Sociology of Education

1–19
From Major Preferences to Ó American Sociological Association 2019
DOI: 10.1177/0038040719887971
Major Choices: Gender and journals.sagepub.com/home/soe

Logics of Major Choice

Natasha Quadlin1

Abstract
Research shows that college students choose majors for a variety of reasons. Some students are motivated
by potential economic returns, others want to take engaging classes, and others still would like opportu-
nities to help people in their jobs. But how do these preferences map onto students’ actual major choices?
This question is particularly intriguing in light of gender differences in fields of study, as men and women
may take divergent pathways in pursuit of the same outcome. Using data from the Pathways through Col-
lege Study (N = 2,639), I show that men and women choose very different majors even when they cite the
same major preferences—what I call gendered logics of major choice. In addition, I use earnings data from
the American Community Survey to assess how these gendered logics of major choice may be associated
with broader patterns of earnings inequality. I find that among men and women who have the same major
preferences, men’s major choices are tied to significantly higher prospective earnings than women’s major
choices. This finding demonstrates that the ways men and women translate their preferences into majors
are unequal from an earnings perspective. Implications for research on higher education and gender are
discussed.

Keywords
higher education, major choice, gender

Major choice is one of the most important deci- weigh fields of study using different criteria.
sions students make in college. Students in differ- Some students have intellectual motivations and
ent majors take different classes that expose them are primarily drawn to majors they find personally
to different content and, in many cases, require engaging (Mullen 2014). Others are more moti-
different amounts of critical thinking (Arum and vated by economic returns and seek out majors
Roksa 2011). Research also highlights the earn- that will allow them to maximize their earnings
ings differentials between individuals with bache- and keep their career options open (Davies and
lor’s degrees in different fields of study. Depend- Guppy 1997; Quadlin 2017). Still others hope to
ing on the majors they choose, students may have identify majors that will allow them to make a dif-
vastly different economic returns throughout their ference in the world. These preferences are not
lifetimes (Kim, Tamborini, and Sakamoto 2015).
For these reasons, major choice affects students’
1
lives not only during college but also many years Ohio State University, Columbus, OH, USA
after college completion (Armstrong and Hamilton Corresponding Author:
2013; Charles and Bradley 2009; DiPrete and Natasha Quadlin, Assistant Professor, Department of
Buchmann 2013; Riegle-Crumb et al. 2012). Sociology, Ohio State University, 238 Townshend Hall,
When choosing a major, students may have 1885 Neil Avenue Mall, Columbus, OH 43210, USA.
different motivations for their decisions and may Email: [email protected]
2 Sociology of Education XX(X)

mutually exclusive; many students hope to find the incorporate prospective earnings data from the
one major that checks all of these boxes, but each American Community Survey (ACS), and I also
major presumably fulfills some criteria better than include an analysis of students’ own expected
others. Precisely how these links operate, earnings from the PtC data. These analyses exam-
however—that is, from students’ preferences to ine whether, given the same major preferences,
the specific majors they see as fulfilling those men and women choose majors with divergent
preferences—has yet to be examined empirically. potential economic returns. If this is the case,
The question of how college students translate then these gendered logics of major choice may
their preferences into actual majors is especially be an important factor that contributes to income
intriguing in light of gender differences in fields inequality among college-educated workers.
of study. A long line of research in the social sci-
ences shows that men and women choose different
college majors (Alon and DiPrete 2015; Bradley
2000; Charles and Bradley 2002; Correll 2001;
BACKGROUND
England and Li 2006; Jacobs 1996; Morgan, Gelb- What Preferences Do Students Have
giser, and Weeden 2013). This literature demon-
strates that gender shapes students’ choice sets,
for Their Majors?
often affecting whether students see majors as Much research in the social sciences has consid-
open or closed to them. It is a distinct possibility, ered how students choose a major. Prior studies
then, that men and women translate their preferen- have focused on many preferences students may
ces into major choices differently. Men may have, but I focus on three preferences that have
believe that certain majors will maximize their received perhaps the most attention in the literature:
economic returns, and women may believe that economic returns, intellectual engagement, and
entirely different majors are better suited for opportunities to help others. These preferences are
achieving this same goal. This is potentially not mutually exclusive; students may try to find
a key mechanism in the relationship between gen- one major that fulfills all these criteria for them.
der and major choice. Research often suggests that Yet, some students may focus on one or more of
shifting students’ preferences is key to equalizing these preferences in their major choice—and the
gender differences in major choice—that women tendency to do so may be patterned along socio-
will be more likely to choose engineering if they demographic lines.
care about money, for example, or men will be The first factor is economic returns, which
more likely to choose education if they care about some research calls ‘‘extrinsic rewards’’ because
helping people. However, if men and women take these are usually tangible benefits students receive
different pathways even when they have the same after completing their education (Ma 2009; Marini
preferences for their major, then attempts to et al. 1996). As the cost of attending college has
change students’ preferences may fare poorly as increased, and the debts some students incur
a solution to equalizing gender gaps in major have risen alongside these costs (Dwyer 2018;
choice. Dwyer, McCloud, and Hodson 2012; Houle
This article addresses two interrelated ques- 2014; Quadlin and Rudel 2015), scholars and
tions: How are college students’ preferences for members of the public are beginning to pay
majors associated with their actual major choices? more attention to the economic returns tied to col-
And to what extent do these links differ for men lege majors. Research shows that bachelor’s
and women? To answer these questions, I use degrees in different fields of study can yield dra-
data from the Pathways through College Study matically different earnings over a lifetime. In
(PtC). The PtC data were collected at three post- fact, the earnings gaps between college graduates
secondary institutions that are diverse in terms of in different majors (i.e., within college) are some-
size, geography, and public/private status. To my times larger than that between high school and col-
knowledge, this is the only large-scale national lege graduates (i.e., between education; Kim et al.
survey with extensive data on students’ preferen- 2015). Degrees in STEM (science, technology,
ces for their majors. In addition to these more engineering, and mathematics) fields, business,
descriptive questions about major choice, I also and health are generally tied to higher earnings,
assess how men’s and women’s major preferences and degrees in the humanities and education are
are associated with projected earnings. To do so, I tied to lower earnings, with the more heterogeneous
Quadlin 3

social sciences in between (Kim et al. 2015). Stu- important criterion because many students use
dents may be aware of these earnings differences their major choice as a precursor to their career
between majors and may use them to help guide choice (Roksa and Levey 2010), and helping others
their major choice (Cebula and Lopes 1982; Davies is a frequently cited way to achieve meaning in
and Guppy 1997; Quadlin 2017). one’s job (Herzog 1982; Lueptow 1980; Marini
In addition to potential earnings as a component et al. 1996). Helping others can take many forms,
of economic returns, students may also consider from teaching or helping children to advocating
the employability of a major, or the extent to for underserved populations to advising on legal
which a major will allow them to keep their career or economic matters. Students may choose different
options open. Employability and projected earn- fields of study depending on the extent to which
ings typically go hand in hand, but they are occa- they weigh altruistic rewards in their major choice
sionally at odds with each other. For example, as well as their preferred method of helping.
some research shows that degrees in applied In summary, research often assesses the factors
fields, such as nursing and education, are associ- students consider when choosing their college
ated with good employment prospects and rela- major—and, in turn, which students are more and
tively high earnings in the early career but low less likely to prioritize these factors. But most
occupational status growth over time (Roksa and research stops here and does not consider how these
Levey 2010). This pattern implies that nursing preferences are associated with the majors students
and education may be more favorable in terms actually choose. This is a key omission because the
of employability but less favorable in terms of pro- majors students perceive as having good economic
jected earnings. Some students may be attuned to returns, employability, intellectual engagement,
employability in their major choice, particularly and opportunities to help others can vary enor-
if they have substantial debts and otherwise lack mously from student to student. These perceptions
a safety net from family or other sources (Quadlin may also be patterned by students’ sociodemo-
2017). But regardless of whether students recog- graphic characteristics—especially gender.
nize the distinction between earnings and employ-
ability, this is an added component of economic
returns that may factor into major choice.
Scholars have also considered the extent to
Gender and Logics of Major Choice
which students prioritize ‘‘intrinsic rewards,’’ or The question of how students’ preferences are
intangible benefits received during or after col- associated with major choice is particularly salient
lege. One commonly cited intrinsic reward is intel- in light of gender. Theories of gender emphasize
lectual engagement. College is often billed as the that expectations about how men and women
best time of one’s life (Armstrong and Hamilton should behave guide decision making in social
2013) not only because of the social side of col- life. Role congruity theory, for example, posits
lege but also because students can engage with that men and women are expected to have differ-
a wide range of academic subjects. Many students ent social roles that align with gendered stereo-
take this opportunity to choose a major that inter- types. Men are expected to occupy agentic roles,
ests them on an intellectual level—especially, as which project power and confidence, whereas
some studies show, if students are relatively priv- women are expected to occupy communal roles,
ileged. Privileged students can afford to weigh which emphasize teamwork and likeability (Eagly
subject matter more heavily than economic returns and Karau 2002). Consistent with role congruity
when choosing their major, presumably because theory, studies show that men and women cite dif-
they have a safety net that can protect them against ferent reasons for choosing majors and careers.
poor employment prospects (Goyette and Mullen Women typically place more emphasis on intrinsic
2006; Quadlin 2017). Yet, most students consider rewards, such as helping others and working with
intellectual engagement on some level, even if it is people, while men typically place more emphasis
not the most important factor. on extrinsic rewards, such as economic returns
A final factor is opportunities to help others, (Konrad et al. 2000; Ma 2009; Zafar 2013).
which is usually categorized with a subset of Some research challenges these premises and
intrinsic rewards known as ‘‘altruistic rewards’’ shows that women today prioritize extrinsic
(Ma 2009; Marini et al. 1996). This is an rewards more than they once did, rendering gender
4 Sociology of Education XX(X)

