Summer Research on Causes of the Gender Gap in STEM Fields

Lena Schaefer, Ben Durham, Salim Alwazir, Van Pham, and Jasper Givens 

The students of BK Lab. From top left to bottom right: Van Pham ’24, Ben Durham ’24, Lena Schaefer ’23, Jasper Givens ’25, and Salim Alwazir ’24

Have you ever walked into a science, technology, engineering, or math (STEM) class only to notice in your class of about twenty (or more) there are only one or two women? Maybe most of your STEM teachers in high school were male. On top of this, maybe you’ve noticed the relative infrequency of female role-models in STEM. Given these common experiences, it isn’t so hard to believe that there aren’t as many women who make it all the way through the STEM education system to STEM careers, or don’t advance as far in their STEM career areas when compared to men. These anecdotal observations may partly explain why women make up only 34% of the STEM workforce. In fact, when looking at survey data from the nationally-representative High School Longitudinal Study (HSLS:09) dataset (Exhibit 1), we see that the number of male students who considered majoring in STEM is almost double that of female students.

Exhibit 1. Tabulation of Student Gender and Students’ Consideration in Majoring in STEM

Perhaps even worse is the fact that this issue has the potential to feed into itself as if women are continually driven away from STEM, there are even fewer female role models for future students. Given this, we want to understand why women leave STEM. Our lab examines possible factors that may influence women’s departure from STEM at the transition from high school to college, within college, and from college to career.

Gender Differences in Preferences over Job Attributes Before and During Careers in STEM

Lena Schaefer ’23

I am a rising senior with an Economics major and a German Studies minor. This is my first summer doing research, and I am working with Professor Blume-Kohout in the Economics department. Last semester I studied abroad in Vienna, Austria and I had such an amazing time, but it feels great to be back on campus for the summer! I am from Atlanta, Georgia so it’s nice to be able to be with other Gettysburg students since I wouldn’t be able to see them if I were working at home and I haven’t seen anyone from school in 6 months. Outside of my academics, I am involved with a few other things on campus. I am on the women’s soccer team, a member of the Chi Omega sorority, and a part of the Garthwait Leadership Institute. One thing I love about being able to do research on campus is that I’m still able to continue the traveling life I have gotten to love from last semester. Over the weekends, I am able to explore new things in and around Gettysburg, along with traveling with other members of X-SIG to further places, such as DC and Baltimore. I was even able to travel to Newport over the Fourth of July weekend and see a place I’ve never visited before. 

This summer I am working on research focusing on why there are so few women working in STEM professions. I am focusing more specifically on differences in the importance workers place on specific job attributes by gender, college major, and occupation. For my research I am using data from the 2019 National Survey of College Graduates (NSCG). The survey asks people from ages 75 and under who indicated they received a bachelor’s degree. The NSCG collects extensive employment information such as job status and importance of job attributes, which is the main focus of this research. It also collects information on education, such as all different levels of degrees that were earned by each individual, along with their demographic information. The NSCG is part of an ongoing data collection program through the National Science Foundation, given every two years using the same participants who are still within the age range, and who are still willing to participate. This allows me to compare how preferences for different job attributes have changed, not only across age groups, but also throughout time. 

There are hundreds of variables included in the 2019 NSCG data set that I used, so to be able to use them in a way that is relevant to my research, I created new variables from the already existing ones using a statistical software called Stata. Within the data set, there are missing observations which I removed from the data that I ended up using in order to keep my results consistent. After going through all of the variables and observations, I ended up using around 91,000 observations per data set to find the best results to answer my question. 

The first research question I am looking into is, within my sample, within each STEM field, is there any evidence of gender differences in importance graduates place on various job attributes. In order to find this answer, using the data set, I ran a regression for each job attribute, split up by field of study, and gender. I also only included participants aged 42 and below. For this question, I am also looking at the difference between various occupations. So far, I have run the regressions and am now working on breaking them down and figuring out the important statistics I was able to find and to include in my results tables. 

