollment in the 1988 cohort.
Discussion and Conclusion
The goal of this study was to understand the effect of family resources on children’s college
enrollment, specifically within the context of a declining number of siblings in families that has
occurred over the past three decades. Utilizing three large nationally representative datasets of
10th graders in 1980, 1990 and 2002, this study aimed to examine the role of number of siblings
and family resources in explaining the female advantage in college enrollment.
The first hypothesis concerned whether the female advantage in college enrollment still held
after accounting for number of siblings and a series of family resources. I found strong evidence to
support that the female advantage in college enrollment increased over time after controlling for
other potential confounders. Students with better family resources had a higher likelihood of
enrolling in college compared to those without the same level of family resources. Results also
confirmed that the number of siblings had a detrimental effect on college enrollment. Female and
male cohort differences were observed, except for the cohort difference between 1988 and 2002.
The gender difference in college enrollment among cohort favored daughters over sons.
The second and third hypotheses concerned whether the female advantage in four-year
college still held and whether the gender gap widened for four-year college compared to two-year
college enrollment, after accounting for number of siblings and a series of family resources
presented in table 3. Results confirmed that the female advantage in four-year college enrollment
still held after controlling for other potential confounders. Through a Hauseman test, I also found
that the female advantage in four-year college enrollment was significantly larger than in two-year

20

college enrollment in the 1988 cohort. Surprisingly, the male disadvantage in four-year college
enrollment was also significantly larger than in two-year college enrollment in the 2002 cohort.
The fourth hypothesis examined whether the gender difference diverged with the number of
siblings over cohorts. Predicted probability states that an increasing trend of gender inequality
between smaller and large families compared the 1980 cohort and the 2002 cohort using the model
3 in table 2. That being said, gender differences in college enrollment were significant when
comparing cohorts, conditional on the number of siblings.
The fifth hypothesis posits that family resources explain the variation in the patterns of
college enrollment across cohort and gender. The evidence for this hypothesis on four-year college
enrollment was fairly strong, particularly for material, socio-psychological, cultural, and social
resources. The positive association between family resources and four-year college enrollment
was strong, despite the fact that parents’ participation in school related activities seemed to have no
effect in predicting two-year college enrollment.
I also found that the role of number of siblings and family resources stratified college
opportunities for daughters and sons in different ways. The results confirmed gender differences
between smaller (one sibling) and large families (four siblings). Females living in a family with
siblings less than two still showed a significant advantage in college enrollment compared to males
with the same number of siblings in the same cohort. This gender gap continued to increase when
the number of siblings increased. However, men almost showed a linear decrease when adding in
any additional siblings at home. This pattern was even more salient for the four-year college
enrollment pattern. This gendered pattern of college enrollment suggests that women are more
resistant to the resource dilution of siblings and may receive more family supports, including
monetary and non-monetary, in sustaining their college enrollment decision.

21

This study has several limitations. First, this study did not account for the selection bias of
parents’ fertility decision, which may affect parents’ ability to raise children and preferences to
invest in daughters over sons. Understanding this potential selection bias requires detailed
information about parents’ attitudes towards obtaining a college degree and family investments in
each of their children (e.g., omitted variables, intra-families variation). Second, due to data
limitations, I did not include some family resource indicators. For example, the measure of family
wealth and parents’ time with their children (Aguiar& Hurst, 2007) may also represent different
dimensions of family investment in children’s college education. I am aware that different
measures of family resources may affect children’s college enrollment differently. Similarly, the
same family resources may have different impacts on daughters’ and sons’ college enrollment
(Eccles& Harold 1993). That is, using different family resource measures or gender-specific
resource measures may alter the findings and conclusions of this study. Third, I operationalize a
number of sibling and family resources as individual characteristics. This strategy is commonly
used in the resource dilution model (Steelman & Powell, 1989, 1993) without accounting for the
intra-family configuration, which may yield different findings (Conley & Glauber 2006) and a
limitation shared by most national representative data without whole family members’ data.
Finally, this study focused on the family-based structural and resource allocation; controlling for
school clustering effects and the effects of specific school contexts is beyond the current scope of
this study.
Despite the limitations above, this study contributes to the existent literature in three ways.
First, substantively, this study confirmed that number of siblings and family resources contribute
to the female advantage in college enrollment. Specifically, I found evidence for the role of
number of siblings and family resources in terms of materials, socio-psychological, cultural, and

22

social resources to explain the changes in the female advantage in college enrollment from 1980 to
2002 while controlling for school achievement, mothers’ education, students’ expectations, and
other covariates. Female advantage in college enrollment was observed across cohorts. Second,
this study identifies the females’ cohort advantage and males’ cohort disadvantage through
two-year and four-year college enrollment. As the results suggest, gender differences in four-year
college enrollment were present in each cohort. In contrast, gender differences in two-year college
enrollment were present only in the 1988 and 2002 cohort. Among those gender differences,
females continued to outpace males in two-year and four-year college enrollment, holding
constant family resources, school achievement, number of siblings, and other covariates. Third,
the female advantage in four-year college enrollment was significantly widened compared to
two-year college enrollment. Importantly, the inclusion of family resources explained more
variation in the female advantage in four-year college than in two-year college enrollment. These
results were consistent with Diprete and Buchmann’s (2013) findings that a gender reversal trend
has shifted in favor daughters toward higher education. This study provided more information
about the gender reversal trend of college enrollment, even after controlling for number of siblings,
mothers’ education, and family SES. Lastly, these results lead to several important questions that
researchers and social demographers need to consider. It is important to better understand why the
transition from high school to college for daughters is smoother than for sons, and the possible
gender bias that may exist through teacher evaluations, low parental expectations, and school
achievement that may reduce sons’ opportunities and motivation to attend college.

23

APPENDICES

24

APPENDIX A

FIGURES FOR CHAPTER ONE

25

.9
.8
.7
.6
.5

College Enrollment Marginal Prediction

College Enrollment by Cohort and Gender conditional on Sibling Size

sib.=0

sib.=1

sib.=2

sib.=3

sib.=4

sib.=5

sib.=6

Number of Siblings

Male HS&B(1980)
Male ELS(2002)
Female NELS(1988)

Male NELS(1988)
Female HS&B(1980)
Female ELS(2002)

Figure 1: Marginal Prediction of College Enrollment by Gender and Cohort within the Number of
Siblings
Source: HSB, NELS, ELS

Predictive Probability by Cohorts and Gender
NELS (1990)

.8
.7
.6

0

1

2

3

4

5

6

.7

.8

.9

ELS (2002)

.6

College Enrollment Probability

.9

HS&B (1980)

0

1

2

3

4

5

6

Number of Sibling at Home
Sons

Daughters

Figure 2: Marginal Prediction of College Enrollment by Gender and Separated Cohort within the
Number of Siblings
Source: HSB, NELS, ELS.

26

.35
.3
.25

Pr(Two year college Prediction)

.4

Adjusted Predictions of cohort by comfemale (Two-year college)

0

1

2
Number of Sibling at Home

HS&B (1980), male
NELS (1990), male
ELS (2002), male

3

4

HS&B (1980), female
NELS (1990), female
ELS (2002), female

Figure 3: College Enrollment Probability in Two-Year College by Gender and Cohort Conditional
on Number of Siblings

.6
.5
.4
.3
.2

Pr(Four Year College Prediction)

.7

Adjusted Predictions of Cohort by Gender (Four-year college)

0

1

2
Number of Sibling at Home

HS&B (1980), male
NELS (1990), male
ELS (2002), male

3

4

HS&B (1980), female
NELS (1990), female
ELS (2002), female

Figure 4: College Enrollment Probability in Four-Year College by Gender and Cohort Conditional
on Number of Siblings

27

APPENDIX B

TABLES FOR CHAPTER ONE

28

Table 1: Variable Description, Mean and Standard Deviation
Variables
College enrollment

Enrolled 2 year college

Enrolled 4 year college

HS cohort GPA
10th grader NaĂŻve
Percentile-Ranking
Student sex
Student White
Student Black
Student Hispanic
Student Asian
Student Native

Description
Whether enrolled in college in
1984 (HS&B); in 1994 (NELS);
in 2006 (ELS)
Whether enrolled in 2-year
college in 1984 (HS&B); in 1994
(NELS); in 2006 (ELS)
Whether enrolled in 4-year
college in 1984 (HS&B); in 1994
(NELS); in 2006 (ELS)
High school overall GPA
composite in 1984 (HS&B); in
1994 (NELS); in 2006 (ELS)
Linear Equity original composite
10th grader percentile-ranking

Family Structure
Number of Siblings
Intact family

Number of Siblings
Living with mother and father
Living with Mother and male
guardian; Father and female
Step-family
guardian
Living with Mother only ; Father
Single family
only
Living with female guardian;
Other Relative/Guardian-family male guardian; or both

29

N

Mean SD

39715

0.68

(.47)

41314

0.20

(.40)

41314

0.43

(.49)

40784

2.70

(.87)

39858

51.21 (20.45)

42339
41968
41968
41968
41968
41968

0.51
0.63
0.12
0.17
0.07
0.01

(.50)
(.48)
(.33)
(.37)
(.25)
(.12)

36893
43166
43166

2.59
0.59
0.11

(1.69)
(.49)
(.32)

43166

0.16

(.37)

43166

0.03

(.17)

Table 1 (cont’d)
Family Resources
Material/Financial Resources

Educational Items

Sum of 4 items: (1) Family has a
daily newspaper, (2) Family has a
computer, (3) Family has more
than 50 books, (4) Has own room

36619

2.89

(.94)

Socio-psychological Resources

How far in school parent want you
to go (HS&B) and (NELS); How
far in school parent wants 10th
Parent educational expectation grader to go (ELS). Scale 1-7.
39874

Student educational
expectation

4.70 (1.65)

How far in school mother want
you to go (HS&B) and (NELS);
How far in school mother wants
10th grader to go (ELS). Scale
1-7.

38445

4.63 (1.64)

(1) How often goes to class
without pencil/paper; (2)How
often goes to class without
books;(3) How often goes to class
without homework done

37118

9.56 (2.08)

(1) how often parents attend
school meetings; (2) how often
parent phoned teacher, counselor;
(3) how often parents attended
school event ; (4) parents acted as
volunteer at r^s school

33920

1.48 (1.32)

Cultural resources

School preparation at home
Social resources

Parent participate in
school-related organization,
activities

30

Table 1 (cont’d)
Mother education
Mother less than high school
Mother high school degree
Mother some college
Mother college graduated and
beyond

38445
38445
38445
38445

0.20
0.32
0.26
0.22

(.40)
(.47)
(.44)
(.42)

Control variables
Region
North east
Midwest
South
West

42828
42828
42828
42828

0.20
0.26
0.34
0.20

(.40)
(.44)
(.47)
(.40)

High school type
Public high school
Catholic high school
Private high school

42760
42760
42760

0.83
0.11
0.06

(.38)
(.31)
(.24)

31

Table 2: Gender Difference in College Enrollment and Family Resources across Three Decades a
HS&B

NELS(1988

ELS(2002

HS&B

Female

College enrollment
Enrolled 2 year college
Enrolled 4 year college
School Achievement
10th grader Relative
High School GPA
Family Structure
Number of Siblings
Intact family
Step-family
Single family
Other
Mother education
Mother less than high
Mother high school
Mother some college
Mother college
Family Resources
Material/Financial
Educational Items (0-4)
Socio-psychological
Parent educational
Student educational
Cultural resources
Parent school
Social resources
Parent participate in
Student White
Student Black
Student Hispanic
Student Asian
Student Native
High school types
Public
Catholic
Private

NELS(1988

ELS(2002

Male

.63*
.27*
.35*

.71*
.28*
.42*

.79*
.30*
.37*

.56
.22
.33

.66
.25
.39

.71
.26
.32

52.72
(20.96
2.39
(.73)

51.02
(23.48)
2.68
(.80)

49.69
(15.95
3.32
(.74)

54.90
(21.63
2.16
(.72)

50.20
(24.66)
2.57
(.81)

49.32
(17.17
3.06
(.79)

2.98
(1.78)
.62
.11
.17
.03

2.57
(1.72)
.57
.09
.12
.02

2.33
(1.56)
.59
.15
.21
.04

2.98
(1.76)
.64
.08
.15
.04

2.40
(1.64)
.58
.08
.10
.02

2.24
(1.49)
.60
.15
.21
.04

.25
.37
.24
.14

.28
.32
.20
.21

.13
.27
.33
.27

.20
.41
.22
.17

.24
.34
.18
.25

.12
.26
.34
.28

2.49
(0.79)

2.78
(0.96)

3.25
(0.85)

2.55
(0.81)

2.89
(0.98)

3.26
(0.87)

4.17
(1.89)
4.13
(1.71)

4.39
(1.43)
4.72
(1.57)

5.48
(1.23)
5.42
(1.33)

4.09
(1.95)
3.98
(1.73)

4.35
(1.41)
4.50
(1.54)

5.29
(1.32)
4.96
(1.49)

9.99
(1.63)
1.40
(1.24)
.60
.14
.21
.03
.02

9.98
(1.64)
1.79
(1.27)
.68
.10
.14
.07
.01

9.65
(2.34)
1.24
(1.34)
.62
.13
.15
.09
.01

9.28
(1.94)
1.47
(1.25)
.58
.14
.23
.03
.02

9.50
(1.83)
1.90
(1.31)
.70
.09
.13
.07
.01

9.06
(2.53)
1.22
(1.34)
.62
.13
.15
.10
.01

0.77
0.20
0.02

0.88
0.06
0.06

0.83
0.08
0.09

0.80
0.16
0.04

0.86
0.07
0.07

0.83
0.08
0.09

Note. Standard deviation (in parentheses)

32

Table 3: Logistic Regression Predicting College Enrollment across Gender and Cohort a
Model-1
Female
Cohort-1980 Female
Cohort-1988 Female
Cohort-2002 Female
Male
Cohort-1980 Male
Cohort-1988 Male
Cohort-2002 Male

[Reference]
0.361***
(0.085)
0.480***
(0.092)
-0.393***
(0.078)
0.009
(0.107)
0.087
(0.086)

Resources dilution
Number of Siblings
School Achievement
Student Relative-percentile rank

Model-2
[Reference]
0.341***
(0.086)
0.444***
(0.092)

Model-3
[Reference]
0.260**
(0.087)
0.229*
(0.099)

-0.389***
-0.352***
(0.079)
(0.081)
-0.018
-0.110
(0.109)
(0.114)
0.038
-0.129
(0.087)
(0.093)
-0.079***
-0.073***
(0.016)
(0.016)

0.030***
(0.002)

Family Resources
Parent educational expectation

0.030***
(0.002)

0.028***
(0.002)
0.221***
(0.019)
0.123**
(0.038)
0.060***
(0.013)
0.094***
(0.023)

Educational Items
Parent school preparation at home
Parent participate in school-related
Family Composition (Ref.= Intact family )
Step-family
-0.377***
(0.081)
Single-family
-0.058
(0.079)
Other Guardians
-0.450**
(0.165)

33

-0.319***
-0.303***
(0.082)
(0.081)
-0.049
-0.026
(0.080)
(0.081)
-0.444**
-0.444**
(0.166)
(0.165)

Table 3 (cont’d)
Mother education (Ref.=Less than High school)
High school mother
Some college mother
College graduated mother
Immigrant status
Family Socioeconomic status (Composite score)
Students' educational expectation

-0.045
(0.077)
0.011
(0.093)
0.278*
(0.120)
0.350*
(0.155)
0.704***
(0.057)
0.474***
(0.021)
Model-1

-0.063
(0.078)
-0.012
(0.094)
0.263*
(0.121)
0.344*
(0.156)
0.685***
(0.058)
0.470***
(0.021)
Model-2

-0.097
(0.079)
-0.064
(0.095)
0.217+
(0.120)
0.340*
(0.156)
0.574***
(0.062)
0.365***
(0.023)
Model-3

Group T-test
Significant female cohort differences
1980 vs. 1988
Yes
Yes
Yes
1980 vs. 2002
Yes
Yes
Yes
1988 vs. 2002
No
No
No
Significant male cohort differences
1980 vs. 1988
Yes
Yes
Yes
1980 vs. 2002
Yes
Yes
Yes
1988 vs. 2002
No
No
No
Significant gender differences
1980 cohort
Female
Female
Female
1988 cohort
Female
Female
Female
2002 cohort
Female
Female
Female
+ p< .1; * p<.0,5; ** p<.01; *** p<.001 ; Standard errors in parentheses.
Note: Reported original logistic regression coefficient.
a
All models adjusted for race group variables, region, high school type. Final analytic sample=
21,553

34

Table 4: Multi-nominal Logistic Regression Model Predicted Two-Year and Four-Year College
Enrollment
ModelModel-1 Model-2 3

Model-1 Model-2 Model-3

Cohort-1988 Female

[Ref.]

[Ref.]

Cohort-1988 Female

0.279** 0.261** 0.209*
(0.093) (0.094) (0.095)
0.102
0.068
-0.124
(0.091) (0.091) (0.097)
-0.407** -0.404** -0.386**
(0.086) (0.086) (0.087)
-0.047 -0.071 -0.133
(0.113) (0.115) (0.119)
-0.381** -0.426** -0.593**
(0.088) (0.089) (0.095)

0.495** 0.473** 0.348**
(0.096) (0.096) (0.097)
-0.057 -0.094 -0.132
(0.099) (0.100) (0.107)
-0.376** -0.372** -0.330**
(0.092) (0.092) (0.094)
0.115
0.086
-0.049
(0.118) (0.120) (0.125)
-0.565** -0.613** -0.852**
(0.096) (0.097) (0.105)

-0.070** -0.065**
(0.017) (0.017)

-0.082** -0.073**
(0.017) (0.018)

Women (Ref. Cohort-1980 Female)

Cohort-2002 Female
Cohort-1980 Male
Cohort-1988 Male
Cohort-2002 Male

[Ref.]

[Ref.]

[Ref.]

[Ref.]

Differential Fertility
Number of Siblings
School Achievement
Student Relative-percentile rank

0.015** 0.015** 0.014*** 0.051** 0.051** 0.049**
(0.002) (0.002) (0.002) (0.002) (0.002) (0.002)

Family Resources
Parent educational expectation

0.158***
(0.019)
0.095*
(0.036)
0.030*
(0.013)
0.033
(0.022)

Educational Items
Parent school preparation at home
Parent participate in school-related

35

0.306**
(0.023)
0.145**
(0.043)
0.071**
(0.015)
0.160**
(0.023)

Table 4 (cont’d)
Family Composition (Ref.=Other Guardians)
Intact family
-0.263** -0.210* -0.207* -0.506** -0.441** -0.396**
(0.081) (0.082) (0.082) (0.086) (0.088) (0.087)
Step-family
-0.052 -0.045 -0.039 -0.093 -0.084 -0.039
(0.079) (0.079) (0.080) (0.084) (0.084) (0.085)
Single family
-0.354* -0.345* -0.349* -0.679** -0.662** -0.645**
(0.168) (0.169) (0.167) (0.183) (0.180) (0.180)
Mother education (Ref.=Less than High
High school mother
-0.032 -0.048 -0.069 0.062
0.043
-0.007
(0.081) (0.082) (0.081) (0.087) (0.087) (0.088)
Some college mother
0.025
0.006 -0.025 0.100
0.077
-0.007
(0.094) (0.095) (0.095) (0.100) (0.101) (0.102)
College graduated mother
0.111
0.098 0.076 0.375** 0.358** 0.286*
(0.121) (0.121) (0.120) (0.125) (0.125) (0.126)
Immigrant status
0.434** 0.429** 0.414** 0.094
0.087
0.096
(0.138) (0.138) (0.138) (0.157) (0.158) (0.157)
Family Socioeconomic status
0.521*** 0.504** 0.428** 0.853*** 0.834*** 0.692***
(0.056) (0.056) (0.060) (0.061) (0.062) (0.066)
Students' educational expectation
0.330*** 0.328** 0.257** 0.609*** 0.605*** 0.473***
(0.022) (0.022) (0.023) (0.023) (0.023) (0.025)
Group T-test
Significant female cohort differences
1980 vs. 1988
yes
yes
yes
yes
yes
yes
1980 vs. 2002
no
no
no
no
no
yes
1988 vs. 2002
yes
yes
yes
yes
yes
yes
Significant male cohort differences
1980 vs. 1988
yes
yes
yes
yes
yes
yes
1980 vs. 2002
no
no
yes
no
yes
yes
1988 vs. 2002
yes
yes
yes
yes
yes
yes
Group T-test
Significant gender differences
1980 cohort
no
no
no
no
no
Female
1988 cohort
Female Female Female Female Female Female
2002 cohort
Female Female Female Female Female Female
+ p< .1; * p<.05; ** p<.01; *** p<.001;
Standard errors in parentheses. Note: Reported original logistic regression coefficient.

