Professional Documents
Culture Documents
46 Demographic Characteristics of High School Math and Science Teachers and Girls (Stearns, 2016)
46 Demographic Characteristics of High School Math and Science Teachers and Girls (Stearns, 2016)
doi: 10.1093/socpro/spv027
Advance Access Publication Date: 8 January 2016
Article
ABSTRACT
Given the prestige and compensation of science and math-related occupations, the un-
derrepresentation of women and people of color in science, technology, engineering, and
mathematics majors (STEM) perpetuates entrenched economic and social inequities.
Explanations for this underrepresentation have largely focused on individual characteristics,
including uneven academic preparation, as well as institutional factors at the college level.
In this article, we focus instead on high schools. We highlight the influence of the intersec-
tion between race and gender of female math and science teachers on students’ decisions to
major in STEM fields. Theoretically, this article extends the political science concept of rep-
resentative bureaucracy to the issue of women’s and disadvantaged minorities’ underrepre-
sentation in STEM majors. We analyze longitudinal data from public school students in
North Carolina to test whether organizational demography of high school math and science
faculty has an association with college major choice and graduation. Using hierarchical pro-
bit models with an instrumental-variable approach, we find that young white women are
more likely to major in STEM fields and to graduate with STEM degrees when they come
from high schools with higher proportions of female math and science teachers, irrespective
of the race of the teacher. At the same time, these teachers do not depress young white or
African American men’s chances of majoring in STEM. Results for African American
women are less conclusive, highlighting the limitations of their small sample size.
Although the relative social position of women has advanced in many ways, gender inequalities in ed-
ucation and the workforce remain prevalent features of American society (Poirier et al. 2009).
Likewise, African Americans are persistently less likely to major in science, technology, engineering,
and mathematics (STEM) fields than whites are (National Center for Science and Engineering
Statistic 2013). While women make up almost half of workers in the U.S. economy, they hold one-
This work was supported by the NSF Science, Technology, Engineering, and Mathematics Talent Expansion Program (STEP)
[Grant Number DUE-0969286]. All errors and interpretations are those of the authors. Direct correspondence to: Elizabeth Stearns,
University of North Carolina at Charlotte, Department of Sociology, 9201 University City Blvd, Charlotte, NC 28223. E-mail:
Elizabeth.stearns@uncc.edu.
C The Author 2016. Published by Oxford University Press on behalf of the Society for the Study of Social Problems. All rights reserved.
V
For permissions, please e-mail: journals.permissions@oup.com
87
88 Stearns, Bottı́a, Davalos, Mickelson, Moller, and Valentino
quarter of science and engineering jobs and African Americans are underrepresented in STEM jobs
as well (U.S. Department of Commerce 2011). Perplexingly, gender inequality in STEM occupations
persists despite the fact that women have surpassed men in the amount of education earned (DiPrete
and Buchmann 2013). Instead, this racial and gender stratification in STEM occupations seems to
have its roots in horizontal stratification in the post-secondary educational system—differences in the
types of majors that students pursue in college.
Research investigating the causes of gender and racial disparities in STEM enrollment and gradua-
tion at the postsecondary level has identified various explanations for these inequalities, emphasizing
both individual-level factors such as prior academic achievement and attitudes (Cronin and Roger
1999; Fox 1998; Ong 2005; Riegle-Crumb et al. 2012), as well as institutional-level factors. Most
studies on STEM inequities that have used an institutional perspective have concentrated on college
characteristics. For example, a noteworthy body of research has been established on the gender com-
position of college faculty in STEM departments, although no consensus has been reached as to
whether any association exists between departmental demography and students’ choice of major
fields of study (e.g., Bettinger and Long 2005; Griffith 2010; Price 2010; Robst, Keil, and Russo
Yet an increasing body of research on outcomes indicates that passive representation is also impor-
tant. This symbolic representation of people whom bureaucracies’ clients perceive to be similar to
themselves may make the services that agencies are providing more attractive to clients, thereby alter-
ing the behavior of the clients whom the bureaucrats are serving, in part because the bureaucrats are
acting as role models. This behavior alteration is seen in a variety of different outcomes both inside
and outside educational settings (Lim 2006; Meier and Nicholson-Crotty 2006; Theobald and
Haider-Markel 2009). In schools, the presence of African American female teachers is associated with
lower teen pregnancy rates among African American girls, due to a combination of role modeling and
of active guidance that teachers give girls regarding romantic choices and the importance of future ed-
ucation (Atkins and Wilkins 2013). Moreover, there are higher test scores for minority students in
schools with more African American teachers (Meier, Wrinkle, and Polinard 1999).