differences in extrinsic reward seeking small or help others (which, as discussed earlier, women
nonexistent (Marini et al. 1996). In addition, gen- often cite as a priority).
der scholars argue that men’s and women’s prefer- The present study builds on earlier research to
ences are constrained by broader demands, such as assess how men and women translate their prefer-
their need for greater workplace flexibility (Blair- ences into actual major choices. This type of
Loy 2003)—a theme I will refer to throughout this inquiry is needed to understand how college stu-
article. Yet, all of this is ultimately separate from dents draw on gendered choice sets as they begin
whether men and women select the same majors to pursue their life goals. Prior research suggests
given the same stated preferences. that reshaping men’s and women’s preferences is
Prior research suggests men’s and women’s an important priority if we are to close gender
major choices are likely to diverge under these cir- gaps in major choice. This study assesses whether
cumstances because they have very different this is the case—or if, alternatively, equalizing
choice sets—that is, the set of fields they see as men’s and women’s preferences is not enough to
open to and appropriate for them.1 Data on U.S. overcome broader gendered constraints in major
bachelor’s recipients show that men are most over- choice.
represented in engineering, physics, computer sci-
ence, and economics, and women are most over-
represented in nursing, education, psychology, Major Preferences, Gender, and
sociology, and English (England and Li 2006).
Much research focuses on gender and STEM Projected Earnings
fields because women are unlikely to choose these In addition to the more descriptive analyses that
majors despite their strong economic returns as predict men’s and women’s major choices, I also
well as the concerted efforts that have been examine how students’ major preferences are
made to increase gender diversity in STEM (Mor- associated with their prospective earnings (mostly
gan et al. 2013; Xie and Shauman 2005). This is using ACS data). These analyses consider whether
not to say women do not consider majoring in gendered logics of major choice potentially con-
engineering, or men do not consider majoring in tribute to income inequality among college gradu-
nursing; research has called attention to people’s ates. Earnings certainly are not the only outcome
experiences as ‘‘tokens’’ in organizations (Kanter that matters for college graduates. College is
1977) and in college majors in particular (Sax also understood as a time for students to develop
1996). Yet, the pool of majors students choose critical thinking skills, make friends, nurture tal-
from is often shaped by gendered notions about ents and interests, and broaden their worldviews
what a ‘‘good’’ or ‘‘appropriate’’ major would be (McCabe 2016; Stevens, Armstrong, and Arum
(Cech 2013; Charles and Bradley 2009). 2008). But because many students and members
Although little research examines the relation- of the public emphasize the economic value of
ships between major preferences and major a college degree, it is important to consider
choices, some studies provide clues as to how whether students make trade-offs between pro-
these patterns might unfold (and, further, how jected earnings and other aims when choosing
these patterns might differ for men and women). a major—and whether these trade-offs are more
These studies focus mostly on STEM majors and pronounced among men or women. For example,
careers, and especially the extent to which stu- even if men and women both emphasize engaging
dents associate intrinsic rewards with STEM. classes in their major choice, they may still choose
Diekman and colleagues (2010) show that women different majors because they have different ideas
perceive STEM careers as impeding communal about what constitutes an engaging class. If the
goals (e.g., helping others) more so than other fields men choose under these circumstances are
careers, and these perceptions partially account associated with higher earnings than the fields
for gender differences in STEM interest. Other women choose, then women may make greater
work finds that girls report greater intention to financial sacrifices in their pursuit of engaging
enroll in STEM if they see these fields as socially classes. This is a potentially important source of
relevant, yet there is no such relationship for boys inequality that has been underexamined in prior
(Kyte and Riegle-Crumb 2017). Taken together, research.
these studies suggest that women may not view To assess these dynamics, I use data primarily
STEM majors favorably if they are motivated to from the PtC, an extensive multisite data set
Quadlin 5

designed to capture students’ experiences in Preferences for Major Choice

higher education. The PtC data set includes rich
information about students’ majors, as well as stu- The main independent variables are students’ pref-
dents’ preferences for their majors, making this erences for their major choice. In the first-term
the only large-scale national data set with suffi- interview, respondents were asked (emphasis in
cient information to conduct this study. Addition- the original): ‘‘How important is each of the fol-
ally, I incorporate prospective earnings by match- lowing to you when choosing a college major?’’
ing students’ majors with earnings data from the Respondents rated six factors on a scale of 1
ACS. Linking these two data sets allows me to (not at all important) to 5 (extremely important):
assess how students’ major preferences are associ- (1) the amount of money earned right after col-
ated with the earnings they can reasonably expect lege, (2) the amount of money earned over the
to receive. course of their career, (3) keeping their career
options open for now, (4) engaging entry-level
classes, (5) engaging advanced classes, and (6)
helping other people in their job or career.4 I com-
DATA AND METHODS bined the two ‘‘money’’ items and the two ‘‘class-
Students were first interviewed for the PtC during es’’ items because they address similar topics and
their first term, in fall 2014 (N = 2,720), and behave similarly in the analyses. As such, the
follow-up interviews occurred in the first-year responses for these items range from 2 to 10,
spring, second-year fall, second-year spring, and and responses for the ‘‘career options’’ and ‘‘help-
third-year spring (Grodsky and Muller 2018). ing others’’ items range from 1 to 5.
Respondents were drawn from three institutions Table 2 shows descriptive statistics for these
selected because they were diverse in terms of major preferences. The means are similar across
size, geography, and public/private status and items, with the lowest point estimate for engaging
because they had developed programs intended to entry-level classes (3.45) and the highest point esti-
attract and retain STEM majors.2 Thus, one purpose mate for helping others (3.96). These average
of the PtC data set is to assess the efficacy of these scores obscure some key gender differences, how-
programs—but, as noted earlier, the PtC data set is ever. Men have higher means for short-term earn-
also uniquely situated for research on the dynamics ings (p \ .001), long-term earnings (p \ .001),
of major choice because it is one of the only and the combined earnings item (p \ .001). Addi-
national data sets with information on the factors tionally, women have a higher mean for helping
students considered when choosing majors. others (p \ .001). This pattern is consistent with
Respondents were selected using stratified proba- past research showing men are attracted to majors
bility samples of first-time first-year students, mak- and careers that offer extrinsic rewards, whereas
ing the sample representative of the populations of women are more likely to seek intrinsic or altruistic
entering students at the three institutions. rewards (Ma 2009; Zafar 2013). Of course, these
Because the PtC data set places special empha- preferences also reflect other gendered con-
sis on STEM majors, students who majored in these straints—for example, jobs that are known to
fields are well represented, as outlined in Table 1. have high extrinsic rewards also tend to involve
In their first term, 30 percent of students (and 36 long hours and poor workplace flexibility (Acker
percent of women) were biology/premedicine 1990; Blair-Loy 2003; Williams, Muller, and
majors, 28 percent (and 44 percent of men) were Kilanski 2012), so women may place less emphasis
engineering majors, and physics/math majors are on extrinsic rewards because they do not view these
also overrepresented compared to national esti- workplaces as compatible with their life goals.
mates.3 For this reason, these data are especially One challenge in using these data is that the
well suited for examining factors that predict entry causal ordering between major preferences and
intro STEM majors. This is not to say the data set is major choices is not always clear. Prior research
unable to assess majors other than STEM; indeed, contends that students begin with a sense of their
many of the factors examined in the data set are priorities and then select majors and careers that
related to the selection of non-STEM majors, as align with those priorities (Cebula and Lopes
discussed in the Results. Yet, the large population 1982; Davies and Guppy 1997; Diekman et al.
of STEM majors is a strength of the PtC data that 2010; Kyte and Riegle-Crumb 2017), so this is
is unmatched by similar data sets. the causal pathway I emphasize. However, it is
6 Sociology of Education XX(X)