36

REFERENCES

37

REFERENCES

Aguiar, M., & Hurst, E. (2007).Measuring trends in leisure: The allocation of time over five
decades.Quarterly Journal of Economics, 122(3), 969-1006.
Andrew, M., &Hause, R. M. (2011).Adoption? Adaption Evaluating the Formation of Educational
Expectation. Social Forces, 90(2), 497-520.
Becker, G. (1981). Alturism in the Family and Selfishness in the Market Place.Economica,
48(189), 1-15.
Becker, G., & Tomes, N. (1976).Child Endowments and the Quantity and Quality of Children.
Journal of Political Economy, 84(4), 143-162.
Behrman, J. R., Pollak, R. A., & Taubman, P. (1989). Family Resources, Family Size, and Access
to Financing for College Education. Journal of Political Economy, 97(2), 398-419.
Behrman, J. R., &Rosenzweig, M. R. (2002). Does Increasing Women’s Schooling Raise the
Schooling of the Next Generation? The American Economic Review, 92(1), 323-334.
Cherlin, A. J. (2010). Demographic Trends in the United States: A Review of Research in the
2000s. Journal of Marriage and Family, 72, 403-419.
Chiappori, P. A., Iyigun, M., & Weiss, Y. (2009).Investment in Schooling and the Marriage Market.
American Economic Review, 99(5), 1689-1713.
Cho, D. (2006). The role of high school performance in explaining women’s rising college
enrollment. Economics of Education Review, 26(4), 450-462.
Conley, D., &Glauber, R. (2006). Parental Educational Investment and Children's Academic Risk:
Estimates of the Impact of Sibship Size and Birth Order from Exogenous Variation in Fertility.
The Journal of Human Resources, 41(4), 722-737. doi: 10.2307/40057288
Crosnoe, R. (2001). Academic Orientation and Parental Involvement in Education during High
School. Sociology of Education, 74(3), 210-230. doi: 10.2307/2673275
Crosnoe, R. (2004). Social Capital and the Interplay of Families and Schools. Journal of Marriage
and Family, 66(2), 267-280.
Degraaf, P. M. (1986). The Impact of Financial and Cultural Resources on Educational-Attainment
in the Netherlands. Sociology of Education, 59(4), 237-246.

38

DiPrete, T. A., &Buchmann, C. (2013). The Rise of Women: The Growing Gender Gap in
Education and What It Means for American Schools. New York: Russell Sage Foundation.
England, P. (2010). The Gender Revolution Uneven and Stalled. Gender & Society, 24(2),
149-166.
Eccles, J. S., & Harold, R. D. (1993).Parent-School Involvement during the Early Adolescent
Years.Teachers College Record, 94(3), 568-187.
Flashman, J. (2013). A Cohort Perspective on Gender Gaps in College Attendance and Completion.
Research in Higher Education, 54(5), 545-570.
Friedkin, N. E., & Thomas, S. L. (1997).Social Positions in Schooling. Sociology of Education,
70(4), 239-255.
Goldin, C., Katz, L. F., &Kuziemko, I. (2006). The Homecoming of American College Women:
The Reversal of the College Gender Gap. The Journal of Economic Perspectives, 20(4),
133-156.
Jacobs, J. A. (1996). Gender Inequality and Higher Education. Annual Review of Sociology, 22,
153-185.
Kalmijn, M. (1994).Mother’s Occupational Status and Children's Schooling. American
Sociological Review, 59(2), 257-275.
Kim, D. H., & Schneider, B. (2005). Social Capital in Action: Alignment of Parental Support in
Adolescents’ Transition to Postsecondary Education. Social Forces, 84(2), 1181-1206.
Lareau, A., & Horvat, E. M. (1999). Moments of Social Inclusion and Exclusion Race, Class, and
Cultural Capital in Family-School Relationships. Sociology of Education, 72(1), 37-53.
Lareau, A., &Weininger, E. B. (2003). Cultural Capital in Educational Research: A Critical
Assessment. Theory and Society, 32(5/6), 567-606.
Lee, J. (2008). Sibling Size and Investment in Children’s Education: An Asian Instrument. Journal
of Population Economics, 21(4), 855-875.
Lee, K. S. (2009). Competition for Resources: A Reexamination of Sibship Composition Models
of Parental Investment. Journal of Marriage and Family, 71(2), 263-277.
Maralani, V. (2013). The Demography of Social Mobility: Black-White Differences in the Process
of Educational Reproduction. American Journal of Sociology, 118(6), 1509-1558. doi: Doi
10.1086/670719
Mare, R., & Chen, M. (1986). Further Evidence on Sibship Size and Educational Stratification.
American Sociological Review, 51(3), 403-412.

39

Mare, R., &Maralani, V. (2006).The Intergenerational Effects of Changes in Women's Educational
Attainments. American Sociological Review, 71(4), 542-564.
McLanahan, S., & Percheski, C. (2008). Family Structure and the Reproduction of Inequalities.
Annual Review of Sociology, 34, 257-276.
NCES. (2014). The Condition of Education 2014 (NCES 2014-083). U.S. Department of
Education.National Center for.Education Statistics. Washington, DC.
Oppenheimer, V. K. (1994). Women’s Rising Employment and the Future of the Family in
Industrial Societies. Population and Development Review, 20(2), 293-342.
Oppenheimer, V. K. (1997). Women’s Employment and the Gain to Marriage: The Specialization
and Trading Model. Annual Review of Sociology, 23(Annual Reviews), 431-453.
Perna, L. W. (2006). Studying College Access and Choice: A Proposed Conceptual Model. In J. C.
Smart (Ed.), HIGHER EDUCATION:Handbook of Theory and Research (Vol. 21, pp.
99-151): Springer.
Powell, B., &Steelman, L. C. (1989). The Liability of Having Brothers: Paying for College and the
Sex Composition of the Family. Sociology of Education, 62(2), 134-147.
Powell, B., &Steelman, L. C. (1993). The Educational Benefits of Being Spaced Out: Sibship
Density and Educational Progress. American Sociological Review, 58(3), 367-381.
Scritchfield, S. A., &Picou, J. S. (1982). The Structure of Significant Other Influence on Status
Aspirations: Black-White Variations. Sociology of Education, 55(1), 22-30.
Schneider, B., & Coleman, J. S. (1993).Parents, Their Children, and Schools. Boulder, CO:
Westview Press.
Steelman, L. C., & Powell, B. (1989).Acquiring Capital for College: The Constraints of Family
Configuration. American Sociological Review, 54(5), 844-855.
Steelman, L. C., & Powell, B. (1991).Sponsoring the Next Generation: Parental Willingness to Pay
for Higher Education. American Journal of Sociology, 96(6), 1505-1529.
Suitor, J. J., Plikuhn, M., Gilligan, M., & Powers, R. S. (2008). Unforeseen Consequences of
Mothers’ Return to School: Children’s Educational Aspirations and Outcomes. Sociological
Perspectives, 51(3), 495-513.
Sweeney, M. M. (2002). Two Decades of Family Change: The Shifting Economic Foundations of
Marriage. American Sociological Review, 67(1), 132-147.

40

Sweeney, M. M. (2011). Whither Opportunity? Rising Inequality, Schools, and Children’s Life
Chances: Russell Sage Foundation.
Sweeney, M. M., &Raley, R. K. (2014).Race, Ethnicity, and the Changing Context of
Childbearing in the United States.Annual Review of Sociology, 40, 539-558.
Sewell, W. H., Hauser, R. M., Springer, K. W., & Hauser, T. S. (2003). As We age: A Review of the
Wisconsin Longitudinal Study, 1957-2001. Research in Social Stratification and Mobility,
20(0), 3-111.
Vogl, T. S. (2016). Differential Fertility, Human Capital, and Development. Review of Economic
Studies, 83(1), 365-401.

41

CHAPTER TWO
SIBLING SIZE, GENDER COMPOSITION, AND PARENT INVESTMENT IN COLLEGE
ENROLLMENT

Introduction
Research indicates that the female advantage in postsecondary enrollment and completion
has increased substantially from the 1980s through today (DiPrete & Buchmann, 2013; Buchmann,
DiPrete,& McDaniel, 2008). This change has been in part attributed to decreasing family size and
increases in women’s labor force participation which has resulted in directing more educational
resources to children’s schooling (Torche, 2011; Dufur, Parcel, Hoffmann, &Braudt,2016). What
has received less attention is the potential importance of structural changes in sibling configuration
and its impact on family supports for college. For example, in the 1980 cohort of the U.S. national
longitudinal High School and Beyond study, 23.1 % of boys and 23.3 % of girls lived in
single-child and two-children families. In comparison, 36.9 % of boys and 35.6 % of girls lived in
single-child and two-child families in the 2002 recent cohort of the national Educational
Longitudinal Study (Chen, Schneider, & Frank, 2015). The probability of youth 18 years old and
younger living with only one sibling increased approximately by 12 % from 1980 to 2002.
Estimates of the gender disparity in college enrollment between small and large families but
overlook sibship composition, that is, whether the family has a son and a daughter or two
daughters are likely to miss important differences in the educational expectations and resources
parents are willing to allocate to their daughters compared to their sons (Bissell-Havran, Loken, &
McHale, 2012). This structural heterogeneity among families coincides with decreases in family
size over the last thirty years (Eggebeen, 1992; Mare & Maralani, 2006). By separating out the
effect of sibling size from sibship gender–position we can gain a deeper understanding of what

42

factors are directly shaping the female advantage in college enrollment and how they have been
changing over time. This study examines the differential effects of sibling size, sibship-gender
composition, and family educational resources allocations on female college enrollment using the
National Educational Longitudinal Study 1988-1994 (NELS 88)and the National Longitudinal
Study of Adolescent Health, 1994-2008 (Add Health).
Background
Resources Dilution Model: Gender-Specific Dilution
The resource dilution model—that is diminished family resources allocation to their children
depending on number of siblings (named sibling size)—introduced by sociologists and
demographers (Anastasi 1956; Blake, 1989, 1981; Downey, 1995, 2001) offers the most widely
accepted explanation for the relationship between sibling size, family resources, and children’s
educational opportunities. While having an additional child at home has a negative effect on
family resource allocation and children’s educational opportunities, the evidence for sibling
gender composition on children’s educational outcomes is mixed and related to gender asymmetry
within the family (Chu, Xie, & Yu 2007; Conley &Glauber, 2006; Lee, 2008). For example,
parents’ perceptions of educational returns or their personal experiences with the expected
gendered roles of boys versus girls may guide resource allocations to favor one child over another.
Previous research by Steelman and Powell (1989) found that the presence of an additional brother
reduced parents’ financial contribution to their daughters’ schooling. Most studies on sibling size
and family resources through the 1990s also showed that gender-specific resource dilution
depended on the presence of larger families and sibling composition (Yamaguchi & Ferguson,
1995).

43

Research also has shown that parents are inclined to have an additional child to achieve a
balanced gender boy/girl composition. Conley and colleagues (2006) used 1990 census data to
examine the causal effect of increasing sibling size on educational outcomes while accounting for
parents’ boy/girl composition preferences. They concluded that sibling size has a more consistent
dilution effect on sons in families with more than two children (where the eldest two children are
of the same gender). Similar effects were found by Kalmijn and van de Werfhouse (2016). Given
that family size continues to decrease (Census 2010; Hobbs &Stoops, 2002), we expect to find
greater resource dilution and gender asymmetric effects in earlier family cohorts than today. We
propose the following hypotheses:
Hypothesis1: The cohort trend of smaller families results in the decline gender-specific
dilution on college enrollment (The narrowing dilution hypothesis).
Sibship composition and Gender Socialization
The concept of sibling gender composition originates from a child’s gender position relative
to other siblings in the home. Previous research on sibling gender composition has focused on the
influence of the number and balance of sisters and brothers within the family on a sibling’s
educational outcomes (Kuo & Hauser, 1995; Butcher & Case, 1994).Recent studies have shown
differential parent expectations, household chores, and resource allocations for similar and
differential gendered sibling dyads (Marks, Lam, & McHale, 2009; McHale & Crouter, 2003a;
Brody & Steelman, 1985) in the household. For example, McHale and Crouter (2003b) found that
in older-brother-younger-sister dyads, younger girls did more housework than their older brothers.
Likewise, in older-sister-younger-brother dyads, older sisters performed more housework than
their younger brothers. These patterns do not appear to be consistent when there are the same
female sibling configurations. Recently, Mark and colleagues (2009) found that girl-girl sibling

44

dyads rather than boy-boy or mixed gender sibling dyads are more likely to be characterized as
having more egalitarian gender role attitudes. We suspect that educational expectations and
resource allocations are likely to depend on the gender composition of the family, such as girl-girl
dyads and gender position, older girl and younger brother.
To capture this concept, we constructed three gendered sibship composition patterns (e.g.,
majority, minority and balanced, see Appendix A). Each position in the household sibship
constellation corresponds to a child’s gender, siblings’ gender, and sibling size. Based on this
constellation, we measure the relationship of the proportion of same-sex siblings at home with
parent attitudes and resource allocations. Given that family size has decreased we expect that
daughters may gain more in the same-gender majority position in a home-based environment,
which in turn affect their educational opportunities.
Hypothesis 2: The trend toward smaller familieshas increased the importance of
sibshipgender- majority position on family expectations, resource allocations and four year
college enrollment (Sibship-gender majority hypothesis).
Differential Parental Resource and Cultural Supports
In addition to sibling composition, we also recognize that there are likely to be other
competing factors in the household, such as their own gendered experiences which are likely to
affect parent attitudes toward their children and their resource allocations. Parent attention and
educational college expectations have been found to be different between sons and daughters in
the same household. Bissell-Havran (2012) found that sisters reported their mothers have higher
educational expectations for them compared to their brothers in the same household. Carter and
Wojtkiewicz (2000) also found higher parental involvement, educational plans, and school
discussions with their daughters in contrast to their sons’. On the other hand, in two children and

45

mixed-sex sibling families, brothers reported higher father involvement compared to their sisters
in the same household. With respect to education and career choices, Bleeker and Jacobs (2004)
found that daughters’ pursuits of science career in engineering and other hard sciences were
positively related to their mothers’ beliefs and perceptions of their math and science achievement
and career opportunities. This effect was not found for sons. However, parental involvement in the
boy-boy families reveals the largest gender stereotype expectations compared to boy-girl and
girl-girl sibling family compositions (Marks, Lam, & McHale, 2009).
In addition to differences in educational expectations, role models, and involvement in school
activities, research has also shown that parents differentially allocate resources, such as parental
granting autonomy, lessons for art and music, summer camps, computers and other electronic
equipment between their sons and daughters. These differential investments are related in part to
parent perceptions of their children’s academic and athletic abilities. The increasing competitive
college landscape is in part pressuring parents to make major investments in their children’s
education experiences both in and out of the home. It is not that parents did not make some of these
choices in the past, we suspect that what has changed is that there may be differences in resource
allocations between families of sons and families of daughters. Parents may be willing to allocate
even more to their daughters to ease their path to a four-year college.
Hypothesis 3: Parents’ today are likely to be more responsive to allocating resources to
their daughters’ education compared to their sons, taking into account the academic ability of
their daughters and sons(Cultural resources hypothesis).

46

Data and Methods
Data
Data for this study were derived from two national longitudinal surveys, the National
Educational Longitudinal Study (NELS: 88) conducted from 1988-1994 and the National
Longitudinal Study of Adolescent Health (ADD Health) conducted from 1995-2008. NELS88
began with 8th graders (aged 14) and included subsequent follow-up surveys with students, their
parents, and schools. Among the original sample, 9,202 students had valid postsecondary
outcomes from the last follow-up. Applying listwise deletion for key covariates, such as sibling
size, family living arrangement, mother education, school performance, types of high school and
family resources; the final analytic sample included 6,364 adolescents. Analyses were conducted
to ensure that the final analytic sample was not biased and representative of the third follow-up
cohort.
ADD Health surveyed high school students who were in 7th through 12th grade from
1994 –1995. Four waves of data were collected in 1995, 1996, 2002, and 2008. We used these four
waves of data and generated a variable to represent whether a student enrolled in a two-year or
four-year college by age 20. We used the high school transcript data to represent students’ level of
mathematics achievement. Listwise deletion for key variables was also conducted, which reduced
the final analytic sample to 5,615 adolescents. The NELS cohort included an oversample of
Hispanic and Asian families, and the ADD Health cohort also included an oversample of Cuban,
Puerto Rican, Chinese, and high SES Black families

47

Measures of Dependent Variables
College Enrollment. The outcome for both surveys was college enrollment two years after
high school graduation. Students are classified into three groups those who were: (a) not enrolled
in postsecondary school (b) enrolled in a two-year college, and (c) enrolled in a four -year college.
Measures of Independent Variables
Sibship-gender Composition. In the NELS data, respondents were asked about the number of
older and younger sisters and brothers at home. We also compared the same question in the parent
survey to with the student one to ensure the accuracy of sibling configuration and their gender
position. In the ADD Health data, respondents were asked to provide information on each member
who regularly resided in the home with their relationship, gender, and age to the sampled student.
We used the first wave of the surveys to construct the sibling size and gender composition
variables.
To examine the effect of a sibship-gender majority position on female advantage in college
enrollment, we first configured the sibling-sex composition with both size and gender information
into twenty-eight possible patterns. Second, we classified sibship-gender position into three
categories: (1) the sibship-gender minority; (2) the sibship-gender balanced, and, (3) the
sibship-gender majority (See the Table 10 for details). In the analysis, we used sibship-gender
minority as the reference group.
Family Resources. Family resources refer to parent expectations, investments and
involvement in their children’s education. We focus on six of these including: (1) extracurricular
activities organized by parents and by schools; (2) social capital—knowing the parents of the
students’ friends (Coleman 1990); (3) parent-child discussion and parent-child relationship; (4)
independence—allowing students with autonomy for some decisions; (5) involvement in

48

monitoring and supervision of student activities such as homework; and (6) education and career
expectations parents had for their students. Variable definitions and sample statistics for all these
resource variables in the two datasets are listed in the Table 11.
Other Covariates. We measured 10th grade school performance using standardized test scores
in mathematics and reading from the NELS data. For the ADD Health data, we recoded the Picture
Vocabulary Test score into a relative percentile rank to represent students’ vocabulary performance
at wave 3. For mathematics performance in the ADD Health data, we used the highest mathematics
credit that students received in wave 3. We constructed this math level variable as follow 0= No
Math; 1= Remedial Math; 2 = General Math; 3 = Pre-Algebra; 4 = Algebra I; 5 = Geometry; 6 =
Algebra II; 7 = Advanced Math including Algebra III, Statistics and Probability, and Discrete
Math; 8 = Pre-Calculus; 9 = Calculus).We also examined school GPA as an indicator of school
performance, which increased the analytic sample size and the final results remain the same as
math performance. Thus, we reported the estimation of the modeling results based on school
performance in vocabulary and the highest math performance at wave 3.
Family living arrangement distinguished between four types of family arrangements in both
the NELS and ADD health data: (1) intact family: living with both biological parents, (2)
step-family: living with biological mother and male guardian or biological father and female
guardian, (3) single family: living in mother only or father only families, (4) other
guardians/relatives: living with relatives or other non-biological guardians. Family income in
NELS was scaled from 0-13 in the first follow-up; in ADD Health, the total household income was
scaled from 0-4 in wave one. Race and ethnicity categorized students’ backgrounds into five
groups, not Hispanic white, black, Hispanic, Asian/pacific-islander, and others. The Hispanic
category included all students of Hispanic origin regardless of whether they reported being