In contrast to the robust literature regarding race, evidence is spottier as to whether bureaucra-
cies representative in terms of gender behave any differently from less representative bureaucra-
cies, with less consensus as to whether gender is a salient identity on which to base
representation (e.g., Meier and Stewart 1992; Meier, Stewart, and England 1989). More recent re-
gender schema—between STEM success and masculinity, thereby increasing the odds that girls and
young women will enroll in and go on to succeed in STEM fields (Ma 2011).
In general, a higher proportion of female math and science teachers may improve girls’ percep-
tions of the cultural setting of STEM classrooms, help them to view STEM as more being welcoming
to women (Fox, Sonnert, and Nikiforova 2009; Statham, Richardson, and Cook 1991), and less
male-biased (National Alliance for Partnerships in Equity 2013). In addition, for girls questioning
whether STEM fields are the right fields for them, the presence of successful women teachers may
show them that barriers to success for them are not insurmountable (Steele 1997). This is the case
regardless of the gender of the instructors the students have in their own classrooms because we ex-
pect the dynamic also encompasses the possibility that girls are aware of the gender composition of
the overall faculty in certain academic areas. It may also be that it is necessary for there to be substan-
tially more women than men teachers in science and math to counter prevalent associations between
math and science and masculinity, especially given that these associations are established by elemen-
tary school (Andre et al. 1999; Meece, Glienke, and Burg 2006). These social aspects of classroom
climates may lead girls and young women to question whether they “belong” in STEM classes, ques-
This encouragement is particularly important, given that it may be needed to offset some of the per-
sistent bias against girls’ math abilities that teachers show (Riegle-Crumb and Humphries 2012).
Teachers may also actively communicate positive feedback to girls, giving them overt and explicit
signs of reassurance that math and science are indeed appropriate fields for them, thus encouraging
higher achievement. In fact, young women’s math scores are significantly higher in high schools
where women make up a greater proportion of the math and science faculty (Keiser et al. 2002) and
they are more likely to enroll in and graduate from college when they attend high schools with more
female high school faculty overall (Nixon and Robinson 1999). It is also possible that female teachers
may sponsor STEM-related clubs or activities. Again, these works do not give insight into whether ef-
fects of faculty gender composition are equally experienced across racial groups.
Finally, although it does not fall neatly on the passive-active representation continuum, if gendered
patterns of pedagogy that are seen among college instructors are similar among high school teachers,
female teachers may be more likely to use student-centered pedagogies such as collaborative learning,
which have consistently been associated with higher achievement levels and more thorough learning
of concepts (Gibbs 1992; Singer 1996; Trigwell et al. 1998; Webber 2012). In addition, they are
DATA
We use a unique data set to address our hypotheses, the North Carolina Roots data set (Stearns et al.
2013). This data set contains longitudinal information on the academic performance of all North
Carolina public school students from seventh grade through college graduation. Additionally, the
data set contains information about the characteristics of the schools that students attended through-
out their educational careers. All of these data originate in administrative records. The North
Carolina Education Research Data Center at Duke University provided data following students from
grades 7 through 12, and the University of North Carolina General Administration provided data on
college experiences.
92 Stearns, Bottı́a, Davalos, Mickelson, Moller, and Valentino
OUTCOME VARIABLES
We use two dependent variables as indicators of success in STEM: (1) a student declared a STEM
major (versus declaring any other non-STEM major) and (2) a student graduated with a STEM ma-
jor in the six years following his or her entrance into the UNC university system (versus the student
graduating from any other major).3 We ran separate analyses for each outcome (those who declare
and those who graduate from a STEM major) given that, in the cohort of NC university system stu-
dents that we analyzed, 24 percent of the students that declared a STEM major never graduated with
such a degree. To define a STEM major, we begin with the categorization utilized by the National
Science Foundation ADVANCE Program (NSF n.d.) where majors such as engineering, physical sci-
ences, earth, atmospheric or ocean sciences, mathematical and computer sciences, and biological and
agricultural sciences are considered to fall within the umbrella of STEM. We modify the ADVANCE
1. Ninety-two percent of the schools included in our sample had ten or more students participating in our analyses. None of the
public schools included in our sample is an all-girls or all-boys school. At the time this cohort of students was attending schools,
there was no single-sex public high school in the state of North Carolina. Wake County opened the state’s first single-sex public
schools in 2012.