Table 1. Descriptive Statistics.

Mean (standard deviation) or percent

Variable Full sample Men Women

Initial major
Physics/math .08 .10*** .06
Biology (including premed) .30 .24 .36***
Engineering .28 .44*** .12
Social sciences .03 .02 .04***
Arts or humanities .04 .03 .06**
Business .05 .05 .05
Education .05 .02 .08***
Health professions .04 .01 .07***
Other .12 .08 .16***
Young-adult prospective earnings (log, from ACS) 10.53 (.27) 10.61*** (.21) 10.44 (.29)
Prime-aged prospective earnings (log, from ACS) 11.05 (.24) 11.13*** (.23) 10.98 (.21)
Own expected earnings at age 30 (log, from PtC 11.01 (.47) 11.06*** (.44) 10.94 (.50)
data)a
Student-level characteristics
Female .50 — —
Race
White .63 .64 .62
Asian .18 .19 .17
Other .19 .17 .21**
First-generation college student .44 .39 .50***
Highest math taken (1–5) 3.15 (1.12) 3.30*** (1.07) 3.00 (1.14)
N 2,639 1,327 1,312

Source: Pathways through College Study (PtC); 2014 American Community Survey (ACS).
Note: Significance levels indicated for group with larger mean or percentage (i.e., men compared to women). Math
categories are geometry or algebra II, precalculus, statistics, calculus, and multivariate calculus. ACS earnings are
assigned as the median earnings among workers who are employed, currently working, in a given age group, and with
a bachelor’s degree in the student’s major.
a
Values are bottom coded at the 5th percentile and top coded at the 95th percentile to minimize skew. Sample sizes are
2,486 (total), 1,280 (men), 1,206 (women).
**p \ .01. ***p \ .001 (two-tailed tests).

also possible that students choose a major and ret- items, respondents may not have given much
roactively list the motivating factors most consis- thought to differentiating between the six items.
tent with the major they have settled on (for sim- A review of the data shows that only 6 percent
ilar discussions, see Nau, Dwyer, and Hodson of respondents assigned the same score to each
2015; Quadlin 2017). Thus, causal ordering cannot item, so straight-lining was not widespread. But
be determined with these data from the first term because there were six items and only a 5-point
of college, nor is the survey worded in such scale, each respondent had at least one tie between
a way that causal ordering can be inferred. I con- items. Accordingly, I refer to this construct as
sider this issue further in the Conclusion. ‘‘preferences’’ for choosing a major rather than
Another challenge is that we cannot be sure ‘‘reasons’’ or another term that implies a definitive
how respondents ranked their preferences or rank. This also has potential implications in light
which preference may have ‘‘won out’’ in their of the gender component of this study: different
major choice. Because respondents were asked to preferences may win out for men and women
respond to each item on a 1-to-5 scale, and they even if respondents list the same criteria as
could assign the same score to some or all of the important.
Quadlin 7

Table 2. Preferences for Major Choice.

Mean (standard deviation)

Major preference Full sample Men Women

Money right after college 3.56 (.99) 3.64*** (.99) 3.49 (.97)
Money over course of career 3.89 (.96) 3.95*** (.97) 3.82 (.95)
Money—combined 7.45 (1.81) 7.59*** (1.82) 7.31 (1.79)
Keeping career options open 3.89 (.90) 3.86 (.90) 3.91 (.90)
Engaging entry-level classes 3.45 (1.02) 3.45 (1.02) 3.46 (1.03)
Engaging advanced classes 3.89 (.91) 3.89 (.91) 3.89 (.92)
Classes—combined 7.35 (1.68) 7.34 (1.66) 7.35 (1.71)
Helping others in your job 3.96 (1.00) 3.74 (1.03) 4.17*** (.91)
N 2,639 1,327 1,312

Source: Pathways through College Study.

Note: Significance levels indicated for group with larger mean or percentage (i.e., men compared to women).
Respondents were asked, ‘‘How important is each of the following to you when choosing a college major?’’ Responses
were entered on a scale of 1 (not at all important) to 5 (extremely important).
***p \ .001 (two-tailed tests).

College Major subject to sorting processes that intensify as stu-

dents progress through college, such as the ‘‘leaky
The initial analyses predict students’ majors dur- pipeline’’ out of STEM fields that disproportion-
ing the first term of enrollment. Students were ately affects women (Bradley 2000; Jacobs
asked to indicate their specific major if they had 1996). About 40 percent of students switched their
already declared one (e.g., biomedical engineer- major at least once during the fielding period,
ing). If they had not declared a major, students which is consistent with national estimates
were asked to indicate which broad category their (Denice 2018). Most major switches were within
eventual major would most likely fall under (e.g., the same or similar categories of majors (e.g.,
engineering).5 To keep responses consistent from psychology to sociology).
between declared and undeclared students, I sorted
the specific majors into the categories the PtC pro-
vided for students who had not declared a major.
This yielded nine categories (descriptions are as Prospective Earnings
listed in the survey): physical sciences (including In later analyses, I consider how students’ major
mathematics), biological/life sciences (including preferences are associated with their prospective
premed), engineering, social sciences (including earnings. Earnings are only one outcome students
psychology and history), arts or humanities consider when choosing a major, but scholars and
(including English, literature, dance, theater, and the public have focused recently on the economic
music), business, education, health professions, value of college in light of rising costs. I assess
and other.6 The specific majors included under prospective earnings to gauge whether gendered
each category are shown in Appendix A in the logics of major choice are potentially related to
online supplement. These categories are generally earnings inequality among college graduates.
consistent with past studies that sort majors for Earnings data come from two sources. I ini-
purposes of predicting earnings (Kim et al. 2015; tially use data on prospective earnings from the
Ma and Savas 2014), although the STEM catego- 2014 ACS (this is the base year of the PtC). The
ries are unusually detailed in the PtC data set due ACS is an ideal data set for this analysis because
to its STEM emphasis. it provides information on earnings and college
The analyses focus on students’ majors from majors for a large sample of college-educated
the first term of college because these reflect stu- workers. Because the earnings used in these anal-
dents’ initial inclinations given their major prefer- yses are derived from students’ major choices, the
ences. Majors from the first term are also less analyses essentially reveal whether, given the
8 Sociology of Education XX(X)