49

multi-racial or white. Mother’s education was categorized into five groups using the parent survey,
ranging from 1 = less than high school to 5 = college graduated and beyond. If this variable was
missing or the mother was not surveyed, we substituted using adolescents’ reports of the education
level of residential mother. High school type was operationalized as two dummy variables as
private or Catholic, with public school as the referent group. Region was operationalized as three
dummy variables, with South as the referent group.
Analytic Methods
To answer if females were more likely than males given their sibling size, sibship
composition, and family resources were more likely to attend a four-year postsecondary school,
we employed a multi-nominal logistic model (MNL) where the dependent variable was one of
three destinations after high school graduation. Our inferences about the possible effect of sibship
size on two-year and four-year college enrollment were based on an observational survey, where
the assignment of children living in a small (one or two children families) versus a large family
was not a random event. Our first step was to conduct a set of descriptive analysis to determine the
distribution of sibling size, sibship gender composition and family resources across cohorts. We
then used the Heckman selection approach to adjust for nonrandom event of small families in each
cohort (Heckman, 1979; Winship and Mare 1992). The two-step approach began with a probit
model predicting the probability of an adolescent living in a small family. Variables used to predict
living in a small family included immigrant status, mothers’ age, religion, parent education, parent
marital status, household income, race, and region. All measures were based on the first follow-up
data in NELS and first wave data in ADD Health.
Following this correction, estimates for two-year and four-year college enrollment were
interpreted as an adjustment to the sample selection of small families. The second stage included

50

the predicated probabilities from the first stage to estimate the likelihood of two-year and four-year
college attendance in the multi-nominal logistic model. The second stage equation was as follows:

Pr (yi=1| xi) = Pi1=
Pr (yi=2| xi) = Pi2=
Pr (yi=3| xi) = Pi3=

1

1+𝑒𝑒𝑒𝑒𝑒𝑒 (𝑥𝑥 𝑖𝑖 𝛽𝛽 2 )+𝑒𝑒𝑒𝑒𝑒𝑒 ⁡
(𝑥𝑥 𝑖𝑖 𝛽𝛽 3 ⁡
)
𝑒𝑒𝑒𝑒𝑒𝑒 ⁡
(𝑥𝑥 𝑖𝑖 𝛽𝛽 2 )

1+𝑒𝑒𝑒𝑒𝑒𝑒 (𝑥𝑥 𝑖𝑖 𝛽𝛽 2 )+𝑒𝑒𝑒𝑒𝑒𝑒 ⁡
(𝑥𝑥 𝑖𝑖 𝛽𝛽 3 ⁡
)
𝑒𝑒𝑒𝑒𝑒𝑒 ⁡
(𝑥𝑥 𝑖𝑖 𝛽𝛽 3 )

1+𝑒𝑒𝑒𝑒𝑒𝑒 (𝑥𝑥 𝑖𝑖 𝛽𝛽 2 )+𝑒𝑒𝑒𝑒𝑒𝑒 ⁡
(𝑥𝑥 𝑖𝑖 𝛽𝛽 3 ⁡
)

𝛽𝛽 2 and 𝛽𝛽 3 represent the effects of two-year and four-year college enrollment with no

postsecondary school enrollment as the referent group. The equation forPi1 was derived from the
constraint that the three probabilities sum to 1 (Powers and Xie, 2000). We then implemented a
model where family resource variables were not included for the purpose of establishing the
auxiliary relationship between number of sibling and female advantage in college enrollment. The
second model further accounted for all family resources to investigate whether the effect of female
advantage in college enrollment could be mediated. We also estimated the model with and without
the inclusion of the Heckman sample control and found that the latter fit the data better based on
the Bayesian information criterion (BIC), and Likelihood ratio test (coefficients without selection
control are available upon request). To further account for the heterogeneity issue, we also ran
additional analyses using generalized ordered logitmodel(Williams, 2009) to check whether the
sibling characteristics were heterogeneous across gender groups (e.g., the number of sibling on
college enrollment may be greater for sons over daughters). In so doing, we confirmed our main
findings still hold and present more informative findings using multi-nominal logistic model
format.

51

Results
Descriptive Results
Table 5 shows the distribution and the means of the focal variables, by survey and gender.
Within each of the two survey cohorts, the trend of female advantage in college enrollment is fairly
consistent. Daughters are more likely to enroll in college compared to sons. There is a significant
decline in number of siblings (p< .05) in all types of families. Sons have a greater chance of living
in sibship gender-majority families whereas daughters have a higher likelihood of living in sibship
gender-minority families. This tendency is consistent across cohorts. Adolescents in the recent
cohort have a greater chance of living in the sibling gender-majority (p< .05) and gender-balance
families (p< .05) in the ADD Health cohort. Cohort changes also showed a positive increase in step
families, private high school attendance, mothers with high school, and college degree.
The mean family resources by sibship gender composition are presented in Table 6. In the
NELS cohort, there is a notable gender difference in parental supervision and parents’ organized
extracurricular activities across the three categories of sibship composition, with families of
daughters reporting higher activities and parental supervision than families of sons. As expected,
parents had higher educational expectation for their daughters than sons in the sibship-gender
majority position. Within each sample of girls and boys, mean differences across positions are
fairly consistent with respect to a better resource allocation in the sibship gender-balanced position.
There are other allocation differences across positions.
One key difference is the gender resources gap between majority position and minority
position. For example, daughters obtain a significantly higher degree of cultural related resources
(p<.001), parent-child discussion (p<.01) and educational expectation (p<.01) in the majority
position, but we find that the absence of a similar tendency for sons in the same position. The

52

resource gap between gender-balanced and gender-minority position shows a similar pattern as
better resource allocation is noted for gender-balanced positions.
In the ADD Health cohort, there is a notable gender difference in cultural-related
extracurricular activities in school and parent respect child’s autonomy across the three categories
of sibship composition. Interestingly, sons receive more autonomy from parents than daughters. In
the sibship-gender majority position, daughters have greater cultural activities (p<.01), parental
closure network (p<.01) and supervision (p<.01) compared to sons. However, daughters in the
sibship minority position have significantly lower parental expectations (p<.05) compared to sons.
Within each subsample of girls and boys, the ADD Health cohort shows the same consistent
pattern as the NELS cohort, with resource allocations in the sibship gender-balanced position
being relatively better than other positions regardless of a student’s gender.
Multi-nominal Regression Results
Our multivariate analysis consists of two steps. The first analysis investigates the effects of
sibling size and family resources in explaining the female advantage in college enrollment by
cohorts. The second investigates the effect of sibship gender position and whether these effects
vary between females and males by cohort. Beginning with Model 1 in the NELS cohort, shown in
Table 7, panel A results indicate a significant female advantage in college enrollment, net of
number of siblings, family SES, and demographic characteristics. Females were 1.39 times
(e0.33=1.39) more likely to enroll in a four year college over no college compared to males and
almost 1.36 times (e0.31=1.36) more likely to enroll in a two-year college over no college.
Comparing the NELS results with the ADD Health shown in Table 7 panel B the pattern of the
female advantage in a four-year college is similar, yet no female advantage was observed in
two-year college enrollment.

53

With respect to the sibling dilution effect, number of siblings was negatively associated with
both four-year and two-year college enrollment across cohorts. The NELS students’ four-year
college reduced 1.1% predicted probability with an additional sibling presence at home and the
NELS students’ two-year college also reduced 0.8% predicted probability with an additional
sibling presence in the family. Likewise, with one sibling presence at home, the ADD Health
students’ four-year college enrollment was associated with 1.2% reduction of predicted probability
and 0.9% reduction of predicted probability in a two-year college enrollment. For example, males
with one sibling had a predicted probability of 50% for attending a four-year college; a predicted
probability of 46% for attending a four-year college if he had three siblings at home (these models
also included the full set of covariates).
Model 2 added the family resource variables and showed that a decline in the gender
significance of 31% in the four-year college and 20% in two-year college enrollment. Five
measures of the family resources variables are significant predictors of four-year college
enrollment, which included cultural-related extracurricular activities, parent-child discussing
school things, parents’ social capital resources, monitoring and supervision, and educational
expectations. Importantly, our results showed that a great reduction of female’s likelihood of four
in four-year college enrollment than for two-year college, suggesting which suggested that the
female advantage in four-year college was mostly derived from the family resources in the NELS
cohort. Correspondingly, in Model 2 in the ADD Health cohort, our results shown a similar
declining pattern of the female advantage in the four-year college enrollment by 22 %, however,
females still retained advantage over males four-year college enrollment. Four measures of family
resources variables are significant predicted four-year college enrollment, which included

54

parent-child relationships, parents’ social resources, parental expectations and parent respect
children’s autonomy.
The second step of the analysis investigated the effect of sibling size and sibship gender
position on college enrollment shown in the NELS cohort in Table 8. As see in model 1, net of
family characteristics and family resources, we found no direct effect of sibship gender-majority
advantage in college enrollment. The next two columns estimated the gender-specific family
context within respect to the number of siblings and sibship gender composition. The male
-specific sibling dilution effect was present in both two-year and four-year college enrollment in
Model 2, suggesting. It suggested that the effect of sibling dilution on college enrollment depended
to some extent on a student’s child’s gender. To cite an example, the results indicated that sibling
dilution effect was more severe for males than for females. We calculated the probability of
four-year college enrollment for males who is average of the other covariates, varying only the
number of siblings. Males with one sibling had a predicted probability of 50% for attending a
four-year college; and males had a predicted probability of 46% for attending a four-year college if
he had having three siblings at home (these models also included the full set of covariates). Results
suggested that males’ college opportunities reduced when having more siblings presented at home.
In contrast, our results showed that females’ four-year college enrollment was 1.52 times
higher than males when in the sibship-gender majority position. This beneficial effect however,
was observed only for four-year college enrollment. We also calculated the probability of four-year
college enrollment for females who is average of the other covariates, varying only the sibling
gender position. Females with a gender majority position had a predicted probability of 60% for
attending a four-year college compared to 51% for females and 54% for males in other sibship

55

positions; suggesting that . Our results imply that the gender-specific sibship context tended to
favor daughters in the sibship gender majority positions tend to favor daughters.
Up to this point, we found that both resource dilution and gender majority benefits were
present in four-year college enrollments. That being said, males’ four year college opportunities
are particularly sensitive to the number of siblings at home whereas and females’ four-year college
opportunities are responsive to their sibling gender composition. Perhaps those daughters who
were singular children and those daughters who single child have more sisters and having sisters
more than brothers at home tend to have greater opportunities to attend a four-year college. In the
NELS cohort, we observed that gender-specific family contexts in terms of sibling dilution and
gender majority socialization were working in determining children’s college enrollment.
This patter also occurs with the ADD Health cohort however the effect is less. As shown in
When come to the ADD Health cohort in Table 9, we found a slightly gender-majority advantage
in four-year college enrollment (p<.1). Model 1 suggests, which suggested that students in sibship
majority positions gained a greater opportunity to attend college compared to other sibship
positions.
The next two columns estimated the gender-specific family context. In Model 2, we found
that sibling dilution effect depended less on gender but diluted affected the students’ likelihood of
attending a two-year college. The pattern of male-specific sibling dilution in a four-year college
was declining from the NELS cohort to ADD health cohort. We found the evidence of the decline
gender specific dilution in males’ four-year college enrollment and no cohort difference on the
declining gender-specific dilution has been suggested for two-year college enrollment. This
finding supports the narrowing dilution hypothesis for four- year college enrollment. This may
reflect the increased importance of smaller families rather than gender for increasing four year

56

college enrollment for both males and females to attend four-year college in the ADD Health
cohort.
In Model 3, we only found a gender-specific beneficial context for the toward sibship gender
majority position and this only occurs for two year college enrollment. However, this effect
operated only in two-year college enrollment, net of family resources allocation and the other
covariates in the model. When calculated the probability of two-year college enrollment for
females who is average of the other covariates, varying only the sibling gender position, females in
a with gender majority position had a predicted probability of 42% for attending a two-year college
compared to 38% for of females in with other positions to attend two-year college.
Results also indicated the increased importance of being in a gender majority position for
attending a four-year college for the ADD health cohort. Students in the sibship gender majority
position are almost 1.2 times more likely to attend a four-year college than their peers who had
greater numbers of opposite-sex siblings at home. However, we observed a significant declining
pattern of the daughter-specific majority effect across cohorts in choosing a four-year college but
this pattern was not present for two-year college enrollment. The daughter-specific majority
position did not operate in a consistent manner but this effect overtook the evidence of sibship
gender majority benefit in the ADD Health cohort with other controls in the model.
To better understand how female advantage in college enrollment depends on family sibship
gender composition and family resource allocations, we ran the full model separated by the three
gender composition contexts. The female advantage in college enrollment was present only in the
sibship gender majority composition controlling for number of siblings. Students’ four-year
college enrollment in the NLES cohort was more responsive to parent-child discussion, parents’
social capital, monitoring, and educational expectations. We also found a sibship-gender majority

57

advantage in the ADD Health cohort particularly for four-year college enrollment. Students’ four
year college enrollment in the ADD Health was also responsive to parents’ social resources, child
autonomy, and educational expectations.
Interestingly, students living in the sibship gender balance families, with greater family
resources, had no female advantage on their college enrollment. Parents’ interactions with their
children may outweigh other factors in mixed gendered families which resulted in a more
egalitarian attitude and resource allocations for daughters and sons to attend four year college.
Conclusion and Discussion
Our analysis of two national longitudinal studies yields three findings. First, we confirmed
resources dilution hypothesis across cohorts and observed the present and decline of the
male-specific resources dilution from the NELS cohort to the ADD Health cohort. This male
“disadvantage” home-based environment is derived from the number of siblings and its negative
association with resource allocations. This result, especially for sons in the NELS cohort, is
consistent with results using twin data and sibling fixed effect model in Sweden (Amin, Petter, &
Dan-Olof, 2011) and Norway (Black, Devereux, & Salvanes, 2005; Ermisch & Paronzato 2009).
Second, female advantage in college enrollment still holds across cohorts except for two-year
college enrollment in the ADD Health cohort. We speculated that structural changes of sibship
configuration might contribute to a daughters’ advantage in college enrollment when she has few
siblings and situating in the gender majority position, particularly for later cohorts. Our results
support this proposition but only for daughters with sibship gender majority position for college
enrollment. This “friendly” home-based environment enhances daughters’ four-year college
enrollment in the NELS cohort and two-year college enrollment in the ADD Health cohort.

58

We also find evidence of a decline in the importance of gender–specific dilution and
socialization processes, particularly for four-year college enrollment. The gendered sibship
majority socialization persists only in two-year college enrollment in the ADD Health cohort.
Such changes point to a distinctive family advantage for students in the ADD Health cohort —
either being a sibship gender majority position or greater female-friendly family with shared
gender egalitarian socialization. More research should be undertaken to identify the influence of
adolescents’ experiences and development that may contribute to college-going mindset and
pre-college preparation as well as the influence of their parents’ education and labor market
experiences. It may be that with the increasing financial opportunities to attend two-year colleges
that the gendered majority position for two-year college enrollment may be seen as an economical
and efficient pathway especially for females as they are more likely to study hard, complete their
degrees, and transfer to four-year institutions if they desire.
Third, descriptive results confirm that family resource allocations varied by gender and
sibship gender composition. This emerging influence of sibling gendered environment and
differential resources allocation were largely confirmed in the multi-nominal models. Our analyses
supported descriptive evidence of daughters-favorable resources allocation in cultural-related,
socio-psychological and parents’ supervision. The empirical evidence further suggested that
greater resources promote two-year and four-year college enrollment and this tendency was
consistent across cohorts. Surprisingly, our results suggested that better resource allocations
toward the daughters and families with gender mixed siblings. However, our results and
multi-nominal models further confirmed that better resources allocation is not necessary to have a
female advantage in college enrollment. Result show that the female advantage in college is
observed only with the daughters who have more sisters at home. Results seem to indicate the

59

female sibship-gender majority socialization does not necessarily combine with the greatest family
resource allocation. Daughters are particularly sensitive to family socialization with moderate
family resources, which subsequently influences their college enrollment. Daughters’
female-friendly family environments may provide supports for parents or daughters to develop
more gender egalitarian attitude toward their higher education decisions.
Current results also point to potential disadvantage of daughters who are in the minority
gender positions and of sons who are in larger families or are exposed to lower level of cultural and
socio-psychological resources at home. Overall, family resource allocation appears to be
particularly important for high school students in the transition into college. The inclusion of
family resources explains 31% of the female advantage in four-year college enrollment in the
NELS cohort and 22% of the female advantage in four-year college enrollment in the ADD Health
cohort. It is worth noting that the inclusion of family resources nearly explains the total female
advantage in four-year college enrollment in the NELS cohort (p<.1) controlling for family
resources), but similar resources did not explain the female advantage in the ADD Health cohort
suggesting a need for further examination of unobserved variables in the later cohort. In the NELS
cohort, family resources explain a greater proportion of the female advantage in four-year college
than two-year college (31% vs. 20%), which is consistent with the previous research that four-year
college is largely dependent on family socio-economic status and resource availability (Bailey,
Jaggars, & Jenkins, 2015; Ma & Baum, 2016). These results highlight the need to examine the
differential resource allocation patterns (Dumais, 2002; Fasang, Mangino, & Bruckner, 2014; Kim
& Schneider, 2005; McHale, Crouter, & Whiteman, 2003a) that contribute to home advantage
toward daughters versus sons in high school achievement, college preparation, and college identity
especially given the trend of smaller families.

60

Our findings point to a departure for future research. It bridges conventional educational
inequality studies that have failed to examine sibship composition and family resources in
explaining the female advantage in college enrollment. With smaller families and less emphasis on
gendered balanced families, females are gaining a new resource opportunity. Mothers, especially
those who have entered the labor force may be particularly sensitive to ensuring their daughters’
future education and career successes and underscoring the importance of four year college
enrollment.
Yet there are some limitations to this work. Although we identified the significance of family
gender-specific context across cohorts, our multivariate analysis were restricted to comparable
variables with similar questions on the two surveys and we may have missed some key measures in
our models. Second, some family transitions, such as divorce or remarriage, may have occurred
during the intervals within each of the surveys. We did however generate a dummy variable to
capture the impact of these family transitions, our main findings still held across the cohorts. Third,
this study focused on the effect of smaller number of siblings and gender composition on college
enrollment, we do not discuss how the covariates in terms of family structure, income, and racial
variation may change over time or any potential interactions among the covariates.
But even with these limitations one of the important values of this study is replicating the
basic models with two different datasets, one of which that was started slightly earlier than the
other. With two datasets it is possible to gain a deeper understanding of the general patterns on the
female advantage in college enrollment based on up to date information about sibling gender
composition by cohort comparative approach. Both datasets point to of the decline of the dilution
model along with reductions in family size and casts a new perspective on accounting for sibship
composition and multiple measures of family resource allocations. The female college advantage

61

is complex and family not just schooling dependent. Being a good student is not quite enough
without a family that expects college attendance and allocates education resources beyond family
income to make that a reality.