2. These campuses are: Appalachian State University, East Carolina University, Elizabeth City State University, Fayetteville State
University, North Carolina Agricultural and Technical State University, North Carolina Central University, North Carolina State
University, UNC-Asheville, UNC -Charlotte, UNC -Chapel Hill, UNC -Greensboro, UNC -Pembroke, UNC -Wilmington,
University of North Carolina School of the Arts, Western Carolina University, and Winston-Salem State University.
3. The sample of students utilized in the analysis for “Graduate with a STEM major” is substantially smaller because we are compar-
ing them to students that actually graduate from any other major. Many students that had missing values in declared majors
fields were also excluded in the analysis for “Declare a STEM major.” We utilized a sample of 12,300 students in 200 high
schools to run models where the dependent variable is “Declare a STEM major” and we used a sample of 10,500 students, com-
ing from 160 high schools to run models in which the dependent variable is “Graduate with a STEM major.”
Teacher Demographics and Girls’ Success in STEM 93
program’s definition to exclude social and behavioral sciences from STEM for this analysis, given the
gender balance that has historically categorized most of the majors in these fields.
INDEPENDENT VARIABLES
Our key independent variables measure the organizational demography of high school math and sci-
ence faculty in various ways. First, we use the proportion of science and math teachers in a student’s
high school who are women. To create this measure, we took information from high school class-
room personnel files and calculated the gender composition of math and science teachers at each
high school between the years 2000 through 2004, years that correspond to the high school careers
of the students in our sample. Based on the State Course Code Subject Area, which indicates the
state-defined subject area to which teachers’ activity can most closely be associated, we determined
each teacher’s subject area. We also used the Outline of the Course Coding Structure for NC Public
Schools 2005–2006, which defines the discipline of the subject; we categorized all teachers in the dis-
ciplines of mathematics and science as math and science teachers. We then determined the propor-
Control Variables
We control for many secondary school characteristics that are frequently associated with student suc-
cess. All of these factors help capture important earlier contextual characteristics of the schools that
could contribute to increased or decreased chances of declaring and/or graduating from a STEM major.
These characteristics include: racial and socioeconomic composition of the school (proportion of white
students at the school and proportion of students on free-reduced lunch at the school); proportion of
female students in the school; proportion of teachers with licenses; proportion of teachers with ad-
vanced degrees; proportion of inexperienced teachers (those with less than three years of experience);
proportion of students in advanced college preparatory courses; and school locale (urban, suburban, or
rural).4 Correlations of school-level variables used in the estimations may be found in Table 2.
At the individual level, we control for demographic and family characteristics that are also fre-
quently associated with student outcomes, including race/ethnicity and SES (defined as whether stu-
dent received free-reduced lunch in seventh grade, received need-based financial aid in college, and is
a first generation college student in his or her family). We also control for whether the student trans-
ferred schools between ninth and eleventh grades. In addition, we include whether the student plans
to attend college, a question that students answered in the ninth grade.5
4. Log of school size, student/teacher ratio, and whether the school was a magnet or charter school were school-level variables in-
cluded in the preliminary analysis, but given their non-significant results, these were not included in the final models presented
in this article.
5. Previous versions of the article also included several controls that an anonymous reviewer pointed out might be collider variables that
would be outcomes of our key independent variable and thereby bias the estimation. These variables included teacher turnover, stan-
dardized test scores, and measures of course-taking patterns (measured by North Carolina’s standardized End-of-Course [EOC] test
scores in algebra 2 and biology; whether the student took the chemistry EOC exam). In light of this comment, however, we have omit-
ted those variables from the models presented in this article. Results are substantively identical when those variables are included.