This Figure shows the results of how likely a female is to say that opportunity for advancement is very important to them compared to men, split up into these individual fields of occupation. To interpret this graph, you must look at where the dots are in relation to the red line, zero. If a dot is very close to the red line, as Biological scientist is in this graph, this means that women in this field are just as likely to say that opportunity for advancement is very important as men are. A dot well above the red line, like Engineer for example, means that women in this field are more likely to say that opportunity for advancement is very important to them. Lastly, a dot below the red line, such as social and related scientists, means that women are less likely to say that the opportunity for advancement is very important to them than men are in this field. The numbers on the side of the graph tell you the percentage points women are more or less likely to say that opportunity for advancement is very important than men. This graph was derived from a table that Stata produced after running regressions with the variables I mentioned earlier. The chart shows the exact numbers, percentage points, and whether or not each result is statistically significant, but the graph shows it in a more visual way. 

Eventually, I will be able to answer two more research questions I have. I want to see if, within each gender, is there a significant difference in importance graduates place on various job attributes over time, with age, or in STEM versus non-STEM careers. I also want to know if across STEM occupations, is there any evidence of systematic differences in the importance of various job attributes. After being able to answer these questions, it can help us understand why there are so few women in STEM careers, even though many more women are graduating with STEM degrees. 

Pathways Through (or Away From) STEM at the College Level

My name is Ben Durham and I am a rising junior mathematical economics and mathematics double major here at Gettysburg College. This is my second summer of X-SIG research with Prof. Blume-Kohout in the economics department. On campus, I have served as a Peer Learning Assistant (PLA) and grader in the math department, and I look forward to working as a PLA again this coming academic year in the math department and the economics department! Outside of my academic interests, I enjoy going to the gym and I compete in powerlifting, so I have a tangential interest in research about strength training and nutrition.

This summer, I’m working on improving our understanding of undergraduate students’ transitions to and from STEM. Understanding why people transition away from STEM majors at the undergraduate level may be useful in determining what could be changed to influence students retention, especially for currently underrepresented groups. Depending on the prevalence of this phenomenon in our sample, I may also be able to study some of the factors associated with switching from non-STEM majors into STEM majors.

To understand these transitions, I make use of a dataset from a large, public, minority serving institution tracking 16,116 students enrolled at this institution from fall 2006 to fall 2015. The data include a student’s declared major over time, any STEM courses a given student takes during their time of enrollment, and several covariates including demographic information, socioeconomic background, and indicators of ability including high school GPA and SAT or ACT students where available. This dataset is unique in its demographic composition, specifically the proportion of Hispanic students. This sample consists of about 40.32% Hispanic students, despite Hispanic students making up only 15% of enrolled college students in 2010. From this data, I can construct a timeline for each student, beginning at enrollment and going through intermediate states such as major declarations until reaching an absorbing state such as graduation or dropout. 

The figure above shows the states I identify in the dataset and possible transitions from intermediate states (outlined in black) to absorbing states (outlined in blue). I use the abbreviation “NS” as a shorthand for “Non-STEM.” To avoid cluttering the diagram, I omit arrows from the intermediate states to Dropout  since a student can transition from any of the intermediate states to Dropout.

Some of the key transitions we would like to understand include the two-way transition between Major STEM and Major NS and the transition between Major STEM and Grad STEM. Examining factors associated with making either transition between Major STEM and Major NS will help us understand what might influence a person already in STEM to leave STEM or what might draw someone into STEM. Examining the transition between Major STEM and Grad STEM will help to understand factors associated with completing a degree for students who are already majoring in STEM.

I will use a competing risks model to identify what factors may play a role in altering students’ trajectories. In this context, the absorbing states Grad STEM, Grad NS, and Dropout are “competing risks” since once an individual enters one of these states, they are no longer “at risk” of transitioning to any of the other states. Some key factors I will consider include instructor race-matching and gender-matching (meaning when the instructor’s race or gender is matched to the student’s race or gender) and classroom gender composition in STEM courses. Conceivably students may be more likely to continue in a field if they have potential role models in the form of instructors of the same race or gender or if they are surrounded by peers that are like them in their degree program. As mentioned, this dataset contains a significant proportion of Hispanic students, which will make it possible to understand race-matching effects which would be much more difficult to do with precision in a dataset containing a more typical proportion of Hispanic students. Additionally, this may allow me to study the even rarer case of simultaneous race and gender matching for non-white students. 