62

APPENDICES

63

APPENDIX A

FIGURES FOR CHAPTER TWO

64

60.00%
a: Majority VS. Minority
b: Minority VS.Balance
c: Majority VS.Balance

Majority
50.00%

Minority
Balance

40.00%
30.00%
20.00%
10.00%
0.00%
NELS 2-year (c,)

NELS 4-year (a,b)

NELS 2-year

NELS 4-year (b, c)

Figure 5: College Enrollment by Three Context of sibship Gender Composition in NELS cohort
Note. “a” indicates significant mean difference between majority and minority position; “b” indicates
significant mean difference between minority and balance position; “c” indicates significant mean
difference between majority and balance position. Statistical significance sets at level of 0.5.
Source: the National Educational Longitudinal Study 1988-1994 (NELS 88)

60.00%
50.00%

Majority

a: Majority VS. Minority
b: Minority VS.Balance
c: Majority VS.Balance

Minority
Balance

40.00%
30.00%
20.00%
10.00%
0.00%
ADD 2-year

ADD 4-year (a)

ADD 2-year

ADD 4-year

-10.00%
Girls

Boys

Figure 6: College Enrollment by Three Context of sibship Gender Composition in ADD Health
cohort
Note. “a” indicates significant mean difference between majority and minority position; “b” indicates
significant mean difference between minority and balance position; “c” indicates significant mean
difference between majority and balance position. Statistical significance sets at level of 0.5.
Source: the National Longitudinal Study of Adolescent Health, 1994-2008 (Add Health)

65

Estimation of College Enrollment in NELS, by Gender and Sibship Positions

.2 .4 .6 .8 1

(2) Sibship Gender Minority Position

0

1

2

3

4

(3) Sibship Gender Balance Position

.2 .4 .6 .8 1

Probability of 4-year College Enrollment

(1) Sibship Gender Majority Position

0

1

2

3

4

Number of siblings
Sons

Daughters

Figure 7: Four-Year College Enrollment Probability by Sibship Gender Composition in the NELS
Cohort

Estimation of College Enrollment in ADD Health Cohort by Gender, Sibship Positions

0

.5

1

Sibship Gender Minority Position

0

1

2

3

4

.5

1

Sibship Gender Balance Position

0

Probability of 2 Year College Enrollment

Sibship Gender Majority Position

0

1

2

3

4

Number of Siblings
Sons

Daughters

Figure 8: Two-Year College Enrollment Probability by Sibship Gender Composition in the ADD
Health Cohort

66

APPENDIX B

TABLES FOR CHAPTER TWO

67

Table 5: Sample Means by Cohort and Sex for Variables Uses in Analysis

Two-year college enrollment(%)
Four-year college enrollment(%)
Sibling gender composition
living in sibshipgender-majority
living in sibshipgender-minority
living in sibshipgender-balance family
Sibling Size
Number of sibling
School performance
Math standardized score\Highest math
Reading standardized score\
Family Resources
Cultural-related classes\extracurricular
Parent-child discusses school things\
Parent social resources (parent
Parent monitor\parent supervisiona
Parents’ support children’s autonomy
Parents’expectation\ college
Race and Ethnicity (%)
White
Black
Hispanic
Asian
Others
Family Living arrangement (%)
Intact family
Step family
Single family
Other guardians
Mother education (%)
Mather less than HS
Mather HS
Mather some college
Mather college graduated
Mather beyond college

NELS (1990-1994)
Overall Girls Boys t-testb
23
23
22

ADD Health (1994-2001)
Overall Girls Boys t-testb
42c
44c
41 *

51

53

48 *

34c

37c

32 *

60
16
24

57
18
25

65 *
13 *
23

64c
8c
28c

62c
9c
28c

68c *
7c
24 *

2.25

2.26

2.24

52.81 52.38 53.27 *
52.69 53.25 52.09 *

1.44c 1.45c 1.42c
6.02 6.16 5.86 *
66.86 66.12 67.62 *

0.73
9.46
3.15
6.14

0.91
9.46
3.16
6.56

0.54 *
9.46
3.13
5.67 *

3.86

3.89

3.84 *

81
9
6
3
0.1

80
9
7
3
0.1

81
9
6
3
0.1

69c
13c
10c
4
3c

69c
14c
10c
4
3c

69c
13c
10c
4
3c

74
11
13
2

73
12
14
2

75
11
13
1*

67c
24c
6c
4c

65c
25c
6c
4c

69c *
21c *
6c
3c *

19
33
22
16
10

20
34
22
14
10

17 *
33
21
18 *
11

20
37c
15c
21c
7c

21
38c
16c
19c
7c

19c *
36c
14c
24c *
7c *

68

0.42 0.49 0.35 *
16.15 16.13 16.16
2.71 2.73 2.69 *
12.35 12.39 12.31
2.69 2.64 2.74 *
2.27 2.28 2.26

Table 5 (cont’d)
High school type (%)
Public High school
89
90
89
82c
83c
81c *
Catholic High school
7
6
7
5c
4c
6c *
Private High school
4
4
4
13c
13c
13c
Student characteristics
Student educational
3.98 4.05 3.90 *
4.19 4.30 4.07 *
a
Household income index
9.57 9.47 9.68 *
2.55 2.53 2.57
Note: Analytic sample means and percentage are weighted.
a.
The scale of the measurements between two cohorts is not consistent.
b.
Statistical significance indicates a gender difference within each cohort.
c.
Significant differences in the z test (t test for continuous variable) between this percentage and the
corresponding percetage for the same demographic characteristics in the NELS sample.
* p<.05

69

Table 6: Mean Difference on Family Resources, by Child’s Gender and Sibship Position in NELS and ADD Health cohort
NELS cohort
Majority
(1) Extracurricular activities
(2) Parent-child discussion/ relationship
(3) Parent know about kid's friends’ parent
(5) Parent supervision /monitor
(6) Parent educational expectation
Minority
(1) Extracurricular activities
(2) Parent-child discussion/ relationship
(3) Parent know about kid's friends’ parent
(5) Parent supervision /monitor
(6) Parent educational expectation
Balance
(1) Extracurricular activities
(2) Parent-child discussion/ relationship
(3) Parent know about kid's friends’ parent
(5) Parent supervision /monitor
(6) Parent educational expectation
ADD Health cohort
Majority
(1) Extracurricular activities
(2) Parent-child discussion/ relationship
(3) Parent know about kid's friends’ parent
(4) Parents respect children’s autonomy
(5) Parent supervision /monitor
(6) Parent educational expectation
Minority
(1) Extracurricular activities
(2) Parent-child discussion/ relationship

Girls
Mean
0.93
9.47
3.13
6.59
3.90

SD.
(1.06) ***
(2.46)
(1.41)
(2.29) ***
(.87) **

Boys
Mean
0.51
9.43
3.12
5.58
3.83

0.74
9.22
3.10
6.43
3.81

(.89) **
(2.61)
(1.53)
(2.52) ***
(.82)

0.57
9.23
3.20
5.81
3.80

0.98
9.59
3.28
6.58
3.91

(1.02) ***
(2.38)
(1.39) **
(2.33) ***
(.89)

0.60
9.65
3.12
5.85
3.88

Girls position Boys position
SD. Majority versus Minority
(.77) ***
(2.24) **
(1.45)
(2.50)
(.85) **
Minority versus Balance
(.89) ***
(2.31) **
**
(1.48) **
(2.41)
(.89) **
Majority versus Balance
(.90)
**
(2.16)
**
(1.36) **
(2.42)
**
(.87)
Majority

0.48
16.06
2.73
2.64
12.40
2.28

(.70) ***
(2.38)
(.67) **
(.99) **
(2.45) **
(.70)

0.35
16.16
2.68
2.72
12.26
2.25

0.46
16.11

(.68) ***
(2.39)

0.29
16.12

70

versus Minority

(.73)
(2.34)
(.72)
(1.05)
(2.42) **
(.72)
**
Minority versus Balance
(.58)
***
(2.20)

Table 6 (cont’d)
(3) Parent know about kid's friends’ parent
(4) Parents respect children’s autonomy
(5) Parent supervision /monitor
(6) Parent educational expectation
Balance
(1) Extracurricular activities
(2) Parent-child discussion/ relationship
(3) Parent know about kid's friends’ parent
(4) Parents respect children’s autonomy
(5) Parent supervision /monitor
(6) Parent educational expectation

2.68
2.56
12.70
2.27

(.75)
(1.01) ***
(2.58) **
(.68) *

2.71
2.80
12.28
2.36

0.52
16.28
2.75
2.66
12.27
2.29

(.79) **
(2.28)
(.64)
(.99) **
(2.46)
(.70)

0.41
16.19
2.71
2.77
12.44
2.27

71

(.65)
(.98)
(2.39) **
(.74)
Majority versus Balance
(.83)
*
(2.26) **
(.71)
(1.00)
(2.47)
(.72)

Table 7: Multinominal Logistic Regression Predicting College Enrollment Pattern Using NELS (1990-1994) and ADD Health
(1995-2008)
Panel-A: NELS 1990-1994 2-year versus non
Model-1 Model-2
Female
0.310** 0.246*
(0.113) (0.123)
Sibling size
-0.087* -0.076*
(0.036) (0.036)
School Performance
Grade 10 math standardized
0.008
0.010
(0.009) (0.009)
Grade 10 reading
0.023** 0.021*
(0.009) (0.009)
Family Resources
(1) Extracurricular activities
0.129+
(0.074)
(2) Parent-child discusses school
0.070**
(0.025)
(3) Parent social resources
0.117**
(0.038)

4-year versus none
Model-1 Model-2
0.334** 0.229+
(0.119) (0.128)
-0.119** -0.099*
(0.039) (0.039)
0.077*** 0.081***
(0.009) (0.009)
0.030*** 0.027**
(0.009) (0.008)
0.197**
(0.072)
0.143***
(0.025)
0.207***
(0.041)

Panel-B: ADD Health
Female
Sibling size
School Performance
The highest level of math
Vocabulary score
Family Resources
(1) Extracurricular

2-year versus none
Model-1 Model-2
0.112a
0.110
(0.075) (0.074)
-0.086** -0.085**
(0.029) (0.029)

4-year versus none
Model-1 Model-2
0.288** 0.226*
(0.102) (0.099)
-0.128** -0.102*
(0.044) (0.043)

0.010* 0.010*
(0.005) (0.005)
0.345*** 0.337***
(0.025) (0.026)

0.027*** 0.025***
(0.006) (0.006)
0.645*** 0.644***
(0.033) (0.034)

0.133**
0.015
(0.051)
(0.064)
(2) Parent-child
0.043**
0.067***
(0.016)
(0.019)
(3) Parent social resources
0.100*
0.149*
(0.046)
(0.063)
(4) Parents’ support children’s
-0.015
0.318***
(0.036)
(0.052)
(5) Parent supervision
0.048*
0.055*
(5) Parent supervision
0.001
0.015
(0.021)
(0.022)
(0.015)
(0.017)
(6) Parent educational
0.236** 0.172*
0.369*** 0.298***
(6) Parent college
0.095+
0.088 0.334*** 0.327***
(0.082) (0.081)
(0.088) (0.087)
(0.056) (0.056)
(0.067) (0.065)
Student educational
0.747*** 0.704***
1.479*** 1.415***
Student college
0.280*** 0.256*** 0.677*** 0.634***
(0.086) (0.089)
(0.100) (0.103)
(0.034) (0.035)
(0.054) (0.055)
Family income 1992
0.113** 0.124**
0.082+ 0.090*
Household income quarter 0.208*** 0.211*** 0.295*** 0.305***
(0.039) (0.040)
(0.045) (0.045)
(0.037) (0.039)
(0.049) (0.051)
+ p< .1; * p<.05; ** p<.01; *** p<.001;Note. NELS analytic sample, N=6364. ADD Health sample, N=5615.Reported multi-nominal logistic
regression coefficient and standard errors in parentheses. Model control for Heckman selection probability, family living arrangement, race, mother
education, high school type, region. a Significant different from NELS cohort * p<.05 (Hausman test applied).

72

Table 8: Multinominal Logistic Model on College Enrollment by Sibship Gender Composition in
NELS
2 year
4-year
a
Model-1 Model-2 Model-3 Model-1 Model-2a Model-3
Female
Sibling size

0.260*
(0.123)
-0.164***
(0.045)

0.247*
(0.122)

0.142
(0.164)
-0.080*
(0.039)

0.222*
(0.110)
-0.097*
(0.041)

-0.001
(0.185)

-0.038
(0.172)
-0.091*
(0.039)

Sibship majority position

-0.031

-0.146

-0.086

-0.271

(0.158)

(0.174)
-0.052
(0.197)

(0.170)

Sibship balance position

(0.155)
0.054
(0.179)

Gender-specific family context
Number of Siblings*Female
Number of Siblings*Male

-0.039
(0.043)
-0.130*
(0.057)

Sibship Majority

-0.048
(0.048)
-0.140*
(0.058)

0.170
0.421*
(0.210)
(0.201)
Note. NELS analytic sample, N=6364. Reported multi-nominal logistic regression coefficient and standard
errors in parentheses. a The interaction between female and sibling size is significant in Model 2. But, to
specifiy gender-specific effect, we exmaine interaction separated by gender. The refernece group is the
single-child families.+ p<.1* p<.05** p<.0*** p<.001

73

Table 9: Multinominal Logistic Model on College Enrollment by Sibship Gender Composition in
ADD Health
2 year
4-year
a
Model-1 Model-2 Model-3 Model-1 Model-2a Model-3
Female

0.115
(0.078)
-0.056b
(0.035)

Sibling size
Sibship majority position
(

f

i

i )

Sibship balance position
Gender-specific family context
Number of Siblings*Female
Number of Siblings*Male

0.217
(0.122)

-0.062
(0.117)
-0.065*
(0.030)

0.375*** 0.444**
(0.101) (0.151)
-0.015b
(0.052)

0.131

-0.162

0.186+b

(0.173)
0.167
(0.175)

(0.118)

(0.112)
0.162
(0.201)

-0.091*
(0.043)
-0.107*
(0.042)

-0.070
(0.053)
-0.009b
(0.064)

0.266*
0.093a
(0.121)
(0.198)
Note. ADD Health sample, N=5615.Reported multi-nominal logistic regression coefficient and standard
errors in parentheses. a The interaction between female and sibling size is significant in Model 2. But, to
specifiy gender-specific effect, we exmaine interaction separated by gender. The refernece group is the
single-child families.bSignificant different from NELS cohort, applying only for female, sibling size and
gender composition variable across cohorts.* p<.05 (Hausman test applied)
Sibship Majority position)*Female

74

APPENDIX C

SUPPLEMENTAL TABLES

75

Table 10: The Classification of Sibhip-sex Majority, Balance and Minority Context at Home
Type

Boy

Girl

1

Sibsize=0

no sibling

Majority

Majority

2

Sibsize=1

1 girl sibling

Balance

Majority

3

Sibsize=1

1 boy sibling

Majority

Balance

4

Sibsize=2

2 girls sibling

Minority

Majority

5

Sibsize=2

1 girl and 1 boy sibling

Majority

Majority

6

Sibsize=2

2 boys sibling

Majority

Minority

7

Sibsize=3

3 girls sibling

Minority

Majority

8

Sibsize=3

2 girl and 1 boy sibling

Balance

Majority

9

Sibsize=3

1 girl and 2 boys sibling

Majority

Balance

10

Sibsize=3

3 boys sibling

Majority

Minority

11

Sibsize=4

4 girls sibling

Minority

Majority

12

Sibsize=4

3 girls and 1 boy sibling

Minority

Majority

13

Sibsize=4

2 girls and 2 boys

Majority

Majority

14

Sibsize=4

1 girl and 3 boys sibling

Majority

Minority

15

Sibsize=4

4 boys sibling

Majority

Minority

76

Table 11: Family Resources Variables
Variable in NELS data

Variable description

Variable Range

(1) extracurricular
activities organized
by parents and by
schools

Parents’ report of whether the focal student
attended cultural-related classes after school
which included studying (a) art, (b) music, (c)
dance, (d) language and (e) the computer.

This measure took
the sum of all five
items, ranging from
0-5.

(2) parent-child
discussion and
parent-child
relationship

Parent reported how frequently a parent
This measure scaled
discussed the following with their teenage child from 0 (never) to 2
(a) selecting courses, (b) school activities, (c) (often). We took the
school studying, (d) grades, (e) ACT/SAT exam, sum of all six items,
and (f) applying to colleges.
ranging from 0-12.

(3) social
capital—knowing the
parents of the
students’ friends

Parents reported “whether he/she knows parents
of teen’s 1st-5th friend” which represents the
number of a student’s five closest friends’ that
parents knew in the NELS data.

(5) involvement in
monitoring and
supervision of
student activities
such as homework

This measure took
the sum of all 5
friends, ranging from
0-5.

The scale of each
Students reported that how often their parent
item ranged from 1-5
monitors: (1) where teen is after school, (2)
and higher scores
where teen goes at night, and (3) what teen does
represented stronger
with free time.
control from parents.
The scale of item
(6) education and career
ranged from 1-7 and
expectations
Students reported educational expectations and
higher scores
parents had for their parent expectation on ‘‘
represented higher
students
expectations.

77

Table 11 (cont’d)
Variable in ADD Health
Variable description
data
Student self-report as to whether the student
(1) extracurricular
attended cultural-related extracurricular
activities organized
activities in school, such as attending a club
by parents and by
focused on learning a second language,
schools
music club, or computer club.
Parented reported (a) how well gets along
(2) parent-child
with the adolescent, (b) makes decisions
discussion and
about the adolescent’s life with the
parent-child
adolescent, (3) trust the adolescent, and (4)
relationship
satisfaction with his or her relationship with
the adolescent.

Variable Range
This measure also
took the sum of all
nine items, ranging
from 0-9.

Answers were
recoded into five
categories and we
took the sum of all
four items, ranging
from 0-20.
Answers were
re-coded into six
(3) social
categories from 0 to
Parents reported in wave one about “how
capital—knowing
5 (represents five or
many parents of [focal child’s] friends have
the parents of the
more friends) to
you talked to in the last 4 weeks?”
students’ friends
align with the same
variable in the
NELS data.
Students reported whether parents allow you
(4)
making your own decisions about (1) the
This measure took
independence—allo
time at home on weekend nights, (2) people the sum of all 4
wing students with
you hang around with, (3) how much
items, ranging from
autonomy for some
television you watch, and (4) which
0-4.
decisions
television programs you watch.
The scale of each
(5) involvement in
item ranged from
Student reported how often each resident
monitoring and
1-5. We combine
mother and father was home: (1) before they
supervision of
mother and father
go to school, (2) after they return from
student activities
information as a
school and (3) when they go to bed.
such as homework
parental supervision
variable.
The scale of student
Students reported how likely is it that you
expectation was a
will go to college?” where higher scores
(6) education and career
five-point scale.
represented higher expectations. And, parent
expectations parents
The scale of parent
reported their college expectations for
had for their students
college expectation
children as "how disappointed would you be
was a three-point
if your kid did not graduate from college?"
scale.

Note. a. All resource measures in NELS came from the second follow-up data.
b
. All resource measures in ADD Health came from Wave 1 data.

78

REFERENCES

79

REFERENCES

Amin, V.,PetterL.,& Dan-Olof, R. (2011). Mothers Do Matter: New Evidence on the Effect of
Parents’ Schooling on Children’s Schooling Using Swedish Twin Data. IZA, (Working Paper
No. 5946).
Anastasi, A. (1956). Intelligence and Family Size. Psychological Bulletin, 43(3), 187-209.
Blake, J. 1989. Family Size and Achievement. Berkeley, CA: University of California Press.
Blake, J. (1981). Family Size and the Quality of Children.Demography, 18(4), 421-42.
Bailey, T. R., Jaggars, S. S., & Jenkins, D. (2015).Redesigning America’s Community Colleges: A
Clearer Path to Student Success. (pp. 229–239). Cambridge, MA: Harvard University Press.
Bissell-Havran, J. M., Loken, E., & McHale, S. M. (2012). Mothers’ Differential Treatment of
Adolescent Siblings: Predicting College Attendance of Sisters Versus Brothers. Journal of
Youth and Adolescence, 4, 1267-1279.
Black, S. E., Devereux, P. J., &Salvanes, K. G. (2005). Why the apple doesn’t fall far:
Inderstanding intergenerational transmission of human capital. American Economic Review,
95, 437-449.
Bleeker, M. A., & Jacobs, J. E. (2004). Achievement in math and science: Do mothers’ beliefs
matter 12 years later? Journal of Educational Psychology, 96(1), 97-109.
Brody, C., & Steelman, L. (1985).Sibling Structure and Parental Sex-Typing of Children’s
Household Tasks. Journal of Marriage and Family, 47(2), 265-273.
Bumpus, M. F., Crouter, A. C., & McHale, S. M. (2001). Parental autonomy granting in
adolescence: Exploring gender differences in context. Developmental Psychology, 37,
163–173.
Butcher, K., & Case, A. (1994). The Effect of Sibling Sex Composition on Women’s Education
and Earnings. The Quarterly Journal of Economics, 109(3), 531-563.
Census (2010).The decreasing trend of family size. (Nov 17, 2016). Retrieved
fromhttps://www.census.gov/newsroom/press-releases/2016/cb16-192.html
Carter, R. S., & Wojtkiewicz, R. A. (2000). Parental involvement with adolescents’ education: Do
daughters or sons get more help? Adolescence, 35(137), 29-44.