94 Stearns, Bottı́a, Davalos, Mickelson, Moller, and Valentino
Dependent variables
Declared a STEM major .164 .371 .298 .457
Graduate from a STEM major .143 .350 .274 .446
Demographic measures
African American student .247 .431 .185 .389 .219 .414 .152 .359
Latino student .018 .131 .016 .123 .017 .130 .015 .121
Asian American student .032 .178 .040 .197 .033 .178 .040 .196
American Indian student .012 .107 .010 .101 .011 .106 .010 .100
Receives need-based financial aid in college .400 .490 .361 .480 .375 .484 .337 .473
ANALYTIC TECHNIQUE
To address our research questions, we use probit6 regression models with random high school effects
and college fixed effects. The dependent variable has a binary distribution and normally distributed
residuals. We control for arbitrary correlation within a school by using Huber-White standard errors
6. We ran logit and probit models for women and men in each specification. The estimates from the two models are consistent.
The signs of the coefficients are the same across models, and the same variables are statistically significant in each model. We
also estimated the Akaike information criterion (AIC) and the Bayesian information criterion (BIC) to compare the goodness of
fit of the probit and logit models. The AIC and BIC favor the probit model over the logit model (see Appendix B). As a result,
we prefer to use a probit model instead of a logit because it facilitates the implementation of instrumental variables methods
with binary dependent variables.
Teacher Demographics and Girls’ Success in STEM 95
Dependent variables
Declared a STEM major .144 .351 .163 .370
Graduate from a STEM major .117 .321 .144 .351
Demographic measures
Receives need-based financial aid in college .633 .482 .303 .460 .608 .488 .289 .454
Receives free-reduced lunch in 7th grade .398 .490 .038 .191 .379 .485 .036 .185
First generation college student .195 .397 .083 .276 .193 .395 .080 .271
Experiences during high school
Transferred school between 9th and 11th .090 .286 .055 .229 .092 .289 .053 .223
clustered by high school, and we account for factors that may vary across colleges by including sepa-
rate dummy variables for each college. North Carolina State University serves as reference category
since it is a flagship STEM college in the UNC system.7
We utilize probit models to examine two outcomes: students’ chances of declaring a STEM major
during the years 2005-2011 and their probability of graduating with a STEM major in those same
years. Probit models with random high school effects and college fixed effects allow us to examine
the influence of high school characteristics that impact college students’ likelihood of declaring and
graduating with a STEM major, taking into consideration the fact that certain groups of students at-
tended the same high schools, and some students attended the same college. The basic econometric
7. We ran separate models using each of the UNC campuses as a reference category; there is no significant difference in our
results.
96 Stearns, Bottı́a, Davalos, Mickelson, Moller, and Valentino
Table 3. Correlations of School Level Variables Used in Probit and IVProbit Estimations
Variable 1 2 3 4 5 6 7 8 9 10 11
1 Proportion students on
free/reduced lunch
2 High school located .12
in a rural area
3 High school located .06 .64
in suburban area
4 High school located .08 .52 .33
in a urban area
5 Proportion white .59 .24 .03 .32
students
6 Proportion female .00 .07 .06 .01 .03
students
model is presented in Equation 1: Probit model assessing probability of enrollment and graduation
with STEM majors by student and secondary school characteristics
Pr yiðjkÞ ¼ 1 j xi ; wj ; zk ¼ U ðc0 þ c1 xiðjkÞ þ c2 wj þ c3 zk Þ
or graduating with a STEM major. However, students may not be sorted randomly into schools
because their parents may have selected their home location based on their school preferences.