Thus far, I have cleaned and transformed the data into the format required to estimate a competing risks model. Competing risk models require the data be formatted such that there is one observation per individual per time period that the individual is “at risk” of transitioning into one of our states of interest. In our case, this means one observation per semester prior to a student graduating with his or her first bachelor’s degree or dropping out. I investigate cases such as students missing information for key variables to identify patterns among these students to explain why their data may not look as I would expect.

Having reconciled most of the outstanding issues with the dataset, the next step is to specify the model or models I will use to understand these transitions, run these models, and then communicate these results in a paper. This is what will occupy me for the remainder of my research this summer.

The Gender Gap in STEM Fields: The Period of High School into College  

We are Van Pham and Salim Alwazir, we are both rising juniors at Gettysburg College majoring in Mathematical Economics and minoring in Data Science. We are also both international students: Van from Vietnam and Salim from Palestine. This summer we are working together on a research project for Professor Blume-Kohout. It’s very fortunate to have a lot of students doing X-SIG this summer, so the social life is not much different from during the year.  In her free time, Van hangs out with her friends and cooks some good food. She also studies some courses from DataCamp to learn statistics and coding in R to better prepare for the next year at Gettysburg. Salim enjoys hanging out with his friends and visiting cities around the U.S. He also enjoys using online platforms such as Udemy and Coursera to learn and sharpen his technical skills. This is our first research experience in economics and we are very excited to share our work with you. Also, neither of us has used Stata to perform statistical analysis and build regression models before, so it is a new and valuable experience. Furthermore, we both feel excited about how this research can contribute to a lot of change on campus and hopefully nationally. 

We are working on understanding the gender gap in STEM (Science, Technology, Engineering, and Mathematics) fields in the period between high school and college. Our work includes data cleaning, statistical analysis, and estimating relationships between variables. We are aiming to understand the gender gap in STEM by exploring the student-teacher interaction, student achievements, and the student’s classroom experience in high school and college. 

We started our work by summarizing multiple research papers and prior literature to develop a general idea of our research questions and hypothesis, as well as the variables we want to extract from the dataset. The dataset we are using is the High School Longitudinal Study: 2009 (HSLS:09). The dataset  includes over 10,000 variables and information about more than 25,000 students from 940 different schools across the United States.  In Fall 2009, the HSLS:09 surveyed 9th grade students (base year), parents, math and science teacher, administrators, and counselors, then followed up with the student respondents three times: in Spring 2012 (11th grade), Summer and Fall 2013 (after most graduated from high school), and in 2016 (about 3 years after high school graduation). Finally, between Spring 2017 and Fall 2018 (about 4 years after high school graduation), the survey collected college transcripts for students who attended college.  Our study focuses on responses from the students, math and science teachers, and parents. 

For the first three weeks of the summer, we looked into the variables and chose which variables we wanted to use. We had a lot of discussions (and maybe some arguments, too). But overall, we have some very good results, and a lot of new ideas that we exchange with each other. From the variables and from the inspirations of some articles we read as well as the professor’s suggestions, we chose to focus on three possible outcomes (considering majoring in STEM, declared STEM major in college, and completed STEM majors), and investigate how the targeted variables relate to our outcome. Examples of these variables are Science/Math teacher beliefs about who is better in  Math/science, students’ beliefs about who is better in  Math/science, whether the teacher listens to students’ ideas, whether the teacher treats men and women differently, and other variables to understand which factors have a significant impact on our outcomes. 

We are currently looking into replicating the framework of Dario Sansone, who was mainly looking into the relationship between high school student’s beliefs about female abilities in math and science and their teacher’s gender, beliefs, and classroom behaviors. He found that these formed beliefs are related to the decisions by female students to take advanced math and science classes in Sansone’s 2019 paper. We are building on his framework to investigate how all the previously mentioned factors might affect a female student’s decision to major in  STEM, and their probability of completing a STEM degree. We use Stata, a general-purpose statistical software package to test for statistical significance and to estimate regression models, and we have envisioned a plan on how we want to design our models. While working, we have found many interesting things to consider. For example, most teachers believe that women and men have the same abilities in math and science. On the other hand, there are some gendered beliefs that are interesting to explore. For instance, the percentage of math teachers who believe that women are better than men in math is higher than the percentage of math teachers who believe that men are better.  However, the percentage of math teachers who believe that women are better in science is lower than the percentage of math teachers who believe that men are better in science. Additionally, the percentage of science teachers who believe that women are better in science is higher than the percentage of science teachers who believe that men are better in science. However, the percentage of science teachers who believe that women are better in math is lower than the percentage of science teachers who believe that men are better in Math (Figure 1). We can conclude that teachers tend to think that women are better in their field of expertise, but overall have a stereotype that men are better in the field they are not teaching. This informs us that there is a gendered stereotype that men are better at science/math, but the experts in the field state otherwise! Therefore, we are aiming at studying how these beliefs and teacher behavior in the class affects female intentions to major in STEM. 