80

Conley, D. (2000). Sibship Sex Composition: Effects on Educational Attainment. Social Science
Research, 29(3), 441-457.
Conley, D.& Rebecca G. (2006). Parental Educational Investment and Children’s Academic Risk:
Estimates of the Impact of Sibship Size and Birth Order from Exogenous Variation in Fertility.
The Journal of Human Resources, 41, 722-737.
Chen, I., Schneider, B., & Frank, K. (2015). The cohort trends ofsibling size and the rise of female
advantages: Examining therelationship between family resources and college enrollment.San
Diego, CA: Population Association of America.
Chu, C. Y. Cyrus, Xie, Yu.,&Yu, Ruoh-rong. (2007). Effects of Sibship Structure Revisited:
Evidence from Intrafamily Resource Transfer in Taiwan. Sociology of Education, 80, 91-113.
DiPrete, T. A., & Buchmann, C. (2013). The Rise of Women: The Growing Gender Gap in
Education and What It Means for American Schools. (pp. 1-68). New York: Russell Sage
Foundation.
Downey, Douglas B. (1995). When Bigger Is Not Better - Family-Size, Parental Resources, and
Childrens Educational Performance.American Sociological Review, 60(5), 746-761.
Downey, Douglas B. (2001). Number of siblings and intellectual development: The resource
dilution explanation. American Psychologist, 56, 497-504.
Dufur, M. J., Parcel, T. L., Hoffmann, J. P., & Braudt, D. B. (2016). Who has the advantage? Race
and sex differences in returns to social capital at home and at school. Research in Social
Stratification and Mobility, 45, 27-40.
Dumais, S. A. (2002). Cultural Capital, Gender, and School Success: The Role of Habitus.
Sociology of Education, 75, 44-68.
Eggebeen, D. J. (1992). Changes in Sibling Configurations for American Preschool-Children.
Social Biology, 39(1-2), 27-44.
Ermisch, J.& Paronzato, C. (2009). Causal Effects of Parents’ Education on Children’s Education.
in Persistence, Privilege, and Parenting: The Comparative Study of Intergenerational
Mobility, edited by T. Smeeding, R. Erikson, and M. Jantti. New York: Russell Sage
Foundation.
Hobbs, Frank & Stoops, Nicole. (2002). Demographic Trends in the 20th Century, from Census
2000 Special Reports, CENSR-4 (November 2002), Figure 5-3.
Fasang, A. E., Mangino, W.& Bruckner, H. (2014). Social Closure and Educational Attainment.
Sociological Forum, 29(1), 137-164.

81

Heckman, J. (1979). Sample Selection Bias as a Specification Error. Econometrica, 47 (1),
153-161
Kalmijn, M., & van de Werfhorst, H. G. (2016). Sibship Size and Gendered Resource Dilution in
Different Societal Contexts. Plos One, 11(8).
Kim, D. H., & Schneider, B. 2005. Social Capital in Action: Alignment of Parental Support in
Adolescents’ Transition to Postsecondary Education. Social Forces, 84(2), 1181-1206.
Kuo, H.-H. D., & Hauser, R. M. (1995). Trends in Family Effects on the Education of Black and
White Brothers. Sociology of Education, 68(2), 136-160.
Lee, J. (2008). Sibling Size and Investment in Children’s Education: An Asian Instrument. Journal
of Population Economics, 21, 855-875.
Ma, J. & Baum, S.(2016). Trends in Community Colleges: Enrollment, Prices, Student Debt, and
Completion. (Research Brief 2016 April). New York, NY: The College Board.
Marks, J. L., Lam, C. B., & McHale, S. M. (2009). Family Patterns of Gender Role Attitudes. Sex
Roles, 61, 221-234.
Mare, R. D., &Maralani, V. 2006. The Intergenerational Effects of Changes in Women’s
Educational Attainments. American Sociological Review, 71(4), 542-564.
McHale, S. M., Crouter, A. C., & Whiteman, S. D. 2003a. The family contexts of gender
development in childhood and adolescence. Social Development, 12, 125-148.
McHale, S. M., &Crouter, A. C. How do children exert an impact on family life? (2003b). In:
Crouter, A.C., Both, A., editors. Children’s influence on family dynamics: The neglected side
of family relationships. Lawrence Erlbaum Associates; Mahwah, NJ, US: pp. 207–220.
Powers, D. A. & Yu X. (2000). Statistical Methods for Categorical Data Analysis. New York:
Academic Press.
Steelman, L. C. & Powell, B.(1989). Acquiring Capital for College - the Constraints of Family
Configuration.American Sociological Review, 54, 844-855.
Steelman, L. C., Powell, B., Werum, R., & Carter, S. (2002). Reconsidering the Effects of Sibling
Configuration: Recent Advances and Challenges. Annual Review of Sociology, 28, 243-269.
Torche, F. (2011). Is a College Degree Still the Great Equalizer? Intergenerational Mobility across
Levels of Schooling in the United States.American Journal of Sociology, 117(3), 763-807.
Winship, C., &Mare, R.D. (1992).Models for Sample Selection Bias.Annual Review of Sociology,
18, 327–50.

82

Williams, R. (2009). Using Heterogeneous Choice Models to Compare Logit and Probit
Coefficients Across Groups. Sociological Methods and Research, 37(4), 531-559.
Yamaguchi, K. &Ferguson, L. R. (1995).The Stopping and Spacing of Childbirths and Their
Birth-History Predictors-Rational-Choice Theory and Event-History Analysis. American
Sociological Review, 60, 272-298.

83

CHAPTER THREE
THE EFFECT OF MARRIED MOTHERS’ EMPLOYMENT ON CHILDREN’S HIGH
SCHOOL ACHIEVEMENT, MATHEMATICS SELF-EFFICACY AND STEM MAJORS

Introduction
Enhancing women’s participation in the STEM fields of computer science and engineering
remains a national priority. Females lag behind their male counterparts in certain STEM fields,
among STEM degree recipients, and in the labor market (Anderson & Kim, 2006; Wang, 2013).
The best method of bringing more women into STEM careers remains unclear; current
explanations for the low number in STEM fields tend to emphasize educational factors, gender
stereotypes and gendered perceptions of STEM careers (Ceci & Williams, 2011; Eccles & Roeser,
2011; Tyson, Lee, Borman, & Hanson, 2007). Some evidence, however, indicates that mothers’
employment status and family social and psychological resources have a greater influence on
women’s entry into STEM fields (Fomby, 2013; Xie, Fang, &Shauman, 2015). Despite substantial
demographic change in delayed parenthood and mothers’ increased participation in the labor force,
the extent to which mothers’ labor force participation is associated with their children’s likelihood
of majoring in a STEM field remains underexplored.
There are several explanations why fewer females than males major in STEM fields,
including biological differences in spatial versus verbal skills (Ceci, Williams, & Barnett, 2009);
parents’ expectations and investment in their children’s schooling (Baker & Milligan, 2016;
Morgan, 2005; Morgan &Weeden, 2013); social and cultural influences (Guiso, Monte, Sapienza,
&Zingales, 2008); non-cognitive skills and peer influence (DiPrete& Jennings, 2012 ; Kulis,
Sicotte, & Collins, 2002); gender-specific social contexts and discrimination (Wang, Eccles, &
Kenny, 2013; Ceci& Williams, 2011) and teacher bias. Despite the debate over the shortage of

84

female role models and supportive communities in the workforce, there is a paucity of research on
when and the extent to which mothers’ employment status encourages their adolescent daughters
to major in a STEM field.
Mothers’ employment status shapes their children’s educational opportunities and aspirations
through parental investment and involvement in children’s schooling and career preparation.
Adolescents with career mothers tend to have better economic and socio-cultural resources than
adolescents with stay-at-home mothers. Career mothers are able to spend more of their income,
time and knowledge investing in educational resources and services for their children, such as
having their children take the SAT, receive tutoring, and attend summer camps and extracurricular
activities.
This study examines the role of mothers’ employment conditions on children’s STEM major.
Specifically, I examine two aspects of mothers’ employment -- working hours and occupational
prestige -- in four cohorts of surveys between 1980 and 2013 to discern whether having a career
mother affects children’s decision to major in a STEM field and how this may change over time.
Career mothers refer to mothers that were part of the labor market during their children’s life stage
at 10th grade to four years after. This study assumes that the role of motherhood has changed as
more mothers increase their economic independence and as female equity rises. Mothers’
employment status and economic resources may promote their children’s STEM career through
three transmitting processes: 1) children’s mathematics self-efficacy, 2) cognitive development,
and 3) family support.

85

Background
Career Mothers and Impacts on Children
Current studies of the relationship between mothers’ employment and children’s educational
outcomes present three arguments over whether paid jobs have a positive, negative or neutral
effect on their children’s educational outcomes (Goldberg, Prause, Lucas-Thompson, &Himsel,
2008). For example, Parcel, Nickoll & Dufur (1996) found that mothers with full-time jobs had a
negative effect on children’s reading and math scores for nine-to-twelve-year-olds compared to
mothers employed part-time. Ruhm (2004) also found a small deleterious effect of full-time
employed mothers on verbal ability and a larger negative impact on school achievement of
five-to-six-year-olds.
In contrast, Haveman, Wolfe & Spaulding (1991) identified that the effect of mothers’
employment on four-to-fifteen-year-old children had a positive association with their children’s
probability of graduating from high school. Johnson and colleagues (2012) found that mothers
with higher job stability and jobs with high cognitive-skill demands tended to reduce children’s
behavioral problems. Jobs requiring high cognitive skills were also expected to increase mothers’
wages, which allowed mothers to contribute to family income and investment in children.
A third group of studies found no clear association between mothers’ employment and their
children’s educational outcomes. For example, Aughinbaugh & Gittleman (2004) examined the
impact of maternal employment on children’s adolescents’ risky behaviors and found no strong
evidence between mother’s employment and children’s behavioral problems using data from the
NLSY79. Baum (2004) found that mothers’ employment during childhood did not affect school
GPA when their adolescent was in high school. However, Baum found a marginal negative effect
on high school grades when mothers were employed during high school students’ adolescent years;

86

this effect was stronger for sons than daughters. Despite considerable research, little is known
about whether mothers’ employment status is associated with children’s STEM decision, and even
less has been explored about the mechanisms through which mothers’ employment conditions
affect their children’s educational outcomes.
What are the key messages conveyed by the aforementioned research? First, a longitudinal
study is desirable to identify family processes that transmit causal impacts from mothers’
employment to children’s outcomes in adolescence. Second, by taking advantage of longitudinal
data, researchers should investigate the mechanisms that may account for the detrimental and
beneficial effect of mothers’ employment on children across children’s life course, such as early
childhood or adolescence. Finally, definitions of mothers’ employment should be expanded to
other factors in terms of earnings, work hours and occupational prestige, rather than limited to a
dichotomous variable and a nationally representative analysis is needed. The goal of this study is
to address these issues by proposing a research framework that is based on the suggestions by
Dufur, Parcel & Troutman (2013) and testing it with four nationally representative surveys.
Previous research has suggested that schooling and parenting are two major mechanisms to
transmit mothers’ influence into children’s development. I argue that it is a contingent effect of
maternal employment on children’s educational outcomes. Whether this effect is positive or
negative may depend on the family processes between mother and children, such as children’s
development on cognitive ability and interests in a STEM career. In this study, I focus on two
processes. The first process is to link mothers’ employment to children’s schooling and school
achievement. The second process is to link mothers’ employment to children’s development in
math self-efficacy.

87

Theoretical Perspectives: the Causal Effect of Career Mother on Children’s Cognitive
Development and Interest in a STEM Career
I proposed three theoretical hypotheses to link maternal employment and children’s STEM
participation. First, the gender asymmetric hypothesis emphasizes the involvement of mothers’
time and care has a strong influence on their children’s educational outcomes and interest in a
STEM career. Therefore, a reduction of mothers’ involvement as a result of their participation in
the labor market has a more detrimental effect on their children’s school achievement. Following
this line of research, and given that more mothers have entered the labor force, has led to the
hypothesis that the potentially negative effects of maternal employment for children may increase
over time.
The economic independence hypothesis describes that changes from “traditional motherhood”
to “modern motherhood,” have led to more economic autonomy allowing for mothers to be
involved in and invest in children’s schooling. For instance, children may need more resources and
family support to indulge in STEM major interests. Career mothers with more moderate gender
role attitudes, more prestigious jobs, and greater participation in the labor market, are more likely
to allocate economic, social, and cultural resources to ease their children’s transition into
demanding math/science careers. Augustine, Cavanagh &Crosnoe (2009) found that maternal
cognitive and psychological skills contribute to the effect of maternal employment on children’s
outcomes. Married career mothers also gained more economic and social resources by
participating in the labor market. College-educated career mothers are associated with more
effective parenting and caring styles than less educated mothers (Kalmijn 1994). For example,
career mothers tend to know how to stimulate their children’s cognitive development, and manage,
monitor, and encourage their children’s academic activities and decisions (Archer et al., 2012).

88

Career mothers could also serve as role models to their children in transmitting career values and
experiences (Kalmijn 1994). Archer and colleagues (2012) found that mothers encouraged
daughters to engage in self-supervision and self-confidence in mastering any subjects in school
that include actual participation in the STEM major courses. Based on this line of reasoning, I
argue that the children of career mothers are more likely to be interested in math/science areas
among high school seniors and a narrowing gender gap in STEM fields.
According to the mothers’ cognitive ability and occupational prestige hypothesis, jobs with
high occupational prestige are more likely to include complex jobs requiring mothers’ high
cognitive skills and development, resulting in better stimulation and mothering on children’s
cognitive development (Cooksey, Menaghan, &Jekielek, 1997; Parcel &Dufur, 2001; Parcel
&Menaghan, 1990). In addition, career mothers with prestigious jobs are more likely to earn more,
and have stable income and professional affiliations with colleagues. As such, these mothers tend
to promote the internalization of behavioral norms, encourage self-efficacy, and facilitate
cooperation in their children, as their occupation tends to value self-supervision and self-worth.
This line of research has suggested that parents in more complex occupations provide more
warmth and cognitive stimulation for their children than do parents in routinized occupations
(Parcel &Menaghan, 1994) and full-time shift position with lower cognitive demands. Mothers’
job stability and job complexity, therefore, have a positive effect on children’s educational
outcomes, cognitive development and interest in mastering STEM fields (Johnson, Kalil,
&Dunifon, 2012). I therefore argue that children of mothers in high prestige careers are associated
with growing school achievement, which in turn, affects children’s interest and opportunities in
STEM career among high school seniors.

89

Career Mother and Family Supports
Parents’ school involvement and support are indicators of family social capital, which refers
to “relations among persons that facilitate action” (Coleman 1988: p.100). In this study, action
indicates that family members built relations within and outside their immediate family that
facilitate attending college and majoring in STEM. Family social capital represents a connection
between parent and teacher, and between family and school (Chin & Phillips, 2004; Kim &
Schneider, 2005; Offer & Schneider, 2007; Crosnoe, 2004, 2009). For example, college educated
mothers are better able to recognize the value of a college education for children, particularly for
daughters (Buchmann & DiPrete, 2013; Raley & Bianchi, 2006), and are more likely to be
equipped with math-specific or science-specific capital to build up a local community of the
“STEM family” (Archer et al., 2012).
In addition, family social capital and support for education varies by children’s gender, school
achievement and mother’s employment. Eccles (2015) found that parents of daughters are more
likely to have cohesive parent-child communication, relationships, and social capital than parents
of sons. In addition, new family social capital may form in the raising of a mother’s position in the
labor market. Although mothers with full-time professional jobs may participate less often in
regular school activities (e.g., PTO meeting, school board), some evidence has indicated that
career mothers are more flexible in arranging and organizing school (Fox, Han, Ruhm, &
Waldfogel, 2013) and leisure activities (e.g., STEM-related programs, science camp, summer
travel) for their children. To date, we know less about whether and how the family support changes
in the dynamics of mothers’ labor force participation. Few researchers have investigated the
connections between the changes in mothers’ employment and how it affects children’s
participation in STEM fields. While previous research has been clear on the positive effect of

90

family social capital on students’ schooling, few pieces of research have investigated whether
family support has increased or declined over time, and what types of new family supports may
take shape.
The theoretical framework in this study integrates the family social capital theory,
gender-specific resources perspective and prior literature on factors closely related to mothers’
employment and students’ selection of math/science college majors. Within this framework,
students’ participation in STEM field is a function of school achievement in the 12th-grade,
exposure to effective family support and, belief in mathematics self-efficacy.
Research Questions
This study will examine the effects of two maternal employment conditions -- work hours and
occupational prestige – on children’s interest and participation in math/science activities and
college major. Three questions about potential changes in mothers’ labor force participation, and
children’s math/science interesting and participation guide this research: 1) As mothers’ labor
force participation increases, what is the relationship between maternal employment and children’s
majoring in a STEM field? 2) To what degree does maternal employment affect children’s
math/science major through math-efficacy and school achievement? 3) Does mothers’
employment affect boys’ and girls’ math/science participation in different ways?
Hypothesis 1a: The negative influence of mother’s employment on children’s participation in
STEM fields increases over time.
Hypothesis 1b: The negative influence of mother’s employment on children’s participation in
STEM fields decreases over time.
Hypothesis 2: The growing number of high-prestige and full-time career mothers strengthens
their influence on children’s education through cognitive stimulation and self-confidence in math.

91

Hypothesis 3: The influence of full-time career and mothers with prestigious jobs affects
daughters more than sons, but this tendency is present only in the recent cohort.
Data and Methods
Data
My sample consists of high school seniors living with their married biological parents. I
exclude single-parent, divorced and legal guardian families to simplify the model and reduce
confounding. In doing so, this study is able to focus on changes and patterns in mothers’
employment on the unit of two-biological family in respect to children’s educational outcomes. I
also exclude families in which the adolescent respondents did not report or whose responses were
missing in the postsecondary outcomes from the last follow-up. Limiting the sample in these ways
may produce some selection bias in the final analytic sample. For example, it slightly changes the
sample toward families who reported higher father’s educational attainment of 2.44 and mother’s
education attainment of 2.33 in the High School and Beyond 1980 in the descriptive statistics
presented in Table 12, compared to an average father’s educational attainment of 2.42 and an
average mother’s education attainment of 2.30 in Table 27.
To track changes in mothers’ employment, I consult four longitudinal studies conducted by
the National Center for Educational Statistics (NCES): High School and Beyond 1980 (HS&B),
the National Education Longitudinal Study 1988 (NELS 88), the Education Longitudinal Study of
2002 (ELS02) and High School Longitudinal Study (HSLS09). HS&B includes 14,670 sophomore
students in 988 schools who reported in the third follow-up survey in 1986. The NELS includes
12,144 10th grade students in the first to fourth follow-up. The ELS includes 16,197 students in
750 schools who reported in the third follow-up survey. The HSLS data surveys 17,800 9th grade
students in 944 schools who reported in both the first and the second follow-up surveys in 2013.

92

I use data from student and parent surveys to compare four cohorts of 15-16 year-olds
interviewed in 1980, 1990, 2002 and 2009. (See the appendix for description of background and
control variables.) I discuss key variables of interest below.
Measures of Dependent Variables
My primary outcome of interest is whether students have enrolled in a math/science
intensified college major two years after high school graduation. This outcome is based on the
requirement of math/science ability to categorize into four nominal choices (George-Jackson,
2014; Lara Perez-Felkner et al., 2015): (a) participated in physics, engineering, mathematics, and
computer science major (PEMC); (b) participated in agricultural, architectural and natural
resources sciences (AAN); (c) participated in biological sciences, psychology, clinical and health
science (BPCH); and (d) enrolled in humanities, education, general curriculum, social and
behavior science (HESB).
Measures of Independent Variables
This study examines three paths by which mothers’ employment conditions (e.g., working
hours and occupational prestige) affect children’s college majors. Mother’s employment was
measured by two dummy variables to capture whether a mother works part-time, full-time, or not
at all. I also recode the occupation category of mother and father in the survey into occupational
prestige scores ((Blau& Duncan, 1967; also See Table 28). The occupational prestige scores come
from Nakao & Treas (1990, 1994) for each of the 503 Census occupations and updated code for the
more recent survey data from Hout and colleagues (2010).
Students’ Mathematics Efficacy. Table 17 lists the indicators of students’ mathematics
efficacy, their factor loadings, and the factors’ measure of reliability. The table also contains the
indicators and their factor loadings, with the measures of reliability across cohorts used in the

93

analysis. Table 26 gives further details on the variables. Career mothers may have more egalitarian
gender role attitudes, and affective beliefs about their children’s schooling and achievement goal
(Eccles, 2015), which may increase children’s interest in STEM fields. As Bleeker& Jacobs (2004)
suggested, daughters’ pursuit of careers in engineering and other math/science fields was
positively related to their mothers’ belief in and perception of their math and science achievement
and their confidence in succeeding in a STEM career. As such, career mothers are more likely to
have a better knowledge of their children, which may increase likelihood of their children
enrolling in STEM fields.
School Achievement. The measure of students’ achievement uses students’ grade point
average in 12th grade among the HS&B, NELS and ELS data. For the HSLS data, the measure of
students’ achievement uses students’ grade point average in the 11th grade. Although transcript
data is another variable to represent students’ school performance, it produces a larger proportion
of missing cases. Thus, I use students’ grade point average to represent students’ school
achievement. I also measure students’ high school achievement by generating Naïve
Percentile-Ranking with 10th graders composite standardized test score (combined bother reading
and math). I recognize that there is a huge difference between different standardized procedures
and test composite scores across datasets. To perform this analysis, I apply a simple linear equating
procedure to generate a percentile-ranking based on standardized test composite score (Friedkin &
Thomas, 1997). This percentile-rank represents the school ranking of students’ standardized
test-score compared to their peers in each survey cohort. I use this NaĂŻve Percentile-Ranking
variable to double check the distribution of GPA across cohorts. The results are similar when using
either the GPA or NaĂŻve Percentile-Ranking of composite standardized test score. Thus, in the
final analytical model, I report school achievement using school average GPA.