To control for the possibility that students are not randomly distributed across high schools
(Conger and Long 2013), we use an instrumental variable approach following Eric Bettinger and
Bridget Long (2005) and Joshua Price (2010). Those studies attempted to determine the impact
of instructor gender and race on student outcomes using an instrumental variable at the depart-
ment or institution level, one unit of analysis above the unit of analysis of their key independent
variable in a typical organization. Their instrumental variable was the percentage of courses
taught by women or a black instructor. In this study, we are attempting to tease out the influence
of a school-level measure (organizational demography) on student outcomes; hence, we use an
instrumental variable at the district level, which is one level above the unit of analysis of our key
independent variable. Using the instrumental variable controls for the possibility that students
are not randomly sorted among high schools. The basic econometric model for the instrumental
variable approach is presented in Equation 2: Probit model with a continuous endogenous ex-
planatory variable:
RESULTS
In the subsequent tables, we first present results from the probit model without an instrumental vari-
able followed by results from models with the instrumental variable. Tables 4 and 5 address the
8. We argue that our instrument, f1 , satisfies the two conditions for a valid instrument. First, the instrument is relevant: Cov fl ; wj
6¼ 0 and h1 6¼ 0. The condition h1 6¼ 0 means that f1 is partially correlated with wj once the other exogenous variables ðxi and
zk Þ have been netted out. We tested this condition by computing the F-statistic testing the hypothesis that the coefficient on the
instrument is zero in the first-stage regression. In all our models the first-stage F-statistic exceeds 10. Therefore, our instrument
is relevant (Stock, Wright, and Yogo 2002). These results are reported at the bottom of each table under first-stage estimation.
Second, the instrument is exogenous: Covðfl ; u1 Þ ¼ 0 and f1 is uncorrelated with the dependent variable, y1 , other than
through its correlation with wj . This condition, sometimes called the exclusion restriction, has two parts. The instrument is as
good as randomly assigned and has an effect on the outcome variable only through the first-stage channel (Angrist and Pischke
2009).
The exclusion restriction, however, cannot be tested directly (Imai, et al. 2011; Kraay 2012; Sobel 2008). As a result, we try to
rule out some of the potential violations theoretically. North Carolina’s school districts tend to be quite large: it is likely that par-
ents could choose their home location based on particular school characteristics such as convenience of location or student de-
mographic information (Dougherty et al. 2009; Goyette, Farrie, and Freely 2012; Hamilton and Guin 2005; Schneider and
Buckley 2002; Theobald 2005), but less likely that they will choose based on district characteristics. When and if families do
move for different educational opportunities, they are more likely to be looking at student demographics including racial and so-
cioeconomic composition than for teacher characteristics (Schneider and Buckley 2002) or relying on sometimes faulty informa-
tion and erroneous perceptions about school quality (Holme 2002). Nevertheless, we include district fixed effects to control for
factors that may vary across districts but do not vary within districts, for instance labor market conditions. We grouped districts
into different categories, based on the number of schools per district, and created separate dummy variables. We created 11
dummy variables, one for districts with one school, another for districts with two schools, and so on. The maximum number of
schools per district was 15 schools, and there were no districts with 10, 12, 13, or 14 schools in our sample.
9. IVProbit is equivalent to an IVRegress procedure for a linear regression analysis. Using maximum likelihood estimation,
IVProbit fits probit models that have one or more regressors that are correlated with the error term (StataCorp 2013).
98 Stearns, Bottı́a, Davalos, Mickelson, Moller, and Valentino
Table 4. Probit and IVProbit Models Determining Students Chances of Declaring a STEM
Major
Declare a STEM Major
Probit IVProbit
Table 4. Continued
Declare a STEM Major
Probit IVProbit
10. As a plausibility check, we also tested the proportion of math and science courses taught by women in the high school. To create
the variable measuring the proportion of math and science courses taught by women, we also used the information from high
school classroom personnel files between the years 2000-2004. We used the same criteria to determine subject area as were
used in the calculation of the proportion of female teachers, but instead of calculating gender composition based on teacher
identification code and gender, we selected the number of course sections taught by each teacher and then determined the pro-
portion of course sections taught by women out of all the math and science course sections at each high school. This variable is
a proxy for the exposure to female math and science teachers. Like the proportion of math and science teachers, this variable
ranged from 0-100, with a mean of .64. Results are substantively similar to those with the proportion of teachers, but the magni-
tude of the coefficients is somewhat smaller, which suggests that there is an additional influence of the organizational demogra-
phy above simply a heightened probability of having a woman teaching a math or science course.