same preferences for their major, men and women Analytic Strategy
choose majors associated with disparate prospec-
tive earnings. Of course, there are other reasons The analyses are presented in two parts. I begin by
one might expect to see gender gaps in earnings, examining how students’ major preferences are
such as the reality that men tend to occupy more associated with their major choices. I used multi-
prestigious and higher-paying positions within nomial logistic regression to determine students’
industries and workplaces. I will discuss these pos- chances of choosing each of the nine major cate-
sibilities further, but here I focus on the earnings gories. All four preferences are included in this
tied to majors to determine whether students model, along with sociodemographic controls.8
choose majors associated with different earnings Then, I used Long and Freese’s (2014) -mtable-
potential at the median. postestimation command to determine students’
For each of the 173 majors in the 2014 ACS, I predicted probability of choosing each major,
identified the median income among individuals over the range of each major preference. For
who were employed and currently working. This example, for the ‘‘money earned’’ item, I calcu-
yielded a sample of about 500,000 college-educated lated students’ predicted probability of choosing
workers. Then, recognizing that majors may be each of the nine major categories at each level
associated with different earnings trajectories, and on the ‘‘money’’ scale, from 2 (not at all impor-
students may be differentially attuned to short- tant) to 10 (extremely important).9 I then tested
and long-term earnings, I calculated median earn- whether students’ predicted probability of choos-
ings for two age groups: young adults (i.e., ages ing a major changed significantly over the range
30 and below) and prime-aged workers (i.e., ages of major preferences. To use the same example,
25 to 54). If a student had declared a major, I this would tell us whether students who rated
assigned earnings equal to the median income money as extremely important were significantly
among workers with a degree in that major. If the more or less likely to choose a given major (e.g.,
student had not declared a major, I assigned earn- business) than those who rated money as not at
ings equal to the median salary among workers in all important. Separate models are shown for
the same category of majors the student expressed men and women to assess whether major preferen-
interest in. See Appendixes B (young adults) and ces are differentially related to major choice by
C (prime-aged workers) in the online supplement gender. Because I focus on results from postesti-
for median earnings by major. mation tests, the underlying regressions are not
In addition, I used data on students’ own shown in the main text; these models are shown
expected earnings at age 30, taken from the first in Table S1 in the online supplement.
PtC survey. I include this analysis to show how I next show how students’ major preferences
students’ major preferences map onto the salaries are associated with prospective earnings and, in
they expect to earn soon after labor market entry. turn, whether these relationships differ by gender.
To reiterate, because prospective earnings are
assigned from the ACS based on students’ majors,
Gender and Sociodemographic these analyses show whether men and women who
have similar priorities ultimately choose majors
Controls tied to disparate earnings. Prospective earnings
I account for several factors associated with major are estimated using quantile regressions, which
choice, major preferences, and/or earnings. Gen- do not assume homoscedasticity or normality,
der is the primary category of interest. Other con- and are often used for models with earnings as
trols include race (coded in the data set as white, an outcome. As before, all four major preferences
Asian, and ‘‘other’’—a broad category because are included in this model, although the preferen-
other groups were small, so the PtC combined sev- ces are presented separately for ease of interpreta-
eral groups to maintain student anonymity); an tion. These results are shown as figures in the main
indicator variable for first-generation college text, with the underlying regression shown in
attendance, which I use as a proxy for socioeco- Table S2 in the online supplement. All analyses
nomic status; and the highest math course taken are limited to respondents with complete data on
in high school, which I use as a proxy for aca- the outcomes and all covariates (60 cases missing
demic preparation.7 on the outcome [2 percent], 21 cases missing on
Quadlin 9

other covariates [less than 1 percent]), bringing the 10.13, p \ .001), and business (difference =
final sample size to 2,639. 10.10, p \ .001). Women’s chances of majoring
in these fields, particularly engineering and busi-
ness, increase substantially as they place more
RESULTS emphasis on earnings. Yet remember that women
generally deemphasize earnings when choosing
How Are Major Preferences a major (see Table 2), which may help explain
why women are underrepresented in engineering
Associated with Major Choice? despite associating this field with high earnings.
Table 3 shows how students’ major preferences Even when women place great emphasis on earn-
are associated with major choices. The table is ings, other preferences may ultimately win out for
separated into four sections, corresponding to the them. Women are also much less likely to major in
four preferences included in the data: money education as they view earnings as increasingly
earned, keeping career options open, engaging important (difference = 20.22, p \ .01).
classes, and helping others in your job. Earnings-motivated women, therefore, may sys-
The top section of Table 3 shows results for tematically opt out of education majors.
‘‘money earned,’’ separately for men and for The second section of Table 3 shows results for
women. The column labeled ‘‘Not at all ‘‘keeping career options open.’’ As men view
important’’ shows students’ predicted probability career options as increasingly important, they are
of choosing each major category when they indi- more likely to major in physics/math (difference =
cated earnings were not at all important. The col- 10.08, p \ .05). Perhaps surprisingly, men are
umn labeled ‘‘Extremely important’’ shows these also more likely to choose education under these
predicted probabilities when students indicated circumstances (difference = 10.02, p \ .05),
earnings were extremely important. Finally, the although this result is substantively small, and
column labeled ‘‘Diff.’’ shows the difference men education majors were uncommon in the
between the two, with tests for whether these pre- PtC sample overall. Turning to results for women,
dicted probabilities are significantly different for we see that when women place more emphasis on
a given major category. career options, they are less likely to major in
As an initial illustration, the first row shows health professions (difference = 20.12, p \ .05).
that men have a 0.17 probability of majoring in Although health majors are often billed as path-
physics/math when earnings are not at all impor- ways to stable and relatively high-paying jobs,
tant to them, and they have a 0.09 probability of this message apparently does not resonate with
majoring in physics/math when earnings are women, or at least the first-year women in the
extremely important. The difference between the PtC sample. Perhaps women believe health majors
two, –0.08, is not statistically significant. In other offer a reliable, but ultimately limiting, pathway to
words, men’s chances of majoring in physics/math employment or that the health field is less flexible
are statistically unchanged regardless of the than other industries.
importance they place on earnings. Two other The third section shows results for ‘‘engaging
major categories are significantly different over classes.’’ As men view engaging classes as
this range, however. Men’s probabilities of major- increasingly important, they are more likely to
ing in engineering (difference = 10.27, p \ .001) major in engineering (difference = 10.23, p \
and business (difference = 10.08, p \ .001) both .01) and less likely to major in business (differ-
increase significantly as they place more emphasis ence = 20.16, p \ .05). Women, however, are
on earnings. Put differently, as men view earnings not significantly more or less likely to choose
as increasingly important, they are considerably any major as they place greater emphasis on
more likely to major in engineering and, to a lesser engaging classes. This does not mean women do
extent, business. This pattern suggests men tend to not value engaging classes. In fact, men and
choose these majors when they are seeking a reli- women are nearly identical in terms of the pre-
able route to economic success. mium they place on intellectual engagement (see
Turning to women, we see they are more likely Table 2), but women’s desire for engaging classes
to choose three majors as they view earnings as does not translate to any particular major. This
increasingly important: physics/math (differ- may be because even when women find subjects
ence = 10.05, p \ .05), engineering (difference = interesting, other unmeasured beliefs about those
10 Sociology of Education XX(X)

Table 3. Effects of Major Preferences on Major Choices, by Gender.

Men Women
Not at all Extremely Not at all Extremely
important important Diff. important important Diff.

Money earned
Physics/math .17 .09 –.08 .02 .07 1.05*
Biology (including premed) .26 .17 –.09 .35 .35 .00
Engineering .27 .54 1.27*** .03 .16 1.13***
Social sciences .01 .01 .00 .08 .03 –.05
Arts or humanities .08 .01 –.07 .09 .04 –.05
Business .00 .08 1.08*** .00 .10 1.10***
Education .13 .00 –.13 .24 .02 –.22**
Health professions .02 .00 –.02 .04 .07 1.03
Other .06 .09 1.03 .15 .16 1.01
Keeping career options open
Physics/math .06 .14 1.08* .05 .06 1.01
Biology (including premed) .16 .23 1.07 .33 .41 1.08
Engineering .57 .45 –.12 .07 .12 1.05
Social sciences .03 .00 –.03 .07 .04 –.03
Arts or humanities .02 .02 .00 .03 .07 1.04
Business .04 .04 .00 .07 .04 –.03
Education .00 .02 1.02* .09 .04 –.05
Health professions .03 .00 –.03 .17 .05 –.12*
Other .09 .09 .00 .12 .19 1.07
Engaging classes
Physics/math .07 .13 1.06 .07 .05 –.02
Biology (including premed) .30 .17 –.13 .41 .38 –.03
Engineering .31 .54 1.23** .07 .13 1.06
Social sciences .01 .01 .00 .02 .06 1.04
Arts or humanities .01 .03 1.02 .06 .05 –.01
Business .18 .02 –.16* .04 .05 1.01
Education .05 .00 –.05 .11 .03 –.08
Health professions .01 .01 .00 .06 .07 1.01
Other .07 .09 1.02 .17 .17 .00
Helping others in your job
Physics/math .15 .08 –.07 .14 .04 –.10
Biology (including premed) .04 .38 1.34*** .17 .43 1.26***
Engineering .62 .36 –.26*** .28 .08 –.20**
Social sciences .00 .02 1.02* .01 .06 1.05**
Arts or humanities .04 .02 –.02 .11 .04 –.07
Business .04 .03 –.01 .08 .03 –.05
Education .00 .02 1.02** .00 .09 1.09***
Health professions .00 .02 1.02 .02 .09 1.07***
Other .11 .07 –.04 .21 .14 –.07

Source: Pathways through College Study.