94

Family Resources. The availability of family resources for children is derived from family
structure, parents’ employment status, and fertility decisions. In this study, four types of family
resources are identified that may contribute to families with a career mother that has more
resources: (1) family preparation for school, (2) parental involvement in schooling, (3) parents’
educational expectations, (4) parents’ savings for the focal child’s education.
Family preparation for school: this measure is derived from each data set with three identical
items: (1) how often the student goes to class without pencil/paper, (2) how often the student goes
to class without books, (3) how often the student goes to class without having done the homework.
I recode the original scale and generate the sum of all three items for each family.
Parental involvement in schooling: this measure is derived from each data set with four
identical activities: (1) how often parents attend school meetings, (2) how often parents phoned a
teacher or counselor, (3) how often parents attended school events, and (4) how often parents
volunteered at their 10th graders’ school. On a scale of 0-4, this measure represents how many
school activities parents participated in at their 10th graders’ school.
Parents’ educational expectation: this measure is derived from each data set with one identical
item. I recode students’ and parents’ educational expectations into seven categories. This question
includes a “don’t know” option on the questionnaire. In this study, I use students’ perception of
their parents’ educational expectations as an indication of social-psychological encouragement. I
assume that the higher the parental expectations, the higher the academic orientation students
perceived from parents. Based on this notion, if students reported that they didn’t know their
parents’ expectations, this implies that students perceive no academic orientation from their
parents.

95

Parents’ saving for the focal child’s education: this measure is derived from each data set, in
which parents report whether they have ever set aside money for a teen’s education. I use a dummy
variable to capture whether parents saved money for the targeted teen.
Measures of Control Variables. (1) Family SES. I include family SES as a composite score in
the model. To make the comparison across three data sets, I generate new family SES composite
scores using the information obtained from each survey, which includes father occupation, father
education, mother occupation, mother education and household income. (2) Race and ethnicity. I
separate this measure into four dichotomous variables, which represent students’ racial and ethnic
background: non-Hispanic white (reference category), Black, Hispanic and Asian/Pacific Islander.
The Hispanic category includes all Hispanic-origin students regardless of whether he/she reported
multi-racial or white. Students who identified as Native American or other race/ethnicity are
excluded from the analyses due to small sample sizes. (4) Mother’s education. Using the parent
survey, I categorize mother’s education into four groups: 1) less-than high school, 2) high school
diploma, 3) some college, and 4) college and beyond. When this variable is missing for some
respondents or when the mother was not surveyed, I use adolescents’ reporting of the education
level of residential mother. (5) Immigrant status. I generate a dichotomous variable to represent
whether or not students were born in the United States. (6) Number of siblings. All four data sets
asked (in various ways) “the number of siblings that 10th grade had” in the HS&B data. In the
NELS data, respondents were asked about the number of siblings in both the student and parent
surveys. In the ELS data, 10th grade respondents were asked about “the number of siblings living
with the 10th grade respondent.” In the HSLS data, respondents were asked about “the number of
siblings living at home.” I use these variables in their original form to represent the number of
siblings, which did not distinguish between full-sibling, half-sibling or adopted sibling.

96

Analytic Methods
This study uses a two-step modeling approach to access whether the gender gap of STEM
major change over time. In the first step, the measurement part of the model is examined and I
report the factor loading, as well as the factor’s measure of reliability in Table 17. All measurement
models across data sets are acceptable (Kline 2011). In the second step, I analyze the multi-group
structural equation model using pooled data across the datasets, where the measurement and
structural parts of the model were simultaneously estimated using Mplus 7.1 software package
(MuthĂŠn&MuthĂŠn, 2012). Given the nature of the nominal outcome, multinomial logistic
regression is used, in which the log odds of the outcomes are modeled as a linear combination of
the predictor variables.
Multi-group analysis is also applied to simultaneously estimate the relationships among
these concepts/factors and to examine hypotheses in the framework of Figure 9. Observations
across four datasets are 23,187. The full information maximum likelihood estimation method is
also adopted to account for missing cases in the independent variables and covariates. Listwise
deletion is used to deal with the missing cases in the exogenous observed variables. Listwise
deletion results in the removal of about 2651 cases from the sample, with a final analytic sample
size of about 20,536 in the SEM model. M-plus package also accounts for survey weight, and
indirect effects through two potential paths that I propose to examine in the research framework.
I then followed the two-stage approach (Anderson & Gerbing, 1988) to implement a model
where the two pathways are not considered in the first model. The purpose of using a two-stage
approach is to establish the auxiliary relationship between mothers’ employment and children’s
STEM participation. The second model takes into account all covariates and the two proposed
mechanisms to investigate if the effect of mothers’ employment on children’s STEM career

97

decision is mediated. I summarize the standardized coefficients in tables, and calculated the
indirect effect of mothers’ employment on STEM career through two pathways (Hayes, 2013;
MuthĂŠn&MuthĂŠn, 2012). With respect to model fit diagnostics, since this study uses the
multinomial logistic regression and maximum likelihood estimation method, it is not as
straightforward to do model fitting diagnostics with multinomial logistic regression models.
Descriptive statistics are reported by each survey data in Table 12-Table 15. First, I
summarize the distribution of STEM participation across cohorts. As shown in the first line of each
table, the proportion of students enrolled in STEM majors drop in the 1990s and slowly increases
in the recent cohort after year 2000. In addition, the proportion of mothers’ labor force
participation is around 42% in the 1980s and increases over 55% in 2009. Students’ school
achievement also slightly increases over cohorts. Among family resources, parents’ educational
expectation and saving efforts for children also increase over time in married-two-biological
families. For the distribution of STEM majors, Table 16 presents the proportion of male and
female participants by gender and within each of the four categories of STEM major fields. A
gender difference test is performed to show whether the difference reaches statistical significance.
Results show that the gender difference is consistent across cohorts and concentrates on the PEMC
majors, HESB major and AAN major in the earlier cohorts.
Next, I perform confirmatory factor analysis (CFA) to analyze the proposed measurement
model. One latent factors and corresponding indicator items are shown in Table 17. Factor loading
and Cronbach’s Alpha are listed in the third and the fifth column in the table. Across four sets of
latent constructs, NELS cohort has relative lower reliability compared to other data sets.
Following the CFA, structural equation modeling (SEM) and multiple-group SEM were
performed using pooled data and cohort-specific data to examine the proposed conceptual model.

98

The path structure in this study is derived from three sets of simultaneously estimated regression
equations. The first set of equations examines how family resources are each influenced by
mothers’ employment conditions. The second set of equations examines how 10th-12th-grade
beliefs in math self-efficacy and 12th-grade average school achievement are each influenced by
mothers’ employment condition and family resources. The final set of equations investigates how
students’ participation in STEM fields is influenced by their self-efficacy beliefs, school
achievement and family environment of support and encouragement.
Multiple-group analyses are employed to examine structural weight invariance across
datasets. Specifically, I explore whether the hypothesized model is equivalent across cohort
(HS&B, NELS, ELS and HSLS), within gender (females and males) and race (White, Asian, and
Black and Hispanic). I focus on the structural patterns of the model; the invariance test centers on
the equivalence of structural path parameters across different cohort of datasets. To describe the
multiple-group approach used in this study, I start with gender-based multiple-group analysis. I
first fit a baseline multiple-group model with factorial equality constraint across gender and the
structural regression coefficients are estimated freely across groups. Following this procedure, I
perform another multiple-group model with all structural regression coefficients across cohorts
constrained to be equal. The result of the chi-square difference test is statistically significant,
which suggests that one or more of the parameters are non-invariant across cohorts. Therefore, I
perform parameters one by one to identify the parameters that should be estimated freely across
cohorts. I list all possibilities in Table 18. In this study, I use the maximum likelihood estimation
(ML) because the nature of the nominal variable, which means a structural invariance test needs to
be obtained based on the corrected chi-square difference (See M-plus website:
https://www.statmodel.com/chidiff.shtml). The baseline model is compared with the

99

constrained-equal between cohorts. If the chi-square difference is significant, it shows that the
given parameter is not equivalent across cohort groups. Therefore, this parameter will be set as a
free estimate in the subsequent models. However, if the test indicates a non-invariance between
groups as an insignificant chi-square difference, then regression coefficients can be constrained to
be equal across cohort groups. By doing this constraint gradually, it helps to identify whether
group differences could result from any other structural weights in the model.
This constraint procedure identifies several non-invariance paths, such as gender, mothers’
employment; school average GPA, self-efficacy factor and two family supports. Therefore, the
final model is set to be free for those estimates across all cohorts. Based on this test, I will need to
account for the structural difference in cohorts and gender. To simplify the analysis, I first focus on
the structural changes and weights across cohorts controlling for gender. After the first set of
models, I report gender difference results in each cohort.
The effect of mothers’ employment on school achievement and math efficacy beliefs is
transmitted through two mediating pathways on children’s STEM major participation. To obtain
the indirect effect, I calculate and test for statistical significance using the M-plus MODEL
CONSTRAINT OPTION
Results of Multiple-Group SEM Analyses
Based on previous analyses, the result indicates that a multiple-group based on cohorts, where
the structural paths from mother employment to self-math efficacy, school achievement, and
STEM majors need to set as free estimates. Table 19-Table 22 presents the final SEM model across
four data sets.
In Table 19, mothers’ full-time job and occupational prestige did not affect children’s
majoring in a STEM field from HS&B, NELS and ELS cohorts. Although there is a slightly

100

negative effect in NELS cohort on the type of AAN field, the trend is consistent from the 1980s to
2002. Until the most recent cohort, mothers’ occupational prestige (p<.05) is positively related to
their children’s participation in PEMC field, which means that a one-unit increase in mothers’
occupational prestige score is associated with a .007 increase in the relative log odds of
participating PEMC major over the relative log odds of those who are participating HESB field.
Mothers’ full-time employment (p<.001) is also positively associated with children’s participation
in a BPCH field. Only two direct effects are present in the HSLS cohort and show positive
connections between mothers’ employment and children’s participation in PEMC and BPCH
fields.
Results reflect that the effect of mothers’ occupational prestige score and mothers’ full0time
employment tend to increase their children’s participation in STEM major in the recent cohort,
HSLS cohort. To facilitate discussion, I created Figure 11 wherein the estimated coefficients were
shown by directed pathways for the sample across cohorts.
Daughters (women) are less likely to participate in the PEMC major and AAN fields except
for the ELS cohort in the Table 21. Over time, the gender gap presents only in the PEMC field in
the both ELS and HSLS cohort. Interestingly, BPCH fields attract more daughters than sons
starting at the HS&B cohort, except in the NELS cohort in the Table 18. This tendency thereafter
resumes with the following ELS and HSLS cohort. School achievement at the 12th grade and math
self-efficacy are positively associated with PEMC major fields among all cohorts. The role of
school achievement in the recent two cohorts, the ELS and the HSLS, becomes more profound and
consistent in all three categories of STEM-related majors. Math self-efficacy is also positively
associated with BPCH major in the NELS and HSLS cohortin the Table 20 and Table 21.

101

The effect of mothers’ employment on children’s educational outcome may also through
indirect paths. In the same tables, I report the indirect effect that may transmit mothers’ influence
through two mediating paths: school achievement and mathematics self-efficacy. Results suggest
that students’ school achievement on PEMC major is more sensitive to mothers’ employment at the
HS&B cohort and spills over to AAN major participation in the NELS and ELS cohorts. The
pattern is consistent: mothers with full-time jobs decrease children’s STEM participation through
school achievement; mothers with higher-prestige jobs increase the STEM participation through
school achievement. Students’ math efficacy does not have a strong indirect effect on the earlier
cohorts, but does mediate students’ participation in PEMC major in the ELS and HSLS cohorts,
particularly for mothers with higher-prestige jobs. This finding supports theoretical hypothesis that
the complexity of mothers’ jobs affects their children’s cognitive ability and self-confidence in
mastering math and science. Coupled with the findings in the previous section, results also
confirmed that school achievement and students’ mathematics efficacy serve as mediation
mechanisms in conveying the effect of mothers’ employment condition on children’s participation
in STEM across cohorts (Figure 12-14).
It is important to note that mothers with full time jobs also have a negative influence on
STEM participation, but this negative effect must go through students’ school achievement.
Full-time career mothers in the NELS and ELS cohort have a negative influence on children’s
participation in all STEM field through decreasing students’ achievement. On the other hand,
full-time career mothers in the ELS and HSLS cohort may also have a negative influence on
children’s participation in PEMC major through students’ math efficacy. This finding implies that
mothers’ employment on children’s STEM participation depends on the schooling and
encouraging processes in relation to participating STEM fields.

102

Table 23-Table 24 reports the structural paths for both boys and girls in four cohorts. For the
direct effect, the results show that only boys with full-time career mothers are more likely to
increase PEMC major participation in the NELS cohort, and girls with full-time career mothers are
more likely to increase participation in BPCH field in the HSLS cohort. The positive indirect
effects are summarized in Table 25 for the comparison of the patterns. Table 25 presents the
indirect paths by which mothers’ employment positively relates to children’s STEM participation.
The result suggests that daughters are more responsive to mothers’ employment in a full-time
career by increasing STEM participation through school achievement. Mothers with high-prestige
jobs are more encouraging of their daughters’ interests in BPCH and PEMC major fields in both
NELS and ELS cohorts. However, in the ELS cohort, this tendency applies only to sons. Sons in
the NELS cohort benefit from a career mother by participating in a PEMC major field. Until the
most recent cohort, daughters continue to gain with full-time career mothers through school
achievement in pursuing their STEM participation. Sons whose mothers hold prestigious jobs may
also benefit from better self-efficacy when entering a PEMC major field. Because of space
constraints, table 5 does not compare boys and girls on each path.
Discussion and Conclusions
I use these findings to examine the research hypotheses of this study. The SEM model in
Table 19-Table 22 show no strong relationship between mothers’ employment and children’s
STEM participation except for the HSLS cohort. The results show a positive effect of full-time
career and high-prestige mothers on children’s participation in BPCH fields. As discussed
previously, the literature suspects that mothers’ employment has an immediate impact on their
children’s educational outcomes (in terms of the reduction of mothering, time spent with children)
or affects children’s behavior. This study shows that mothers’ employment has no effect on

103

children’s STEM participation in the earlier cohort, which did not support the first hypothesis of
the gender asymmetric perspective. But in the most recent cohort- HSLS (2009), a positive effect
is present between mothers’ full-time job and children’s participation in a BPCH major, and
mothers’ high-prestige job and children’s participation in PEMC majors. The results imply that the
promotion effects of mothers’ economic independence and job complexity emerge in the late
cohort.
In the second hypothesis, I anticipate that students’ math self-efficacy and high school
achievement mediate the effects of mothers’ employment on children’s STEM participation. I find
support for this hypothesis. The findings show that mothers’ occupational prestige transmits
positive impacts on STEM participation through children’s school achievement in the recent
cohort. In respect to the mediation process, I also find this beneficial effect of high-prestige
mothers on STEM participation is present from the HS&B, NELS to ELS, which indicates a
positive relation between mothers’ job complexity, children’s cognitive development, and STEM
participation. Likewise, mothers with high-prestige jobs may also have a positive influence on
their children’s self-efficacy, but this effect only appears in later cohorts, such as ELS and HSLS.
Overall, I find that mothers with high-prestige jobs transmit their influence through both school
achievement and math-efficacy, which support both the economic independence hypothesis and
job complexity hypothesis.
It is also worth noting that mothers’ full-time job, however, has negative impacts on
children’s school achievement, resulting in the negative indirect association between full-time
career mothers, school achievement and children’s STEM participation. Such change points to
distinctive family processes that are different between full-time career mothers and high-prestige
career mothers— either higher levels of job complexity or greater economic resources may

104

compensate for potential lack of mothering or involvement in children’s education. This tendency
also reveals a gender difference in how daughters and sons react to mothers’ cognitive ability with
a complex job and greater economic resources with a high-prestige job. More work should be
performed to distinguish the influence of full-time career mothers from high-prestige career
mothers on adolescents’ experiences and development in family processes.
In the third hypothesis, I examine whether the transmission processes of the mothers’
employment on children’s STEM major varies by gender. I find that sons are more sensitive to
mothers’ employment since the HS&B cohort and experience more negative influences during the
NELS cohort. Up until the ELS and HSLS cohort, mothers’ full-time and prestige-jobs are
facilitators that help sons’ STEM entrance, particularly entrance in the PEMC majors. On the other
hand, daughters became sensitive to mothers’ employment at the NELS cohort. School
achievement and mathematics-self-efficacy work to transmit mothers’ influence on children’s
STEM field participation, particularly for the PEMC major field. However, there is no evidence
present for the reinforcement of high prestige mothers on daughters’ STEM in the HSLS cohort.
Furthermore, the structural paths vary by gender and also vary by the fields, such daughters’
preference to attend BPCH field more than sons, and sons’ likelihood to attend PEMC fields is
much higher than daughters’. Gender differences by cohorts on the two processes are concentrated
on the HS&B and NELS cohort, which may imply that children suffer relatively severe gender bias
during those two time periods because of mothers’ employment. For example, sons are more
sensitive to mothers’ employment conditions. Mothers’ full-time employment tends to reduce sons’
school achievement more than it does daughters’.
While this study improves understanding of recent changes in mothers’ employment on
children’s participation in a STEM field, it suffers from some limitations that require further

105

discussion. Although I identify the significance of the difference between cohorts and gender
groups, multi-group SEM analysis is restricted to comparable variables with similar questions
between the four cohorts of the survey. Second, some other family structural factors, such as
multi-generational families or the presence of other relatives at home, may have different
structural constraints in response to mothers’ employment and its consequences. Third, this study
focuses on the effect of mothers’ employment, children’s school achievement and the development
of self-efficacy in a STEM field; little attention was paid to a more comprehensive specification of
variables and models in each cohort. In this study, I do not discuss how the covariates in terms of
family structure, immigrant status, and racial variation may change over time or any potential
interactions among the covariates. Although I control for race and ethnic variables in the model, a
notable racial and immigrant variation occurred between cohorts with respect to their participation
in college or STEM fields. Future research aiming to understand social change in terms of gender
inequality in STEM fields would benefit from a consideration of race/ethnic and immigrant
variation.

106

APPENDICES

107

APPENDIX A

FIGURES FOR CHAPTER THREE

108

Family background
Sibling size
Family SES
Mother education
Female
Race
Home-based
environment

Mother part-time

Family
processes
Parental
school
involvement

School
achievement
12th grade

Family
culture for
school value

PEMC major
AAN major

Mother full-time
Mother occupation
prestige

Family
educational
expectation
Parent
saving effort
for a child

BPCH major
HESB major
Self-math
efficacy
In 10th -12th
grade

2 year after high
school
graduation
Diagram of structural model for the multi-group structural equation modeling analysis on STEM majors
10th grade

12th grade

Figure 9:Theoretical Framework
Note: PEMC: physics, engineering, mathematics, and computer science.
AAN: agricultural, architectural and natural resources sciences.
BPCH: biological sciences, psychology, clinical and health science.
HESB: humanities, education, general curriculum, social and behavior science.