100 Stearns, Bottı́a, Davalos, Mickelson, Moller, and Valentino
Table 5. Probit and IVProbit Models Determining Students Chances of Graduating with a
STEM Major
Graduate with a STEM Major
Probit IVProbit
Table 5. Continued
Graduate with a STEM Major
Probit IVProbit
11. The dependent variable used here equated physical science and engineering with biology, disciplines that have unequal repre-
sentation of women. We also ran separate models for declaring and/or graduating with biology as a major, and declaring and/
or graduating with a physical science and engineering as a major. On the whole, our findings are consistent for both of these de-
pendent variables when it comes to major declaration, but only for physical sciences/engineering when we consider graduation.
The results for declaring or graduating from a physical sciences/engineering major are stronger than they are for declaring a
STEM major in general (results available from authors).
12. To give the reader some indication of effect size, we have included Appendix D, which presents marginal effects from Tables 4
and 5 for female students. It shows that if the proportion women math and science teachers were 1, girls would have a probabil-
ity of declaring a STEM major of 19 percent and an 18 percent probability of graduating with a STEM major. Both of these
probabilities are higher than female students currently have at the average value of proportion women math and science teach-
ers (.636), representing a 19 percent increase in chances of declaring and a 29 percent increase in chances of graduating with
STEM. These increases are far from being insignificant in magnitude and highlight an important avenue to increase female par-
ticipation in STEM.
102 Stearns, Bottı́a, Davalos, Mickelson, Moller, and Valentino
Table 6. Probit and IVProbit Models Determining Students Chances of Declaring a STEM
Major
Declare a STEM Major
Table 7. Probit and IVProbit Models Determining Students Chances of Graduating with a
STEM Major
Graduate with a STEM Major
significant association between the proportion of black math and science teachers and STEM out-
comes for white or for black female students.
Nevertheless, these tables do not answer the question of whether teacher race also might matter,
such that white female students might be more influenced by white female teachers and African
American female students by African American female teachers. Therefore, in results not shown
(available from authors), we tested whether the proportion of white female math and science teachers
was associated with STEM outcomes for white female students and found that it was not. In other
words, it appears that it is sufficient for the teachers to be female for their presence to be associated
with white girls’ choice of major: they do not also need to be white. We also tested whether the per-
centage of black female teachers was significantly associated with STEM outcomes for African
American girls and found non-significant results as well.
DISCUSSION
Drawing on a theory of representative bureaucracy to explain why math and science teachers may
constitute resources for students in science, technology, engineering, and mathematics (STEM)
relationship to white or African American young men’s choice of major or probability of graduation
with STEM degrees, results that are encouraging from a policy perspective and instructive for the the-
ory of representative bureaucracy. This theory predicted that there might be a negative association
between the percentage of female math and science teachers and young men’s STEM outcomes, an
association produced, in part, if active representation on the part of female math and science teachers
harms young men’s pathways into STEM by privileging young women. It is possible that this lack of
findings is due to the other societal supports that young men are able to find for their interest in
STEM, support that comes from family and friends, for example. It is also possible that female teach-
ers are using different pedagogies within the classroom that are able to engage young men in math
and science topics, engagement that offsets any otherwise negative influence that female math and
science teachers might have on their STEM plans.
The non-significant findings with regard to organizational demography and African American
young women merit some attention as well. It is possible that, as with young men, there are other so-
cietal supports that encourage the potential interest that African American women might have in
STEM (Hanson 2004, 2007; O’Brien et al. 2015). An alternative is that the presence of female math
Furthermore, it is important to acknowledge that we can only generalize our results to the students
in this study due to the nature of our sample being restricted to the pool of students who attended
secondary public school in North Carolina and later pursued their undergraduate studies in the UNC
system. It is possible that there is some selection bias due to the fact that some students attend col-
lege outside the state: as less than 6 percent of graduating seniors intended to attend an out-of-state
institution (North Carolina Department of Public Instruction 2009), we consider this to be a rela-
tively mild bias.