Note: Multinomial logistic regressions; predicted probabilities reported. ‘‘Not at all important’’ is the predicted
probability of choosing a major when a factor is rated not at all important; ‘‘Extremely important’’ is the predicted
probability of choosing a major when a factor is rated extremely important. These columns may not add to 100 due to
rounding. ‘‘Diff.’’ is the difference in predicted probabilities between ‘‘Not at all important’’ and ‘‘Extremely important.’’
The models include all four factors simultaneously as well as controls for race, whether the respondent was a first-
generation college student, and highest math course taken in high school (values held at their means in calculating
predicted probabilities). Sample sizes are 1,327 (men) and 1,312 (women).
*p \ .05. **p \ .01. ***p \ .001 (two-tailed tests; predicted probabilities for ‘‘not at all important’’ and ‘‘extremely
important’’ are significantly different).
Quadlin 11

fields prevent them from majoring in them—such are oriented toward engaging classes or helping
as perceptions of engineering and other STEM others. Women often major in biology/premedi-
fields as fostering a ‘‘chilly climate’’ toward cine, social sciences, education, or health when
women (Herzig 2004). they place a premium on helping others. Notably,
The bottom section of Table 3 shows results for women are overrepresented in education and
‘‘helping others.’’ This item yields many signifi- health despite low enrollment among earnings-
cant associations for both men and women. As and career-motivated women, respectively. This
men view helping others as increasingly impor- pattern suggests women may ultimately choose
tant, they are more likely to major in social scien- majors they associate with altruistic rewards,
ces (difference = 10.02, p \ .05), education (dif- even if they know they are sacrificing extrinsic
ference = 10.02, p \ .01), and especially rewards. This type of trade-off is not nearly as
biology/premedicine, which elicits a 0.34 increase common for men as it is for women, thus revealing
in predicted probabilities between those who see distinctly gendered logics of major choice.10
helping others as not at all important versus These results raise the question, how might
extremely important (p \ .001). Furthermore, these gendered logics of major choice contribute
men are less likely to major in engineering as to inequality? In the next section, I consider earn-
they view helping others as increasingly important ings projections among men and women who
(difference = 20.26, p \ .001). Men often associ- share the same major preferences.
ate an engineering major with high earnings and
engaging classes, as noted earlier, so perhaps
men see engineering as offering material and intel- Major Preferences
lectual benefits but less in the way of benfits to
society. Supplementary analyses suggest this per-
and Prospective Earnings
ception varies across different engineering fields. Figure 1 shows how the four major preferences—
Men in civil, agricultural, and biomedical engi- money earned, career options, engaging classes,
neering place more emphasis on helping others, and helping others—are associated with men’s
whereas male electrical engineers have the lowest and women’s prospective earnings. The labor mar-
point estimate for this item. Yet, this overall trend ket has changed considerably over the past few
implies that the most altrusitically motivated men decades, and earnings for newly minted graduates
may not gravitate toward engineering. in some fields are not in line with what previous
Opportunities to help others are particularly generations had come to expect (see, e.g., Kalle-
salient for women’s major choices. As women per- berg 2011). To account for these structural
ceive helping others as increasingly important, changes, the two panels in Figure 1 represent pro-
they are more likely to major in biology/premedi- spective earnings in two age groups: young adults
cine (difference = 10.26, p \ .001), social scien- (ages 30 and below) and prime-aged workers (ages
ces (difference = 10.05, p \ .01), education 25 to 54). Prospective earnings are taken from the
(difference = 10.09, p \ .001), and health profes- 2014 ACS and are assigned equal to the median
sions (difference = 10.07, p \ .01). All four of salary among workers with a bachelor’s degree
these majors are female dominated in the PtC sam- in the student’s major (see Appendixes B and C
ple (see Table 1), and they tend to be female dom- in the online supplement).11 Underlying regres-
inated nationwide (England and Li 2006). This is sions are shown in Table S2 in the online
consistent with the notion that women are particu- supplement.
larly likely to be attracted to fields that offer Figure 1 shows a clear pattern: the majors men
altrustic rewards (see Table 2). In addition, women choose are associated with significantly higher
are less likely to major in engineering as they view earnings than the majors women choose, regardless
helping others as increasingly important (differ- of men’s and women’s major preferences. This is
ence = 20.20, p \ .01), which is in line with generally consistent whether the outcome is
the results for men. young-adult or prime-aged earnings. Results of ear-
Overall, these data provide insight into the lier analyses showed that men and women choose
majors men and women associate with several different majors even when they have the same
key criteria. Men may major in engineering or major preferences. Here, we see that when men
business if they wish to maximize their economic and women both prioritize economic returns—or
returns, and they may opt out of these fields if they most any of the criteria included in the data set
12 Sociology of Education XX(X)

Figure 1. Prospective earnings (from American Community Survey [ACS]) over the range of major pref-
erences, by gender, N = 2,639. (Panel A) Prospective earnings calculated using young-adult college grad-
uates (ages 30 and under). (Panel B) Prospective earnings calculated using prime-aged college graduates
(ages 25 to 54).
Sources: Pathways through College Study; 2014 ACS.
Note: Dashed lines = women; solid lines = men. Results are derived from quantile regressions shown in the
online supplement; logged prospective earnings shown. The model includes controls for race, whether the
respondent was a first-generation college student, and highest math course taken in high school. Earnings
are assigned as the median earnings among workers who are employed, currently working, in a given age
group, and with a bachelor’s degree in the student’s major from the 2014 ACS.

for that matter—men’s major choices are associ- of the distributions. In the most extreme case,
ated with significantly higher earnings than only 32 students (1 percent) indicated that keeping
women’s. their career options open was not at all important,
One exception emerges in the top-right panel: thus rendering any gender differences in prospec-
men and women who indicate helping others is tive earnings nonsignificant. But notwithstanding
extremely important have statistically indistin- the small cell sizes, many point estimates of pro-
guishable prospective young-adult earnings. This spective earnings are similar for men and women
appears to happen because men’s prospective who see these criteria as unimportant. In an excep-
earnings decline over the range of the ‘‘helping tion, men who view engaging classes as relatively
others’’ item. In other words, as men place increas- unimportant (i.e., a score of 3 on a 2-to-10 scale)
ing emphasis on helping others, they tend to select have significantly higher prospective earnings
majors associated with lower prospective young- than their women counterparts for both young-
adult earnings. Women’s prospective earnings, adult and prime-aged earnings. Supplementary
conversely, are relatively flat. Once this outcome analyses show this is mostly due to the men engi-
is expanded to prime-aged earnings in the bottom neers in this cell who placed little emphasis on
panel, men who emphasize helping others have engaging classes but had high prospective
higher prospective earnings than like-minded earnings.
women, presumably because men’s majors have In addition to these main results using ACS
better earnings trajectories. But despite this excep- data, I conducted two more analyses to examine
tion, the overall pattern in Figure 1 shows that the prospective earnings outcome further. First, I
men’s and women’s logics of major choice are replicated Figure 1 using sex-specific median
unequal from an earnings perspective. earnings—that is, men (women) are assigned the
Another notable pattern is that gender gaps in median earnings among men (women) in their
prospective earnings are generally not significant major (see Figure S1 in the online supplement).
at the lower ends of these items. This is partly This analysis might provide a better estimate of
a function of small cell sizes at the lower ends students’ eventual earnings because it accounts
Quadlin 13