109

100.0%
90.0%
80.0%
70.0%
60.0%
50.0%
40.0%
30.0%
20.0%
10.0%
0.0%

HESB
BPCH
AAN
HS&B(1980)

NELS(1990)

ELS(2002)

HSLS(2009)

HS&B(1980)

Sons

NELS(1990)

ELS(2002)

HSLS(2009)

Daughter

Figure 10: Main Direct Effect of Mother with Full-time Job and Mothers’ Occupational Prestige across Cohorts

110

PEMC

HS&B
1980

Mother full-time

Mother full-time

Mother occupation
prestige

Mother occupation
prestige

School
achievement
12th grade

Self-math efficacy
In 10th -12th grade

School
achievement
12th grade

Self-math efficacy
In 10th -12th grade

No effect
No effect
No effect
-.024*
No effect
No effect

Computer science
Natural science
Health science
Computer science
Natural science
Health science
Computer science

.001*
No effect
No effect
No effect

Natural science
Health science
Computer science

No effect

Natural science

No effect

Health science

Figure 11: Indirect Effect of Mother with Full-time Job and Mothers’ Occupational Prestige
on Children's Participation in STEM Major in the HS&B Cohort

NELS
1990

Mother full-time

Mother full-time

Mother occupation
prestige

Mother occupation
prestige

School
achievement
12th grade

Self-math efficacy
In 10th -12th grade

School
achievement
12th grade

Self-math efficacy
In 10th -12th grade

-.147***
-.081***
No effect
No effect
No effect
No effect

Computer science
Natural science
Health science
Computer science
Natural science
Health science
Computer science

.016***
.009***
.004*
No effect

Natural science
Health science
Computer science

No effect

Natural science

No effect

Health science

Figure 12: Indirect Effect of Mother with Full-time Job and Mothers’ Occupational Prestige
on Children's Participation in STEM Major in the NELS Cohort

111

ELS
2002

Mother full-time

Mother full-time

Mother occupation
prestige

Mother occupation
prestige

School
achievement
12th grade

Self-math efficacy
In 10th -12th grade

School
achievement
12th grade

Self-math efficacy
In 10th -12th grade

-.104***
No effect
-.058***
-.062**
No effect
No effect
.007***
.007***
.004***

Computer science
Natural science
Health science
Computer science
Natural science
Health science
Computer science
Natural science
Health science

.002***

Computer science

No effect
No effect

Natural science
Health science

Figure 13: Indirect Effect of Mother with Full-time Job and Mothers’ Occupational Prestige
on Children's Participation in STEM Major in the ELS Cohort

HSLS
2009

Mother full-time

Mother full-time

Mother occupation
prestige

Mother occupation
prestige

School
achievement
12th grade

Self-math efficacy
In 10th -12th grade

School
achievement
12th grade

Self-math efficacy
In 10th -12th grade

-.046*
No effect
-.029*
-.039**
No effect
No effect
No effect
No effect
No effect

Computer science
Natural science
Health science
Computer science
Natural science
Health science
Computer science
Natural science
Health science

.002***

Computer science

No effect
.002***

Natural science
Health science

Figure 14: Indirect Effect of Mother with Full-time Job and Mothers’ Occupational Prestige
on Children's Participation in STEM Major in the HSLS Cohort

112

APPENDIX B

TABLES FOR CHAPTER THREE

113

Table 12: Descriptive Statistics for the HS&B Teens Living with Married-Two-Biological
Parents families
Mean (%) Std. Dev.
Min
Max
Whether major in STEM
Daughter (Ref. sons)
Number of siblings
SES composite score
The average level of mother
education
The average level of father education
Mother occupation prestige
Mother occupation prestige
% of mother with full-time job
% of mother with part-time job
Student education expectation
Parent education expectation
the average math efficacy
NaĂŻve percentile ranking
Average GPA
Family preparation for school
parental school involvement
Parent saving efforts for children
White
Asian
Black
Non-white Hispanic
Whether as an immigrant

24.88%
0.492
2.898
0.031

0.500
1.725
0.738

0
0
0
-2.658

1
1
6
1.972

2.337

0.985

1

4

2.444
39.881
43.670
42.05%
25.81%
4.151
4.202
0.691
55.825
2.334
9.650
1.473
27.44%
65.63%
3.246%
9.587%
21.538%

1.100
16.737
10.767

7.440%

114

1.701
1.888
0.345
21.133
0.736
1.797
1.255

1
4
0
73.51
0
73.51
0
1
0
1
1
7
1
7
0
1
5.065056 99.76766
1
4
1
12
0
4
0
0
0
0

1
1
1
1

0

1

Table 13: Descriptive Statistics for the NELS Teens Living with Married-Two-Biological
Parents families
Mean (%)
SD
Min
Max
Whether major in STEM
Daughter (Ref. sons)
Number of siblings
SES composite score
The average level of mother
education
The average level of father
education
Mother occupation prestige
Father occupation prestige
% of mother with full-time job
% of mother with part-time job
% of father with full-time job
% of father with part-time job
Student education expectation
Parent education expectation
the average math efficacy
NaĂŻve percentile ranking
Average GPA
Family preparation for school
Parental school involvement
Parent saving efforts for children
White
Asian
Black
Non-white Hispanic
Whether as an immigrant

13.71%
0.515
2.349
0.029

0.500
1.602
0.802

0
0
0
-3.09

1
1
6
2.53

2.417

1.105

1

4

2.505

1.166

1

4

39.194
40.903
49.46%
19.83%
88.45%
2.58%
4.689
4.436
2.974
51.977
2.680
9.788
1.878
37.01%
71.34%

14.274
14.433

0
0
0
0
0
0
1
1
1
0
1
1
0

73.51
73.51
1
1
1
1
7
7
4
100
4
12
4

0

1

0
0
0
0

1
1
1
1

7.970%
6.937%
13.757%
7.137%

115

1.538
1.393
0.547
24.038
0.801
1.746
1.302

Table 14: Descriptive Statistics for the ELS Teens Living with Married-Two-Biological
Parents Families
Mean (%)
SD
Min
Max
Whether major in STEM
Daughter (Ref. sons)
Number of siblings
SES composite score
The average level of mother
education
The average level of father
education
Mother occupation prestige
Father occupation prestige
% of mother with full-time job
% of mother with part-time job
% of father with full-time job
% of father with part-time job
Student education expectation
Parent education expectation
the average math efficacy
NaĂŻve percentile ranking
Average GPA
Family preparation for school
Parental school involvement
Parent saving efforts for children
White
Asian
Black
Non-white Hispanic
Whether as an immigrant

10.33%
0.499
2.044
0.173

0.500
1.351
0.765

0
0
0
-2.11

1
1
6
1.98

2.843

0.998

1

4

2.907

1.028

1

4

45.015
45.493
54.55%
19.78%
88.36%
3.21%
5.324
5.442
1.570
52.221
3.304
9.503
1.390
45.07%
71.06%

16.145
12.167

0
0
0
0
0
0
1
1
0
0
1
1
0

73.51
73.51
1
1
1
1
7
7
3
100
4
12
4

0

1

0
0
0
0

1
1
1
1

10.199%
6.538%
12.204%
10.712%

116

1.360
1.218
0.843
16.096
0.758
2.404
1.390

Table 15: Descriptive Statistics for the HSLS Teens Living with Married-Two-Biological
Parents Families
Mean (%)
SD
Min
Max
Whether major in STEM
Daughter (Ref. sons)
Number of siblings
SES composite score
The average level of mother
education
The average level of father
education
Mother occupation prestige
Father occupation prestige
% of mother with full-time job
% of mother with part-time job
% of father with full-time job
% of father with part-time job
Student education expectation
Parent education expectation
the average math efficacy
NaĂŻve percentile ranking
Average GPA
Family preparation for school
Parental school involvement
Parent saving efforts for children
White
Asian
Black
Non-white Hispanic
Whether as an immigrant

36.35%
0.491
1.515
0.262

0.500
1.409
0.795

0
0
0
-1.93

1
1
6
2.88

2.875

1.016

1

4

2.824

1.055

1

4

46.857
42.198
55.82%
16.40%
75.35%
4.15%
5.451
5.487
1.191
49.622
2.794
12.842
2.038
47.77%
69.79%

12.482
16.727

0
0
0
0
0
0
1
1
0
0
0
1
0

73.7
73.7
1
1
1
1
7
7
3
100
4
12
4

0

1

0
0
0
0

1
1
1
1

8.776%
6.323%
15.107%
6.987%

117

1.564
1.336
0.704
16.210
0.819
2.375
1.263

Table 16: The Distribution of STEM Participation across Cohort of Teens
Physics,
Humanities,
Biology, Psyc Agriculture,
Engineering,
Education,
hology, Clini Architecture
Mathematics,
Social science
c & Health & Natural
and
& Behavior
(BPCH)
Resources
Computer
science
(AAN)
science
(HESB)
Sons

Daughter

Total N

HS&B(1980)

58.15%

6.11%

11.39%

24.35%

2177

NELS(1990)

75.49%

2.99%

5.18%

16.34%

2375

ELS(2002)

76.12%

8.92%

2.03%

12.94%

3061

HSLS(2009)

53.83%

17.37%

2.63%

26.17%

3305

HS&B(1980)

70.74%*

14.86%*

7.39%*

7.00%*

2543

NELS(1990)

89.62%*

1.07%

4.73%

4.58%*

2708

ELS(2002)

72.27%*

22.87%*

1.19%

3.67%*

3516

52.70%

39.50%*

1.64%

6.15%*

3592

HSLS(2009)

* p<0.05, Z-statistics to indicate the gender difference in each cohort
Total valid cases in STEM major outcome in each cohort: HS&B= 4630, NELS= 5083, ELS=6577, HSLS=6897
Overall observations= 23,187

118

Table 17: Math self-Efficacy Beliefs across HS&B, NELS, ELS and HSLS data
Cohort
HS&B

Variables
YB035E
YB035F
YB035G
YB035H

NELS

F1S63D
F1S27A
F1S28A

ELS

BYS89A
BYS89B
BYS89L
BYS89R
BYS89U

HSLS

S2MTESTS
S2MTEXTBOOK
S2MSKILLS
S2MASSEXCL

Indicators
At Ease in math class
Feel tense about math assignment
Math class doesn't scare me
Dread math class
Cronbach’s Alph =.723
Mathematics is one of r's best subjects
Often work hard in math class
Often feel challenged in math class
Cronbach’s Alph =.516
Can do excellent job on math tests
Can understand difficult math texts
Can understand difficult math class
Can do excellent job on math assignments
Can master math class skills
Cronbach’s Alph =.933
Confident can do an excellent job on (spring 2012) math tests
Certain can understand (spring 2012) math textbook
Certain can master skills taught in (spring 2012) math course
Confident can do excellent job on (spring 2012) math assignments
Cronbach’s Alph =.893

119

scale
0--1
0--1
0--1
0--1

Factor
0.782
0.72
0.746
0.707

1--6
1--6
1--6

0.779
0.751
0.487

1--4
1--4
1--4
1--4
1--4

0.874
0.879
0.899
0.896
0.892

1--4
1--4
1--4
1--4

0.901
0.822
0.886
0.889

Table 18: Cohort Structural Weight Invariance Tests
Corrected
Model

△χ (△df)
2

Baseline (Free structural paths and factoral constrained)
Constrained female
Constrained asian
Constrained balck
Constrained hispanic
GPA
Constrained Math self-efficacy
Constrained family supports
Constrained family supports to GPA
Constrained family supports to efficacy
Gender groups
Baseline (Free structural paths and factoral constrained)
Constrained GPA
Constrained Math self-efficacy
Family supports

Number of
Free

Log-likelihood
value a

54.96 (9)*
18.56(9)
21.62(9)
8.30(9)
57.75(9)*
94.63(19)*
124.53(45)
60.51(15)*
67.39(15)*

219
210
210
210
210
210
200
174
204
204

-43872.5
-43927.5
-43891.1
-43894.1
-43880.8
-43930.2
-43947.1
-43997
-43933
-43939.9

77.94(9)*
75.36(9)*
77.91(9)*

109
100
100
100

-30906
-31116.3
-31113.7
-31114.8

* p<0.05
Note. A significant △χvalue indicates that the estimate is non-invariant across groups.
Corrected chi-square test calculate based on the log-likelihood value and the adjusted formula on https://www.statmodel.com/chidiff.shtm

120

Table 19: Final SEM model by Cohort of the High School & Beyond
HS&B
Daughter (Ref. sons)
Mother with full-time job
Mother occupation prestige
The latent of math efficacy (efficacy)
School achievement
Parent education expectation

PEMC major
S.E.
ß
-1.792 (.096) ***
-0.022 (.114)
-0.004 (.003)
0.355 (.049) ***
0.171 (.068) *
-0.05 (.035)
0.085 (.093)

Indirect effect
STEM ← Achievement ← Mother occupation prestige 0.001
STEM ←Achievement ←Mother full-time
-0.024
STEM ←Efficacy ← Mother occupation prestige
0.000
STEM ←Efficacy ←Mother full-time
-0.021
STEM ← parent expectation ← Mother occupation
0.000
STEM ← parent school involvement ← Mother
0.000
STEM ← home preparation for school ← Mother
0.000
STEM ← parent saving efforts ← Mother occupation
0.000

(.000) *
(.010) *
(.000)
(.010)
(.001)
(.000)
(.000)
(.000)

AAN major
S.E.
ß
-1.268 (.092) ***
0.002 (.113)
-0.002 (.003)
0.094 (.043) *
0.073 (.070)
0.009 (.033)
-0.075 (.099)
0.000
-0.010
0.000
-0.006
0.000
0.000
0.000
0.000

(.000)
(.010)
(.000)
(.004)
(.001)
(.000)
(.000)
(.000)

BBPCH major
ß S.E.
0.487 (.108) ***
0.17 (.125)
-0.001 (.003)
0.059 (.046)
-0.044 (.075)
0.038 (.042)
0.051 (.098)
0.000
0.006
0.000
-0.006
0.000
0.000
0.000
0.000

(.000)
(.011)
(.000)
(.004)
(.001)
(.000)
(.000)
(.000)

+ p< .1; * p<.05; ** p<.01; *** p<.001;
Note. Reported standardized coefficient.
a
the measurement of latent factors is omitted from the table.
b
the endogenous variables, 10th/12th-grade math self-efficacy beliefs, school achievement, and family supports, serve as both a dependent and an
independent variable.

121

Table 20: Final SEM model by Cohort of the NELS (1988)
NELS(1988)
Daughter (Ref. sons)
Mother with full-time job
Mother occupation prestige
The latent of math efficacy
School achievement
Parent education expectation
Parent saving efforts
Indirect effect
STEM ← Achievement ← Mother occupation prestige
STEM ←Achievement ←Mother full-time
STEM ←Efficacy ← Mother occupation prestige
STEM ←Efficacy ←Mother full-time
STEM ← parent expectation ← Mother occupation prestige
STEM ← parent school involvement ← Mother occupation
STEM ← home preparation for school ← Mother occupation
STEM ← parent saving efforts ← Mother occupation

PEMC major
ß S.E.
-1.37 (.118) ***
-0.11 (.133)
0.001 (.004)
0.426 (.041) ***
0.434 (.080) ***
-0.117 (.075)
0.208 (.119) †
0.016
-0.147
0.000
0.000
-0.002
0.000
0.000
0.001

(.002) ***
(.034) ***
(.000)
(.000)
(.001)
(.001)
(.000)
(.001) +

AAN major
ß S.E.
-0.337 (.146) *
-0.311 (.177) †
-0.003 (.005)
0.055 (.047)
0.697 (.117) ***
0.223 (.105) *
0.331 (.167) *
0.009
-0.081
0.000
0.001
0.004
0.000
0.000
0.001

(.002) ***
(.023) ***
(.000)
(.004)
(.002) *
(.001)
(.000)
(.001)

BPCH major
ß S.E.
-1.023 (.242) ***
-0.251 (.268)
0.007 (.009)
0.21 (.072) *
-0.059 (.160)
0.036 (.145)
0.682 (.267) *
0.004
-0.038
0.000
0.002
0.000
0.000
0.000
0.003

(.002) *
(.022)
(.000)
(.009)
(.003)
(.001)
(.003)
(.001) *

+ p< .1; * p<.05; ** p<.01; *** p<.001;
Note. Reported standardized coefficient.
a
the measurement of latent factors is omitted from the table.
b
the endogenous variables, 10th/12th-grade math self-efficacy beliefs, school achievement, and family supports, serve as both a dependent and an
independent variable.

122

Table 21: Final SEM model by Cohort of the ELS (2002)
ELS (2002)
Daughter (Ref. sons)
Mother with full-time job
Mother occupation prestige
The latent of math efficacy
School achievement
Parent education expectation
Parent saving efforts
Indirect effect
STEM ← Achievement ← Mother occupation prestige
STEM ←Achievement ←Mother full-time
STEM ←Efficacy ← Mother occupation prestige
STEM ←Efficacy ←Mother full-time
STEM ← parent expectation ← Mother occupation
STEM ← parent school involvement ← Mother
STEM ← home preparation for school ← Mother
STEM ← parent saving efforts ← Mother occupation

PEMC major
S.E.
ß
-1.125 (.142) ***
-0.015 (.166)
0.002 (.005)
0.493 (.072) ***
1.297 (.156) ***
0.026 (.078)
0.171 (.142)
0.007
-0.104
0.002
-0.062
0.000
-0.002
0.000
0.001

(.001) ***
(.027) ***
(.000) ***
(.019) **
(.001)
(.001) *
(.000)
(.001)

AAN major
S.E.
ß
-0.494 (.294)
0.509 (.416)
0.004 (.012)
-0.094 (.150)
1.026 (.309) ***
-0.281 (.186)
-0.273 (.305)
0.007
-0.093
-0.001
0.021
0.000
0.000
0.000
-0.001

(.002) ***
(.047)
(.001)
(.020)
(.001)
(.001)
(.000)
(.001)

BPCH major
S.E.
ß
0.849 (.105) ***
0.055 (.123)
0.001 (.004)
0.049 (.049)
0.653 (.098) ***
0.05 (.057)
0.091 (.102)
0.004
-0.058
0.000
-0.006
0.000
0.000
0.000
0.001

(.001) ***
(.016) ***
(.000)
(.006)
(.001)
(.001)
(.000)
(.001)

+ p< .1; * p<.05; ** p<.01; *** p<.001;
Note. Reported standardized coefficient.
a
the measurement of latent factors is omitted from the table.
b
the endogenous variables, 10th/12th-grade math self-efficacy beliefs, school achievement, and family supports, serve as both a dependent and an
independent variable.

123

Table 22: Final SEM Model by Cohort of the HSLS (2009)
HSLS(2009)
Daughter (Ref. sons)
Mother with full-time job
Mother occupation prestige
The latent of math efficacy
School achievement
Parent education expectation
Parent saving efforts
Indirect effect
STEM ← Achievement ← Mother occupation prestige
STEM ←Achievement ←Mother full-time
STEM ←Efficacy ← Mother occupation prestige
STEM ←Efficacy ←Mother full-time
STEM ← parent expectation ← Mother occupation
STEM ← parent school involvement ← Mother
STEM ← home preparation for school ← Mother
STEM ← parent saving efforts ← Mother occupation

PEMC major
S.E.
ß
-1.386 (.099) ***
-0.044 (.105)
0.007 (.004) *
0.492 (.050) ***
0.719 (.080) ***
0.09 (.048) *
0.026 (.092)
0.005
-0.046
0.002
-0.039
0.002
0.001
0.000
0.000

(.001)
(.018) *
(.001) **
(.016) *
(.000) *
(.001)
(.000)
(.000)

AAN major
S.E.
ß
-0.366 (.196)
0.281 (.247)
-0.003 (.008)
0.022 (.105)
0.691 (.185) ***
0.156 (.108)
0.11 (.213)
0.005
-0.04
0
0.001
0.001
0.001
0.000
0.000

(.001)
(.010)
(.000)
(.008)
(.001)
(.001)
(.000)
(.000)

BPCH major
S.E.
ß
0.718 (.074) ***
0.311 (.087) ***
0.003 (.003)
0.087 (.036) *
0.446 (.062) ***
0.247 (.037) ***
0.059 (.027) *
0.003
-0.029
0.002
-0.006
0.002
0.001
0.000
0.002

(.001)
(.012) *
(.000) *
(.004)
(.000) ***
(.001)
(.000)
(.000) *

+ p< .1; * p<.05; ** p<.01; *** p<.001;
Note. Reported standardized coefficient.
a
the measurement of latent factors is omitted from the table.
b
the endogenous variables, 10th/12th-grade math self-efficacy beliefs, school achievement, and family supports, serve as both a dependent and an
independent variable.