Our findings point to the importance of focusing on the secondary educational careers years—as
well as policies that maintain retention during college—in combating the under-representation of
women in the STEM fields. This is not an issue that can be solved by simply reaching students once
they are already in college: changes to the structural characteristics of the secondary schools children
attend can make profound impacts in the equitable demographic distribution of students who pursue
STEM fields. Our findings show evidence that suggests that a policy that advocates for even more fe-
male secondary math and science teachers in high schools would be effective in increasing the num-
ber of white female STEM majors. Based upon the evidence presented here, policies that increase
Appendix A. SAT Math and Reading Scores of Students who Did and Did Not Matriculate
into the UNC System
Total Did Not Matriculated Young Young Young Young
who Planned Matriculate into the Women who Women who Men who Men who
to Attend into the UNC System Did Not Did Matriculate Did Not Did Matriculate
4-year UNC System Matriculate into the Matriculate into the
College and into the UNC System into the UNC System
Took UNC System UNC System
SAT
Note: Both SAT Math and SAT Reading range from 0 to 80.
Notes: AIC stands for Akaike information criterion. BIC stands for Bayesian information criterion. Given two models, the one with the smaller
AIC or BIC fits the data better.
Teacher Demographics and Girls’ Success in STEM 107
Proportion female Prob > chi2 Prob > chi2 ¼ Prob > chi2 ¼ Prob > chi2 ¼
math and ¼ .0125 .7741 .0455 .7675
science teachers
Proportion female Prob > chi2 ¼ Prob > chi2 ¼ Prob > chi2 ¼ Prob > chi2 ¼
math and .2817 .0338 .76680 .0312
science teachers
Declare Graduate
REFERENCES
Andre, Thomas, Myrna Whigham, Amy Hendrickson, and Sharon Chambers. 1999. “Competency Beliefs, Positive
Affect, and Gender Stereotypes of Elementary Students and their Parents about Science versus Other School
Subjects.” Journal of Research in Science Teaching 36:719–47.
Angrist, Joshua D. and Joürn-Steffen Pischke. 2009. Mostly Harmless Econometrics: An Empiricist’s Companion.
Princeton, NJ: Princeton University Press.
Atkins, Danielle N. and Vicky M. Wilkins. 2013. “Going Beyond Reading, Writing, and Arithmetic: The Effects of
Teacher Representation on Teen Pregnancy Rates.” Journal of Public Administration Research and Theory 23:771–
90.
Bem, Sandra L. 1993. The Lenses of Gender: Transforming the Debate on Sexual Inequality. New Haven, CT: Yale
University Press.
Bettinger, Eric P. and Bridget T. Long. 2005. “Do Faculty Serve as Role Models? The Impact of Instructor Gender on
Women Students.” American Economic Review Papers and Proceedings 95(2):152–57.
Black, Angela Rose and Nadine Peacock. 2011. “Pleasing the Masses: Messages for Daily Life Management in African
American Women’s Popular Media Sources.” American Journal of Public Health 101(1):144–50.
Bottia, Martha, Elizabeth Stearns, Roslyn Mickelson, and Stephanie Moller. 2015. “Growing the Roots of STEM
Majors: Female Math and Science High School Faculty and the Participation of Students in STEM.” Economics of
Education Review 45:14–27.
Byrne, Eileen. 1993. Women and Science: The Snark Syndrome. London, UK: Falmer Press.
Conger, Dylan and Mark C. Long. 2013. “Gender Gaps in College Enrollment: The Role of Gender Sorting across
Public High Schools.” Educational Researcher 42:371–80.
Correll, Shelley J. 2001. “Gender and the Career Choice Process: The Role of Biased Self-Asessments.” American
Journal of Sociology 106:1691–1730.
108 Stearns, Bottı́a, Davalos, Mickelson, Moller, and Valentino
Cronin, Catherine and Angela Roger. 1999. ‘‘Theorizing Progress: Women in Science, Engineering, and Technology in
Higher Education.’’ Journal of Research in Science Teaching 36:637–61.
Dee, Thomas. 2005. “A Teacher Like Me: Does Race, Ethnicity, or Gender Matter?” American Economic Review
95:158–65.
——. 2007. “Teachers and the Gender Gaps in Student Achievement.” Journal of Human Resources 42:528–54.
DiPrete, Thomas A. and Claudia Buchmann. 2013. The Rise of Women: The Growing Gender Gap in Education and What
it Means for American Schools. New York: Russell Sage Foundation.
Dougherty, Jack, Jeffrey Harrelson, Laura Maloney, Drew Murphy, Russell Smith, Michael Snow, and Diane Zannoni.