Next steps will include building our regression model to estimate the effect of the chosen variables on female students declaring and completing a STEM degree. Also, we will be writing our data section where we will explain our key variables and provide some insights about the descriptive statistics, which inform initial results. 

The Impact of Values on STEM Interest and Careers

Hi, I’m Jasper Givens and I am currently a rising sophomore. Outside of school I enjoy playing the guitar, making video essays, and hiking in my home state of Washington. I am majoring in Mathematical Economics and Computer Science and hope to become a machine learning engineer after college. This summer I am working with Professor Blume-Kohout of the economics department on researching the values of people in STEM occupations and undergraduate programs. My research is composed of two main parts: data from the 2012 General Social Survey, and my own research experiment conducted at Gettysburg over the summer. 

Why does it matter?

According to the American Action Forum, the U.S. will be 1.1 million STEM workers short in 2024.  This vital profession group includes scientists, engineers, statisticians and computer scientists. Due to the importance of these groups to the economy, and the increasing need for STEM workers, it is necessary to understand the differences between STEM workers and the general population. Furthermore, STEM careers are some of the worst in terms of gender disparity; women are especially underrepresented in non-life science STEM careers. By investigating the values of STEM workers, we hope to provide information which will allow teachers, professors, and policy makers to better understand how to develop a strong and diverse STEM workforce.

How do we measure values?

In order to measure the characteristics of STEM majors, it was first necessary to establish a metric by which to measure STEM and non-STEM individuals. For this purpose, I am using the Portrait Value Questionnaire (PVQ) instrument first described by Shalom Schwartz in 2002. The PVQ is based on Schwartz’s theory of human values, a system which contains 10 human values which Schwartz theorizes all human cultures share to some extent. The original version of the PVQ contains 40 questions with 4 questions each corresponding to one value. This allows for a consistent measurement of each of the 10 values.

Below is a diagram depicting the 10 Schwartz Human Values. They are organized by their relationship to each other such that adjacent values are more closely related while opposite values are at odds with each other. The test uses a 6-point Likert scale which ranges from “very much like me” to “not like me at all”. Across various experiments, Universalism and Benevolence are usually the highest while Power and Achievement are usually the lowest. You can take the PVQ yourself here.

Figure 1: The 10 Schwartz human values displayed graphically

The 2012 General Social Survey Data

The 2012 GSS contained the PVQ21, a shortened (21 question) version of the original PVQ. The survey also asked whether participants had ever considered a career in science or in engineering, and about their current occupation. Using these variables, I ran linear regressions on the normalized mean of each of the 10 values with career category (never considered STEM, considered STEM, in STEM career) as the dependent variable. The data points from these regressions were used to perform t-test comparisons, which gives the probability that the difference exists given that the independent and dependent variables are uncorrelated. Self-Direction was found to be higher amongst those who considered STEM careers and those who are currently in STEM careers, while Security and Tradition were lower in those same groups.  

What about students at Gettysburg?

For the second part of my research, I administered the PVQ21 to X-SIG students and faculty mentors at Gettysburg College. Students participating in the X-SIG program are usually STEM majors. I am using these data to see if the values most strongly held among STEM students at Gettysburg follow the same pattern as STEM workers in the GSS data.

Below are the results of this work so far, comparing the responses of Gettysburg X-SIG students to those from the General Social Survey.

Figure 2: Schwartz Human Values in GSS and Gettysburg X-SIG respondents

For the remainder of the summer, I will conduct further research on the Gettysburg X-SIG data. Existing literature has shown that men and women on average place importance on different human values; however, this difference has decreased over time. I will examine the relationship between men and women in STEM fields in my data. In doing so I hope to establish whether differences in human values can explain some of the gender disparity in STEM occupations. 

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s