Figure 2. Own expected earnings at age 30 (from Pathways through College Study) over the range of
major preferences, by gender, n = 2,486.
Source: Pathways through College Study.
Note: Dashed lines = women; solid lines = men. Results are derived from quantile regression shown in the
online supplement; logged expected earnings shown. The model includes controls for race, whether the
respondent was a first-generation college student, highest math course taken in high school, and major.

for the fact that men’s earnings tend to be higher data on the outcome (n = 2,486), but the shape
than women’s even when they attain bachelor’s and patterning of results is similar to the results
degrees in the same field. In general, results with generated with ACS earnings. It is immediately
sex-specific prospective earnings are consistent apparent that both men and women appear to over-
with those shown in Figure 1, although gender estimate their earnings. The y-axes in Figure 2 are
gaps in pay are much larger. Note that sex-specific nearly in line with the y-axes for prime-aged work-
median earnings generally align with non-sex- ers—not young-adult workers—in Figure 1. This
specific median earnings for young adults, but is a potentially important dynamic, especially con-
these two projections are much more divergent sidering the observed gender gaps in expected
for prime-aged workers. This pattern is consistent earnings. Women tend to expect lower earnings
with research showing the gender pay gap is than men regardless of their major preferences,
smallest at labor market entry (Marini and Fan but women still expect higher earnings than what
1997). Yet overall, results are similar regardless is realistic (especially if we take sex-specific earn-
of whether prospective earnings are sex specific. ings into account). This indicates women are cog-
Finally, I recreated these analyses using an nizant of their relative placement in the income
alternative outcome: students’ own expected earn- distribution but perhaps not their absolute place-
ings at age 30, measured in the first wave of the ment. Women recognize they may have to sacri-
PtC data set. Figure 2 shows these results.12 The fice pay to receive other benefits, such as greater
sample size is reduced slightly due to missing workplace flexibility, but their expected earnings
14 Sociology of Education XX(X)

are still greater than what is likely to materialize. patterns suggest men’s logics of major choice point
To be fair, men also appear to overestimate their them toward STEM fields more often than wom-
earnings, so their expectations are not accurate, en’s, which echoes prior research on gender and
either. But women’s mismatch may be more STEM. Men tend to perceive and experience the
impactful as they enter the workforce and must benefits associated with STEM fields, whereas
reconcile their expectations with the realities of women tend to perceive and experience the con-
the contemporary labor market. straints these fields can impose, both in higher edu-
cation and in the workplace.
In addition, perceptions of altruistic rewards
are distinctly related to major choice, particularly
DISCUSSION for women. Women who prioritize altruism are
How are students’ major preferences associated increasingly likely to major in biology/premedi-
with their major choices? How do these patterns cine, social sciences, education, and health. Nota-
differ for men and women? And to what extent bly, all of these majors are female dominated, both
might these gendered logics of major choice con- in the PtC data and according to national esti-
tribute to earnings inequality? Using data from the mates. This congruence between helping others
PtC, this study builds on prior research by showing and major choice suggests altruistic rewards are
how students’ broader aims are filtered through at the forefront of many women’s major choices.
gendered choice sets as they select their majors. This is not a new observation; many scholars
This is a unique perspective on major choice that have noted women’s tendency to choose majors
goes beyond preferences for extrinsic and intrinsic and careers that will give them opportunities to
rewards to understand how students actually put help other people as well as these fields’ greater
these preferences into action. receptiveness toward women. But this is a note-
I find that the major preferences students often worthy pattern because no other criterion, for
cite—potential amount of money earned, employ- men or women, aligns with major choice to the
ability, intellectual engagement, and opportunities same extent as women’s preference for altruisti-
to help others—are often associated with different cally oriented fields.
major choices by gender. One source of common Aside from showing how students’ major prefer-
ground for men and women is their perceptions ences are associated with their actual major choices,
of majors that will lead to strong economic these analyses are useful for understanding to what
returns. Among students who place a premium extent men and women see the various major pref-
on earnings, both men and women are more likely erences as in conflict—and which preferences ulti-
to choose majors such as business and engineer- mately ‘‘win out.’’ Table 4 provides a visual of these
ing. Women also show a modest effect in the phys- patterns. Among men, engineering is positively
ics/math category (as well as a negative associa- associated with money earned and engaging classes
tion between earnings motivation and education, but negatively associated with helping others. This
discussed further later). is a key pattern because men are strongly overrepre-
This may seem like a potentially important pat- sented in engineering, thus suggesting men may be
tern in terms of recruiting women to STEM fields, satisfied with majors that are mostly oriented
but there are two main counterpoints to keep in toward extrinsic rewards. Similarly, business for
mind. First, men place more emphasis on extrinsic men is positively associated with money earned
rewards than do women—so even if women associ- but negatively associated with engaging classes.
ate STEM with high earnings, this is a relatively These patterns imply that when men make
small component of many women’s major choices. a trade-off in their major choice, they are more
Prior research shows that women deemphasize likely to prioritize economic returns over other
earnings, at least in part because many high-paying rewards, such as engaging classes or altruism.
jobs entail long hours and poor flexibility. Second, For women, some very different trends emerge.
men frequently cite other reasons for choosing Education is positively associated with helping
STEM majors that do not necessarily emerge for others but negatively associated with money
women. Men are more likely to choose physics/ earned. Similarly, health is positively associated
math when they want to keep their career options with helping others but negatively associated
open, and men are more likely to choose engineer- with career options. The fact that women are over-
ing when they want to take engaging classes. These represented in education and health, despite these
Quadlin 15

Table 4. Summary of Associations between Major Preferences and Major Choices, by Gender.

Money Keeping career Engaging Helping others

Major earned options open classes in your job

Men
Physics/math 1
Biology (including premed) 1
Engineering 1 1 –
Social sciences 1
Arts or humanities
Business 1 –
Education 1 1
Health professions
Other
Women
Physics/math 1
Biology (including premed) 1
Engineering 1 –
Social sciences 1
Arts or humanities
Business 1
Education – 1
Health professions – 1
Other

Source: Pathways through College Study.

Note: This table summarizes results reported in Table 3.

combinations of perceived rewards, underscores the fields women choose when they want to have
the importance women place on altruism. Even the strongest economic returns, the widest oppor-
though many women perceive education and tunities, the most engaging classes, and the most
health as having little to offer in terms of earnings opportunities to help people are lower paying
or employability, women are still disproportion- than the fields men choose to meet these criteria.
ately likely to major in these fields—presumably This is a key factor in inequality that has been
because they value altruism and because these underexamined in prior research. Of course, earn-
fields are receptive to women and offer workplace ings are not the only thing that matters for major
flexibility. This also highlights a particular facet of choice, and women’s majors could very well sur-
the PtC data: because respondents did not rank pass men’s in terms of life satisfaction, flexibility,
their major preferences, but rated each preference or other criteria. It is also important to note that the
on a 1-to-5 scale, the data do not explicitly tell us earnings projections are based on more specific
which major preference ‘‘won out.’’ It appears that fields than the analyses with the PtC data, and
extrinsic rewards often win out for men, and altru- the selection of specific fields within broader
istic rewards often win out for women, but we can- fields is likely a source of stratification that is
not know for sure without data on students’ not captured elsewhere in the analyses.
explicit rankings. Future research can take this Overall, this study demonstrates the impor-
scale construction into account. tance of gendered choice sets as a mechanism
By incorporating data on prospective earnings, that constrains college majors. Even if men and
I show how these gendered logics of major choice women cite the same preferences for their major,
may be implicated in broader patterns of earnings the majors they choose are highly divergent,
inequality. I find that men’s major choices are tied with women’s majors being lower status and lower
to significantly higher prospective earnings than paying. Prior research on major choice suggests
women’s major choices, even when men and that shifting students’ preferences is key to equal-
women cite the same preferences. That is to say, izing gender differences in major choice. The
16 Sociology of Education XX(X)