2009. “School Choice in Suburbia: Test Scores, Race, and Housing Markets.” American Journal of Education
115:523–48.
Downey, Douglas B. and Shana Pribesh. 2004. “When Race Matters: Teachers’ Evaluations of Students’ Classroom
Behavior.” Sociology of Education 77:267–82.
Ehrenberg, Ronald G., D. Goldhaber, and Dominic J. Brewer. 1995. “Do Teachers’ Race, Gender, and Ethnicity
Matter? Evidence From NELS88.” Industrial and Labor Relations Review 48:547–61.
Ehrenberg, Ronald G. and Dominic J. Brewer. 1994. “Do School and Teacher Characteristics Matter? Evidence from
High School and Beyond.” Economics of Education Review 13(1):1-17.
Eisenhart, Margaret, Elizabeth Finkel, and Scott F. Marion. 1996. “Creating the Conditions for Scientific Literacy: A
Lee, James D. 1998. “Which Kids Can ‘Become’ Scientists? Effects of Gender, Self-Concepts, and Perceptions of
Scientists.” Social Psychology Quarterly 61(3):199–219.
Legewie, Joscha and Thomas A. DiPrete. 2014a. “The High School Environment and the Gender Gap in Science and
Engineering.” Sociology of Education 87(4):259–80.
——. 2014b. “Pathways to Science and Engineering Bachelor’s Degrees for Men and Women.” Sociological Science
1:41–48.
Lim, Hong-Hai. 2006. “Representative Bureaucracy: Rethinking Substantive Effects and Active Representation.” Public
Administration Review 66:193–204.
Ma, Yingyi. 2011. “College Major Choice, Occupational Structure, and Demographic Patterning by Gender, Race and
Nativity.” The Social Science Journal 48:112–29.
Maltese, Adam V. and Robert H. Tai. 2011. “Pipeline Persistence: Examining the Association of Educational
Experiences with Earned Degrees in STEM among U.S. Students.” Science Education 5:877–907.
McGrady, Patrick B. and John R. Reynolds. 2012. “Racial Mismatch in the Classroom: Beyond Black-White
Differences.” Sociology of Education 86:3–17.
Meece, Judith L., Beverly Bower Glienke, and Samantha Burg. 2006. “Gender and Motivation.” Journal of School
Psychology 44:351–73.
Meier, Kenneth J. and Jill Nicholson-Crotty. 2006. “Gender, Representative Bureaucracy, and Law Enforcement: The
Riegle-Crumb, Catherine, Barbara King, Eric Grodsky, and Chandra Muller. 2012. “The More Things Change, the
More They Stay the Same? Prior Achievement Fails to Explain Gender Inequality in Entry into STEM College
Majors Over Time.” American Educational Research Journal 49(6):1048–73.
Riegle-Crumb, Catherine and Chelsea Moore. 2013. “The Gender Gap in High School Physics: Considering the
Context of Local Communities.” Social Science Quarterly 95:253–68.
Riegle-Crumb, Catherine and Melissa Humphries. 2012. “Exploring Bias in Math Teachers’ Perceptions of Students’
Ability by Gender and Race/Ethnicity.” Gender & Society 26(2):290–322.
Robst, John, Jack Keil, and Dean Russo. 1998. “The Effect of Gender Composition of Faculty on Student Retention.”
Economics of Education Review 17(4):429–39.
Robinson, Christine R. 1983. “Black Women: A Tradition of Self-Reliant Strength.” Women and Therapy 2:135–44.
Schneeweis, Nicole and Martina Zweimuller. 2012. “Girls, Girls, Girls: Gender Composition and Female School
Choice.” Economics of Education Review 31(4):482–500.
Schneider, Mark and Jack Buckley. 2002. “What do Parents Want from Schools? Evidence From the Internet.”
Educational Evaluation and Policy Analysis 24(2):133–44.
Schuck, Julie A. 1998. Factors Contributing to the Under-Representation of Women in Physics-Based Fields: Final Report to
the Alfred P. Sloan Foundation. Ithaca, NY: Cornell University.
Singer, Ellen R. 1996. “Espoused Teaching Paradigms of College Faculty.” Research in Higher Education 37:659–79.