124

Table 23: Final SEM Model by Cohort and Gender (reported structural paths only) in the HS&B (1980) and NELS (1988)
HS&B
ß
-0.107
0.002

S.E.
(.056)
(.001)

Female
ß S.E.
0.019 (.034)
0.002 (.001) *

GPA← Mother occupation prestige -0.246 (.035) ***
GPA← Mother full-time
0.004 (.001) ***

-0.143
0.004

(.037) ***
(.001) ***

-3.461 (1.044) *
0.402 (.032) ***

-3.565 (1.125) *
0.334 (.034) ***

PEMC major ←Mother full-time
PEMC major ←Mother occupation
PEMC major ←GPA
PEMC major ←Efficacy

0.107
0.001
0.066
0.448

(.206)
(.005)
(.124)
(.087) ***

-0.075
-0.005
0.240
0.351

(.142)
(.003)
(.085) ***
(.061) ***

0.119
-0.007
0.049
0.19

(.245)
(.008)
(.006) ***
(.135)

0.001
0.04
-0.076
0.06

(.005)
(.004) ***
(.086)
(.069)

AAN major ←Mother full-time
AAN major ←Mother occupation
AAN major ←GPA
AAN major ←Efficacy

0.009
-0.001
0.037
0.044

(.176)
(.005)
(.110)
(.065)

0.039
-0.004
0.120
0.200

(.155)
(.004)
(.094)
(.063) ***

-0.49
0.003
0.029
0.551

(.233) *
(.008)
(.006) ***
(.112) ***

0.013
-0.01
0.018
-0.219

(.272)
(.008)
(.006) **
(.141)

BPCH major ←Mother full-time
BPCH major ←Mother occupation
BPCH major ←GPA
BPCH major ←Efficacy

0.193
-0.003
-0.106
0.048

(.147)
(.004)
(.090)
(.053)

0.149
0.004
0.079
0.091

(.465)
(.017) *
(.012) *
(.235)

-0.251
0.022
0.006
-0.291

(.330)
(.010)
(.007)
(.174)

Efficacy ←Mother full-time
Efficacy ← Mother occupation

Female
ß S.E.
-0.044 (.052)
.001 (.001)

NELS
Male

(.239)
(.006)
(.140)
(.098)

-0.088
-0.039
0.027
-0.453

Male
ß S.E.
-0.035 (.036)
0.001 (.001)

+ p< .1; * p<.05; ** p<.01; *** p<.001;
Note. Reported standardized coefficient.
a
the measurement of latent factors is omitted from the table.
b
the endogenous variables, 10th/12th-grade math self-efficacy beliefs, school achievement, and family supports, serve as both a dependent and an
independent variable.

125

Table 24: Final SEM Model by Cohort and Gender (reported structural paths only) in the ELS (2002) and HSLS (2009)
ELS

Efficacy ←Mother full-time
Efficacy ← Mother occupation

Female
ß S.E.
-0.139 (.050) *
0.003 (.001) *

Male

HSLS
Male
ß S.E.
-0.112 (.045) *
0.005 (.001) **

ß
-0.128
0.004

S.E.
(.051) *
(.001) **

Female
ß S.E.
-0.038 (.045)
0.002 (.001)

-0.091
0.007

(.031) *
(.001) ***

-0.074 (.033) *
0.006 (.001) ***

-0.062 (.033)
0.008 (.001) ***

0.045
0.001
1.202
0.432

(.163)
(.005)
(.135) ***
(.085) ***

-0.018 (.199)
0.009 (.007)
1.409 (.206) ***
-0.547 (.094) ***

-0.098 (.120)
0.008 (.004)
0.64 (.086) ***
0.469 (.058) ***

GPA← Mother occupation prestige
GPA←Mother full-time

-0.093 (.031) *
0.006 (.001) ***

PEMC major ←Mother full-time
PEMC major ←Mother occupation
PEMC major ←GPA
PEMC major ←Efficacy

0.066
-0.009
0.994
0.601

(.284)
(.008)
(.281) ***
(.124) ***

AAN major ←Mother full-time
AAN major ←Mother occupation
AAN major ←GPA
AAN major ←Efficacy

-0.4
-0.007
0.977
-0.073

(.516)
(.013)
(.491) *
(.240)

0.787
-0.011
1.072
-0.399

(.427)
(.012)
(.289) ***
(.201)

0.206 (.349)
0.009 (.012)
0.915 (.313) **
-0.084 (.164)

0.135 (.319)
-0.006 (.010)
0.501 (.223) *
-0.279 (.139) *

BPCH major ←Mother full-time
BPCH major ←Mother occupation
BPCH major ←GPA
BPCH major ←Efficacy

0.147
-0.003
0.516
0.104

(.124)
(.004)
(.098) ***
(.054) *

0.074
.0000
0.86
-0.094

(.193)
(.006)
(.148) ***
(.093)

0.325 (.104) ***
0.001 (.003)
0.387 (.077) ***
-0.064 (.042)

0.163 (.143)
0.007 (.005)
0.594 (.100) *
0.134 (.065) *

+ p< .1; * p<.05; ** p<.01; *** p<.001;
Note. Reported standardized coefficient.
a
the measurement of latent factors is omitted from the table.
b
the endogenous variables, 10th/12th-grade math self-efficacy beliefs, school achievement, and family supports, serve as both a dependent and an
independent variable

126

Table 25: Summary of Positively Significant Indirect Effect between Mother Employment, Math-Efficacy and School Average
GPA by Girls and Boys Group over Cohort
FEMALE
MALE
PEMC major ←GPA← Mother full-time (+)
AAN major ←Efficacy ←Mother full-time (+)

HS&B
NELS

ELS

AAN major ←Efficacy ← Mother occupation prestige(+) AAN major ←GPA← Mother full-time(+)
PEMC major ←GPA← Mother full-time(+)
AAN major← GPA← Mother full-time(+)
BPCH major ←GPA ←Mother full-time(+)
PEMC major ←Efficacy ← Mother occupation
prestige(+)
PEMC major ←GPA← Mother full-time(+)
AAN major← GPA← Mother full-time(+)
BPCH major ←GPA← Mother full-time(+)

PEMC major ←Efficacy ← Mother occupation
prestige(+)
PEMC major ←GPA← Mother full-time(+)
AAN major← GPA← Mother full-time(+)
BPCH major ←GPA← Mother full-time(+)

PEMC major ←GPA← Mother full-time(+)
AAN major← GPA← Mother full-time(+)
BPCH major ←GPA← Mother full-time(+)

PEMC major ←Efficacy ← Mother occupation
prestige(+)
PEMC major ←GPA← Mother full-time(+)
AAN major← GPA← Mother full-time(+)
BPCH major ←GPA← Mother full-time(+)

HSLS

+ p<0.10, * p<0.05, ** p<0.01, *** p<0.001
Note. Reported standardized coefficient. Gender group difference test:△χvalue of female versus male > 3.84.

127

APPENDIX C

SUPPLEMENTAL TABLES

128

Table 26: Four STEM Categories
2
digits
Category
CIP
codes
PEMC
major

AAN
major

BPCH
major

HESB

Family of fields

11

Computer and Information Sciences and Support Services

14
27
40

Engineering
Mathematics and Statistics
Physical Sciences

1

Agriculture, Agriculture Operations, and Related Services

3
4

Natural Resources and Conservation
Architecture and Related Services

26

Biological and Biomedical Sciences

42
51
9
13
16
19
23
24
30
31
38
45
49
50
52

Psychology
Health Professions and Related Clinical Sciences
Communication, Journalism, and Related Programs
Education
Foreign Languages, Literatures, and Linguistics
Family and Consumer Sciences/Human Sciences
English Language and Literature/Letters
Liberal Arts and Sciences, General Studies and Humanities
Multi/Interdisciplinary Studies
Parks, Recreation, Leisure, and Fitness Studies
Philosophy and Religious Studies
Social Sciences
Transportation and Materials Moving
Visual and Performing Arts
Business, Management, Marketing, and Related Support Services
History (New): Study and Interpretation of Past Events, Institutions, Issues,
and Cultures
General Curriculum

54
99

129

Table 27: Average Demographic Characteristics in Each Survey
(Including all types of families)
Whether major in STEM
Daughter (Ref. sons)
Number of sibling
SES composite score
The average level of mother education
The average level of father education
Mother occupation prestige
Father occupation prestige
% of mother with full-time job
% of mother with part-time job
%White
%Asian
%Black
%Non-white Hispanic
NES Cohort
Whether major in STEM
Daughter (Ref. sons)
Number of sibling
SES composite score
The average level of mother education
The average level of father education
Mother occupation prestige
Father occupation prestige
% of mother with full-time job
% of mother with part-time job
%White
%Asian
%Black
%Non-white Hispanic

130

Mean
0.24

Std.
(.429)

0.50
2.98
-0.07
2.30
2.42
40.19
43.24
47%
24%
70%
3%
14%
22%

(.500)
(1.776)
(.758)
(.991)
(1.094)
(16.215)
(10.676)

0.13
0.52
2.49
-0.05
2.38
2.42
38.87
39.47
53%
18%
70%
7%
10%
13%

(.334)
(.499)
(1.684)
(.812)
(1.097)
(1.164)
(14.035)
(15.488)

Table 27 (cont’d)
ELS(2002)
Whether major in STEM
Daughter (Ref. sons)
Number of sibling
SES composite score
The average level of mother education
The average level of father education
Mother occupation prestige
Father occupation prestige
% of mother with full-time job
% of mother with part-time job
%White
%Asian
%Black
%Non-white Hispanic

0.09
0.50

(.280)
(.500)

2.30
0.04
2.75
2.77
44.16
44.23
58%
17%
65%
9%
12%
14%

(1.531)
(.753)
(.994)
(1.041)
(15.522)
(12.018)

0.29
0.49
1.63
0.05
2.78
2.78
43.79
34.64
0.29
55%
14%
65%
8%
10%
16%

(.456)
(.500)
(1.517)
(.780)
(1.018)
(1.056)
(15.744)
(21.868)
(.456)

HSLS Cohort

Whether major in STEM
Daughter (Ref. sons)
Number of sibling
SES composite score
The average level of mother education
The average level of father education
Mother occupation prestige
Father occupation prestige
Whether major in STEM
% of mother with full-time job
% of mother with part-time job
%White
%Asian
%Black
%Non-white Hispanic

131

Table 28: Mother Occupation and Occupational Prestige Score
HS&B/ NELS/ ELS/HSLS
Mother/female guardian's current/most recent occupation: 2-digit ONET codea, b
Occupation

Categories

Occupation
prestige score

0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16

NO JOB
CLERICAL
CRAFTSMAN
FARMER, FARM MNGR
HOMEMAKER
LABORER
MANAGER, ADMNTR
MILITARY
OPERATIVE
PROFESSIONAL
PROFESSIONAL,DOCTOR
PROPRIETOR,OWNER
PROTECTIVE SERVICE
SALES
SCHOOL TEACHER
SERVICE
TECHNICAL

0
38.15
38.51
35.57
33.93
33.38
53.52
33.09
35.94
62.24
64.38
44.15
48.4
35.77
73.51
34.95
40.43

Note: a. Parents’ occupation coding scheme see using 2-digit ONET code to align with HS&B code. In ELS
and HSLS data also provide 6-digit ONET code for more category description.
b.
The ONET Database is in compliance with the mandate that all federal agencies collecting occupational
information use the Standard Occupational Classification System external site (SOC). The ONET Database
uses the basic 6-digit numerical coding structure of the SOC as its framework, adding a 2-digit extension
(sequentially numbered beginning with ".01") to differentiate unique ONET occupations within the SOC
system.

132

REFERENCES

133

REFERENCES

Anderson, Eugene & Kim, Dongbin. (2006). Increasing the success of minority students in science
and technology. American Council on Education, Washington, DC.
Archer, L., DeWitt, J., Osborne, J., Dillon, J., Willis, B. & Wong, B. (2012). Science Aspirations,
Capital, and Family Habitus: How Families Shape Children’s Engagement and Identification
With Science. American Educational Research, Journal 49, 881-908.
Anderson, J. C., & Gerbing, D. W. (1988). Structural equation modeling in practice: A review and
recommended two-step approach. Psychological bulletin, 103(3), 411-423.
Augustine, J. M., Cavanagh, S. E., & Crosnoe, R. (2009).Maternal Education, Early Child Care
and the Reproduction of Advantage. Social Forces, 88(1), 1-29.
Augustine, J. M. (2014). Mothers’ Employment, Education, and Parenting. Work and Occupations,
41(2), 237-270.
Aughinbaugh, A., & Gittleman, M. (2004).Maternal employment and adolescent risky behavior.
Journal of Health Economics, 23(4), 815-838.
Baker, M., & Milligan, K. (2016). Boy-Girl Differences in Parental Time Investments: Evidence
from Three Countries. Journal of Human Capital, 10(4), 399-441.
Baum, C. L. (2004). The long-term effects of early and recent maternal employment on a child's
academic achievement. Journal of Family Issues, 25(1), 29-60.
Blau, P. M. & Duncan, O. D. (1967).The American Occupational Structure. New York: Wiley and
Sons.
Bleeker, M. A., & Jacobs, J. E. (2004). Achievement in math and science: Do mothers' beliefs
matter 12 years later? Journal of Educational Psychology, 96(1), 97-109.
Ceci, S. J., Williams, W. M., & Barnett, S. M. (2009). Women’s Underrepresentation in Science:
Sociocultural and Biological Considerations. Psychological Bulletin, 135(2), 218-261.
Ceci, S. J. & Williams, W. M. (2011). Understanding current causes of women’s
underrepresentation in science. Proc Natl Acad Sci USA, 108, 3157-3162.
Coleman, J. S. (1988). Social Capital in the Creation of Human Capital. American Journal of
Sociology, 94, S95-S120.
Chin, T., & Phillips, M. (2004). Social Reproduction and Child-Rearing Practices: Social Class,

134

Children’s Agency, and the Summer Activity Gap. Sociology of Education, 77(3), 185-210.
Crosnoe, R. (2004). Social Capital and the Interplay of Families and Schools. Journal of Marriage
and Family, 66, 267-280.
Crosnoe, R. (2009). Family-School Connections and the Transitions of Low-Income Youth and
English Language Learners from Middle School into High School. Developmental
Psychology, 45, 1061-1076.
Cooksey, E. C., Menaghan, E. G., & Jekielek, S. M. (1997). Life-course effects of work and family
circumstances on children. Social Forces, 76(2), 637-665.
DiPrete, T. A., & Jennings, J. L. (2012).Social and behavioral skills and the gender gap in early
educational achievement. Social Science Research, 41(1), 1-15.
DiPrete, T. A., & Buchmann, C. (2013). The Rise of Women: The Growing Gender Gap in
Education and What It Means for American Schools. New York: Russell Sage Foundation.
Dufur, M. J., Parcel, T. L., &Troutman. K. (2013). Does capital at home matter more than capital at
school? Social capital effects on academic achievement. Research in Social Stratification and
Mobility, 31, 1-21.
Eccles, J. S., & Roeser, R. W. (2011). Schools as Developmental Contexts During Adolescence.
Journal of Research on Adolescence, 21(1), 225-241.
Eccles, J. S. (2015). Gendered socialization of STEM Interests in the Family. International Journal
of Gender, Science and Technology, 7(2), 117-132.
Fomby, P. (2013). Family Instability and College Enrollment and Completion. Population
Research and Policy Review, 32, 469-494.
Fox, L., Han, W. J., Ruhm, C., &Waldfogel, J. (2013). Time for Children: Trends in the
Employment Patterns of Parents, 1967-2009. Demography, 50(1), 25-49.
Friedkin, N. E., & Thomas, S. L. (1997).Social Positions in Schooling. Sociology of Education,
70(4), 239-255.
Goldberg, W. A., Prause, J., Lucas-Thompson, R., &Himsel, A. (2008). Maternal employment and
children’s achievement in context: a meta-analysis of four decades of research. Psychological
Bulletin, 134(1), 77-108.
Guiso, L., Monte, F., Sapienza, P., &Zingales, L. (2008).Culture, Gender, and Math.Science,
320(5880), 1164.
Haveman, R., Wolfe, B., & Spaulding, J. (1991).Childhood Events and Circumstances Influencing
High-School Completion. Demography, 28(1), 133-157.

135

Hout, M., Smith, T. W. & Marsden, P. V. (2010). Prestige and Socioeconomic Scores for the 2010
Census Codes GSS Methodological Report No. 124.
Hayes, A. F. (2013). Introduction to mediation, moderation, and conditional process analysis: A
regression-based approach: Guilford Press.
Johnson, R. C., Kalil, A., & Dunifon, R. E. (2012). Employment patterns of less-skilled workers:
links to children's behavior and academic progress. Demography, 49(2), 747-772.
Kalmijn, M. (1994).Mother’s Occupational Status and Children's Schooling. American
Sociological Review, 59(2), 257-275.
Kline, R. B. (2011). Principles and Practice of Structural Equation Modeling, (3rd. Ed ed.). New
York: Guilford Press.
Kim, D. H. & Schneider, B. (2005). Social Capital in Action: Alignment of Parental Support in
Adolescents’ Transition to Postsecondary Education. Social Forces, 84, 1181-1206.
Kulis, S., Sicotte, D., & Collins, S. (2002). More than a Pipeline Problem: Labor Supply
Constraints and Gender Stratification across Academic Science Disciplines. Research in
Higher Education, 43(6), 657-691.
Perez-Felkner, L., McDonald, S. K., Schneider, B., & Grogan, E. (2012). Female and Male
Adolescents' Subjective Orientations to Mathematics and The Influence of Those
Orientations on Postsecondary Majors. Developmental Psychology, 48(6), 1658-1673.
Morgan, S. L. (2005). On the Edge of Commitment: Educational Attainment and Race in the
United States. Stanford: Stanford University Press.
Morgan, S. L., Gelbgiser, D., & Weeden, K. A. (2013).Feeding the pipeline: Gender, occupational
plans, and college major selection. Social Science Research, 42(4), 989-1005.
MuthĂŠn, L. K.,& MuthĂŠn, B. O. (2012). Mplus User's Guide. Los Angeles, CA: MuthĂŠn& MuthĂŠn.
Nakao, Keiko, & Judith Treas. (1990).Computing 1989 Occupational Prestige Scores. GSS
Methodological Report MR70.
http://publicdata.norc.org:41000/gss/DOCUMENTS/REPORTS/Methodological_Reports/M
R070.pdf
Nakao, Keiko, & Judith Treas. (1994). Updating Occupational Prestige and Socioeconomic Scores:
How the New Measures Measure Up. Sociological Methodology, 24, 1-72.
Offer, S. & Schneider, B. (2007).Children's Role in Generating Social Capital. Social Forces, 85,
1125-1142.

136

Parcel, T. L., & Menaghan, E. G. (1990).Maternal Working-Conditions and Childrens Verbal
Facility - Studying the Intergenerational Transmission of Inequality from Mothers to
Young-Children. Social Psychology Quarterly, 53(2), 132-147.
Parcel, T. L., & Dufur, M. J. (2001). Capital at home and at school: Effects on child social
adjustment. Journal of Marriage and Family, 63(1), 32-47.
Parcel, T. L., &Menaghan, E. G. (1994).Parents’ jobs and children’s lives. New York: Aldine de
Gruyter.
Raley, S., & Bianchi, S. (2006). Sons, daughters, and family processes: Does gender of children
matter? Annual Review of Sociology, 32, 401-421.
Ruhm, C. J. (2004). Parental Employment and Child Cognitive Development. The Journal of
Human Resources, 39(1), 155-192.
Tyson, W., Lee, R., Borman, K. M., & Hanson, M. A. (2007). Science, Technology, Engineering,
and Mathematics (STEM) Pathways: High School Science and Math Coursework and
Postsecondary Degree Attainment. Journal of Education for Students Placed at Risk, 12(3),
243-270.
Wang, M.-T., Eccles, J. S., & Kenny, S. (2013). Not Lack of Ability but More Choice: Individual
and Gender Differences in Choice of Careers in Science, Technology, Engineering, and
Mathematics. Psychological Science, 24(5), 770–775.
Wang, X. (2013). Why Students Choose STEM Majors: Motivation, High School Learning, and
Postsecondary Context of Support. American Educational Research Journal, 50(5),
1081-1121.
Xie, Y., Fang, M., & Shauman, K. (2015). STEM Education. Annual Review of Sociology, 41,
331-357.

137