patterns observed here, however, suggest that any College students weigh multiple factors when
interventions designed to change students’ prefer- choosing a field of study. In addition to the many
ences may not have the same outcomes for men perceptible costs and benefits students consider,
and women. For example, an intervention they are also constrained by gendered conceptions
designed to encourage high earnings might have of what a good or appropriate major would be.
different consequences by gender. Different fields This study demonstrates that gendered logics of
come to mind when men and women think of eco- major choice send students on different pathways
nomic returns, engaging classes, and opportunities while also contributing to inequality throughout
to help others. Students are likely to choose majors the life course. Scholars should continue to investi-
they believe meet these criteria but that are also gate these gendered logics and the role they play in
aligned with their gendered choice set—thus deep- college experiences and outcomes.
ening inequality among college graduates.
Throughout this article, I have posited that the
most likely causal pathway is from major preferen- RESEARCH ETHICS
ces to major choices. This is the mechanism The research reported in this article was reviewed and
assumed in most prior research. However, it is approved by the Institutional Review Board at the Uni-
also plausible that students work backward: stu- versity of Wisconsin-Madison. All human subjects
dents might choose a major and then list the moti- gave their informed consent to participate in the Path-
vating factors they see as most consistent with the ways through College Study, and adequate steps were
major they have already chosen. An alternative taken to protect participants’ confidentiality.
possibility is that students were not answering
for themselves when they answered the questions
ACKNOWLEDGMENTS
about major preferences—instead, they may have
indicated which criteria should be more and less I am grateful to Chandra Muller, Eric Grodsky, Cather-
important for college students broadly defined. ine Riegle-Crumb, Brian Powell, Emma Cohen, and
This is unlikely because it would have required members of the Pathways through College Research Net-
students to think abstractly, and the PtC survey work for their thoughtful comments and assistance.
capitalized you in this question (‘‘How important Research reported in this article was supported by the
Pathways through College Research Network funded
is each of the following to YOU when choosing
by the National Science Foundation under grant numbers
a college major?’’), but some students might DUE 1317196 and DUE 1317206. Any opinions, find-
have interpreted the question this way. ings, and conclusions or recommendations expressed in
A key strength of this study is its use of the PtC this material are those of the author and do not necessar-
data set, which is the only large-scale national sur- ily reflect the views of the National Science Foundation.
vey (to my knowledge) that examines the factors
students considered in their major choice. Yet,
because respondents were drawn from institutions ORCID iD
that emphasize STEM entry and persistence, they Natasha Quadlin https://orcid.org/0000-0002-5854-
differ from the general population of college stu- 8250
dents. The percentage of students majoring in phys-
ics, math, engineering, and several other fields is
relatively high, and the percentage majoring in SUPPLEMENTAL MATERIAL
non-STEM fields is relatively low. In a more repre-
Supplemental material is available in the online version
sentative sample, perhaps students would be less of this journal.
likely to choose STEM fields given their various
major preferences. Economically motivated stu-
dents might be especially likely to choose majors NOTES
other than STEM, as STEM fields are typically per-
1. Other work uses the term choice sets differently.
ceived as challenging, and most students would
Alon and DiPrete (2015) conceptualize a choice
choose an easier major if their primary goal was set as the broad collection of fields students consider
to make money and they were not already inclined before they make their initial major choice. These
toward STEM. Future research can examine these choice sets are less sex segregated than students’
dynamics in a broader sample to understand major declared majors because they are relatively uncon-
choice in the general population. strained. Here I use the term more narrowly, to refer
Quadlin 17

to the fields that students realistically consider given biology/premedicine in the final wave but not in the
gendered constraints on major choice. first wave). I do not overemphasize these findings,
2. Institutions are redacted to maintain student however, because the sample size is considerably
anonymity. smaller by the final wave (n = 1,592). Results are
3. For comparison, about 20 percent of entering first- shown in Table S4 in the online supplement.
years at four-year colleges in the National Longitu- 11. These earnings projections are made for students’
dinal Survey of Youth 1997 (NLSY-97) had specific majors, whereas the analyses in Table 3
declared a major in any science, technology, engi- are based on larger categories of majors. Supple-
neering, and mathematics (STEM) discipline, which mental analyses using broad major categories to
includes physics/math, biology, and engineering produce Figure 1 are consistent with those shown.
(Quadlin 2017). Note that the analyses in Table 3 cannot speak to
4. A factor analysis of these items returns two factors: how students choose specific majors within these
one emphasizing earnings and the other emphasiz- larger categories, and this process could account
ing engaging classes. I chose not to use the factor for further inequality. This is discussed further in
analysis in order to retain data on the items that the Discussion.
were poorly represented in the two factors. 12. The underlying regression is shown in Table S3 in
5. Approximately 8 percent of students in the analytic the online supplement.
sample were undeclared. This figure is lower than
the 25 percent of entering first-years who were unde-
clared in the NLSY-97, although this pattern is con- REFERENCES
sistent with research showing that less privileged stu-
dents (such as the often-first-generation students in Acker, Joan. 1990. ‘‘Hierarchies, Jobs, Bodies: A Theory
the Pathways through College Study data set) are of Gendered Organizations.’’ Gender & Society 4:
unlikely to enroll in college without a major (Quadlin 139–58.
2017). The oversampling of STEM students also Alon, Sigal, and Thomas A. DiPrete. 2015. ‘‘Gender Dif-
likely yielded a higher percentage of declared stu- ferences in the Formation of a Field of Study Choice
dents than would be found in the general population Set.’’ Sociological Science 2:50–81.
of college attendees. Supplemental analyses with Armstrong, Elizabeth A., and Laura T. Hamilton. 2013.
a control for students who had not yet declared Paying for the Party: How College Maintains
a major are consistent with those shown. Inequality. Cambridge, MA: Harvard University
6. Because premed was explicitly listed with biology, Press.
one potential source of error is that undeclared stu- Arum, Richard, and Josipa Roksa. 2011. Academically
dents chose ‘‘biology’’ when their prospective major Adrift: Limited Learning on College Campuses. Chi-
should actually be classified with ‘‘health’’ (this might cago: University of Chicago Press.
be the case with pre-pharmacy or pre-veterinary med- Blair-Loy, Mary. 2003. Competing Devotions: Career
icine, for example). This is ultimately a small number and Family among Women Executives. Cambridge,
of students (39 in undeclared biology, 19 in unde- MA: Harvard University Press.
clared health), and results are virtually identical Bradley, Karen. 2000. ‘‘The Incorporation of Women
regardless of how they are sorted. into Higher Education: Paradoxical Outcomes?’’
7. I also explored using a control for expected college Sociology of Education 73:1–18.
grade point average, but this item was missing for Cebula, Richard J., and Jerry Lopes. 1982. ‘‘Determi-
about 40 percent of the sample. Analyses with this nants of Student Choice of Undergraduate Major
control are generally consistent with those shown Field.’’ American Educational Research Journal
but have smaller sample sizes. 19:303–12.
8. I conducted supplementary analyses using the four Cech, Erin A. 2013. ‘‘The Self-Expressive Edge of Occu-
major preferences in separate models. Substantive pational Sex Segregation.’’ American Journal of
conclusions from these models are equivalent to Sociology 119:747–89.
those discussed in the main text. Charles, Maria, and Karen Bradley. 2002. ‘‘Equal but
9. Controls are included in these calculations, and the Separate? A Cross-National Study of Sex Segrega-
controls are held at their means. Additional analyses tion in Higher Education.’’ American Sociological
with marginal effects averaged across observations Review 67:573–99.
are similar to those shown. Charles, Maria, and Karen Bradley. 2009. ‘‘Indulging
10. In supplementary analyses, I used students’ major Our Gendered Selves? Sex Segregation by Field of
preferences from the first wave to predict majors Study in 44 Countries.’’ American Journal of Sociol-
in the final wave (third-year spring). Patterns are ogy 114:924–76.
generally consistent regardless of which majors Correll, Shelley J. 2001. ‘‘Gender and the Career Choice
are used, with a few exceptions (e.g., men’s prefer- Process: The Role of Biased Self-Assessments.’’
ence for money earned is negatively associated with American Journal of Sociology 106:1691–1730.
18 Sociology of Education XX(X)

Davies, Scott, and Neil Guppy. 1997. ‘‘Fields of Study, Kim, ChangHwan, Christopher R. Tamborini, and
College Selectivity, and Student Inequalities in Arthur Sakamoto. 2015. ‘‘Field of Study in College
Higher Education.’’ Social Forces 75:1417–38. and Lifetime Earnings in the United States.’’ Sociol-
Denice, Patrick. 2018. ‘‘Choosing and Changing Course: ogy of Education 88:320–39.
Postsecondary Students and the Process of Selecting Konrad, Alison M., J. Edgar Ritchie, Pamela Lieb, and
a Major.’’ Paper presented at the Pathways through Elizabeth Corrigall. 2000. ‘‘Sex Differences and
College Research Network, Madison, WI. Similarities in Job Attribute Preferences: A Meta-
Diekman, Amanda B., Elizabeth R. Brown, Amanda M. analysis.’’ Psychological Bulletin 126:593–641.
Johnston, and Emily K. Clark. 2010. ‘‘Seeking Con- Kyte, Sarah Blanchard, and Catherine Riegle-Crumb.
gruity between Goals and Roles: A New Look at 2017. ‘‘Perceptions of the Social Relevance of Sci-
Why Women Opt Out of Science, Technology, Engi- ence: Exploring the Implications for Gendered Pat-
neering, and Mathematics Careers.’’ Psychological terns in Expectations of Majoring in STEM Fields.’’
Science 21:1051–57. Social Sciences 6:19.
DiPrete, Thomas A., and Claudia Buchmann. 2013. The Lueptow, Lloyd B. 1980. ‘‘Social Change and Sex-Role
Rise of Women: The Growing Gender Gap in Educa- Change in Adolescent Orientations toward Life,
tion and What It Means for American Schools. New Work, and Achievement: 1964–1975.’’ Social Psy-
York: Russell Sage. chology Quarterly 43:48–59.
Dwyer, Rachel E. 2018. ‘‘Credit, Debt, and Inequality.’’ Long, J. Scott, and Jeremy Freese. 2014. Regression
Annual Review of Sociology 44:237–61. Models for Categorical Dependent Variables Using
Dwyer, Rachel E., Laura McCloud, and Randy Hodson. Stata, 3rd ed. College Station, TX: Stata Press.
2012. ‘‘Debt and Graduation from American Univer- Ma, Yingyi. 2009. ‘‘Family Socioeconomic Status,
sities.’’ Social Forces 90:1133–55. Parental Involvement, and College Major Choices:
Eagly, Alice H., and Steven J. Karau. 2002. ‘‘Role Con- Gender, Race/Ethnic, and Nativity Patterns.’’ Socio-
gruity Theory of Prejudice toward Female Leaders.’’ logical Perspectives 52:211–34.
Psychological Review 109:573–98. Ma, Yingyi, and Gokhan Savas. 2014. ‘‘Which Is More
England, Paula, and Su Li. 2006. ‘‘Desegregation Consequential: Fields of Study or Institutional
Stalled: The Changing Gender Composition of Col- Selectivity?’’ Review of Higher Education 37:
lege Majors, 1971–2002.’’ Gender & Society 20: 221–47.
657–77. Marini, Margaret Mooney, and Pi-Ling Fan. 1997. ‘‘The
Goyette, Kimberly A., and Ann L. Mullen. 2006. ‘‘Who Gender Gap in Earnings at Career Entry.’’ American
Studies the Arts and Sciences? Social Background Sociological Review 62:588–604.
and the Choice and Consequences of Undergraduate Marini, Margaret Mooney, Pi-Ling Fan, Erica Finley, and
Field of Study.’’ Journal of Higher Education 77: Ann M. Beutel. 1996. ‘‘Gender and Job Values.’’ Soci-
497–538. ology of Education 69:49–65.
Grodsky, Eric, and Chandra Muller. 2018. ‘‘Pathways McCabe, Janice M. 2016. Connecting in College: How
through College Study’’ [Data set]. Madison: Univer- Friendship Networks Matter for Academic and Social
sity of Wisconsin. Success. Chicago: University of Chicago Press.
Herzig, Abbe H. 2004. ‘‘Becoming Mathematicians: Morgan, Stephen L., Dafna Gelbgiser, and Kim A. Wee-
Women and Students of Color Choosing and Leaving den. 2013. ‘‘Feeding the Pipeline: Gender, Occupa-
Doctoral Mathematics.’’ Review of Educational tional Plans, and College Major Selection.’’ Social
Research 74:171–214. Science Research 42:989–1005.
Herzog, A. Regula. 1982. ‘‘High School Seniors’ Occu- Mullen, Ann L. 2014. ‘‘Gender, Social Background, and
pational Plans and Values: Trends in Sex Differences the Choice of College Major in a Liberal Arts Con-
1976 through 1980.’’ Sociology of Education 55: text.’’ Gender & Society 28:289–312.
1–13. Nau, Michael, Rachel E. Dwyer, and Randy Hodson.
Houle, Jason. 2014. ‘‘A Generation Indebted: Young 2015. ‘‘Can’t Afford a Baby? Debt and Young Amer-
Adult Debt across Three Cohorts.’’ Social Problems icans.’’ Research in Social Stratification and Mobil-
61:448–65. ity 42:114–22.
Jacobs, Jerry A. 1996. ‘‘Gender Inequality and Higher Quadlin, Natasha. 2017. ‘‘Funding Sources, Family
Education.’’ Annual Review of Sociology 22:153–85. Income, and Fields of Study in College.’’ Social
Kalleberg, Arne L. 2011. Good Jobs, Bad Jobs: The Rise Forces 96:91–120.
of Polarized and Precarious Employment Systems in Quadlin, Natasha Yurk, and Daniel Rudel. 2015.
the United States, 1970s to 2000s. New York: Rus- ‘‘Responsibility or Liability? Student Loan Debt and
sell Sage. Time Use in College.’’ Social Forces 94:589–614.
Kanter, Rosabeth Moss. 1977. ‘‘Some Effects of Propor- Riegle-Crumb, Catherine, Barbara King, Eric Grodsky,
tion on Group Life: Skewed Sex Ratios and and Chandra Muller. 2012. ‘‘The More Things
Response to Token Women.’’ American Journal of Change, the More They Stay the Same? Prior
Sociology 82:965–90. Achievement Fails to Explain Gender Inequality in
Quadlin 19

Entry into STEM College Majors over Time.’’ Amer- Xie, Yu, and Kimberlee A. Shauman. 2005. Women in
ican Educational Research Journal 49:1048–73. Science: Career Processes and Outcomes. Cam-
Roksa, Josipa, and Tania Levey. 2010. ‘‘What Can You bridge, MA: Harvard University Press.
Do with That Degree? College Major and Occupa- Zafar, Basit. 2013. ‘‘College Major Choice and the
tional Status of College Graduates over Time.’’ Gender Gap.’’ Journal of Human Resources 48:
Social Forces 89:389–416. 545–95.
Sax, Linda J. 1996. ‘‘The Dynamics of ‘Tokenism’: How
College Students Are Affected by the Proportion of
Women in Their Major.’’ Research in Higher Educa- Author Biography
tion 37:389–425.
Stevens, Mitchell L., Elizabeth A. Armstrong, and
Richard Arum. 2008. ‘‘Sieve, Incubator, Temple, Natasha Quadlin is an assistant professor of sociology
Hub: Empirical and Theoretical Advances in the at Ohio State University. Her research focuses on social
Sociology of Higher Education.’’ Annual Review of inequality in the contemporary United States, with an
Sociology 34:127–51. emphasis on access and returns to education. Current
Williams, Christine L., Chandra Muller, and Kristine projects examine perceptions of responsibility for col-
Kilanski. 2012. ‘‘Gendered Organizations in the lege costs, income inequality among college graduates,
New Economy.’’ Gender & Society 26:549–73. and public attitudes toward gender and sexuality.