The Roots of STEM Achievement: An Analysis of Persistence and Attainment in STEM Majors

Abstract

This article analyzes persistence and attainment in postsecondary science, engineering, technology, and math (STEM) education using data from the Beginning Postsecondary Students Longitudinal Study. Ability is shown to have a consistent impact on STEM performance. Self-efficacy has large estimated impacts, and there is evidence of strong bias against women. High school math preparation and attending small colleges increase the likelihood of noninterested students switching to STEM fields. Overall, there is little evidence that collegiate educational experiences affect persistence or attainment. The results indicate that policies to improve high school math preparation and address the gender gap would be most effective.

JEL Classifications: I21, I28

Keywords

STEM persistence STEM attainment gender bias postsecondary education math preparation

Science, engineering, technology, and math (STEM) remain vital fields for the U.S. economy due to both the demand for STEM jobs and the continual advancement of technology associated with STEM areas. In 2009, President Obama launched the “Educate to Innovate” campaign, a movement dedicated to making American students the highest achievers internationally in STEM. The campaign states that “reaffirming and strengthening America’s role as the world’s engine of scientific discovery and technological innovation is essential to meeting the challenges of this century” (The White House Office of the Press Secretary, 2009). The White House proposes to meet these goals by increasing STEM literacy, improving STEM teaching, and providing opportunities for groups underrepresented in STEM fields (The White House Office of the Press Secretary, 2009).

One of the major strategies focuses on increasing retention of undergraduate STEM majors. The president’s advisors on science and technology report that retaining STEM majors is the most cost-effective and fastest policy option to increase STEM professionals (Olson, Riordan, & Executive Office of the President, 2012). They make five recommendations:

Catalyze widespread adoption of empirically validated practices, advocate and provide support for replacing standard laboratory courses with discovery-based research courses, launch a national experiment in post-secondary mathematics education to address the math preparation gap, encourage partnerships among stakeholders to diversify pathways to STEM careers and create a Presidential Council on STEM Education. (Olson et al., 2012, pp. ii-iii)

Will these recommendations actually increase retention of STEM majors? Many studies have pointed to precollegiate characteristics such as SAT math scores, high school grade point average (GPA), high school rank, and grades during the first semester of college as significant predictors of aptitude and persistence in STEM fields (Crisp, Nora, & Taggart, 2009; Heilbronner, 2011; Thompson & Bolin, 2011; Whalen & Shelley, 2010). A critical analysis of persistence and attainment of STEM degrees is clearly warranted.

This article analyzes the factors that influence success in STEM majors. Through regression analysis of data from the “Beginning Postsecondary Students Longitudinal Study” (BPS) of 2003-2009 from the National Center of Education Statistics, the impact of ability, self-efficacy, and postsecondary educational experiences of college students on persistence and attainment in STEM fields are estimated. The results show that preexisting ability, rather than educational experiences, is a strong predictor of success in STEM fields among interested and noninterested students, which is consistent with the majority of the literature. Furthermore, the estimates here indicate that the most effective policies to increase STEM attainment would be to increase high school math preparation and address the gender gap in STEM fields. Little evidence is found that educational experience significantly affects STEM persistence or attainment.

Background

STEM jobs grew 3 times faster than non-STEM jobs between 2001 and 2011, reaching a total of 7.6 million STEM workers by 2010 (Langdon, McKittrick, Beede, Khan, & Doms, 2011). Of these STEM workers, two thirds held a postsecondary degree (Langdon et al., 2011). This rapid growth is projected to continue by the U.S. Department of Commerce. Between 2008 and 2018, the United States will need approximately 1.3 million additional STEM workers (Langdon et al., 2011).

To keep up with this increase of STEM jobs, American colleges and universities need to increase production of STEM degrees. In 2009, only 24% of incoming college students declared a STEM major (Shapiro & Sax, 2011) and less than half of the students who declare a STEM major graduate with STEM degrees (Price, 2010). Although students enter the STEM “pipeline” in elementary school, and can “leak” out of the pipeline at various points in their academic and professional careers, the majority of STEM attrition occurs during students’ undergraduate careers (Heilbronner, 2011). Analysis of the factors affecting persistence and attainment in STEM degrees is needed to inform policy.

Recruitment Versus Retention

At first glance, it seems that policy makers should focus on retention policies because STEM majors attract more incoming college students than most other fields but less than half of these initial STEM majors graduate with a degree in a STEM field (Chen & Soldner, 2013). Therefore, programs that provide faculty mentors, academic support, peer support, undergraduate research opportunities, and professional development similar to the Howard Hughes Medical Institute (HHMI) Professors Program at Louisiana State University (LSU) should be created (Bayer Corporation, 2012). Accordingly, the “Engage to Excel” report from the president’s advisors on science and technology puts a clear focus on retention of current STEM majors as the most cost-effective policy to increase STEM graduates (Olson et al., 2012).

However, recruitment may be a fruitful strategy as well. By recruiting high-ranking high school students to STEM majors through summer field studies, high school outreach programs, scholarships, internships, and high school–college mentoring programs, STEM departments could make students who possess the capabilities to achieve in STEM more interested in and better prepared for collegiate STEM courses (Bayer Corporation, 2012). By creating interest in high-achieving students, retention rates may increase as many studies have found that academic achievement and STEM preparation in high school is a strong predictor of success in STEM degree attainment.

The current policy focus on retention may be misguided, especially if ability is the key determinant of STEM success. If current students are not completing STEM degrees due to lack of ability, programs aimed at retention are unlikely to be successful. According to Heilbronner (2011), four factors influence the decisions of students to declare a STEM major and to persist in the field: interest, ability, self-efficacy, and educational experiences. We will consider each in turn.

Interest

People who are interested in STEM tend to take pleasure in working with ideas and hands-on problem solving (Carnevale, Smith, & Melton, 2011). Early and continued experience with these realistic and investigative interests is extremely advantageous in maintaining interest in STEM fields (Kokkelenberg & Sinha, 2010). Students who were drawn to STEM at an early age and continued to be interested throughout high school were more likely to declare a STEM major. Furthermore, students who planned to major in STEM before they graduated high school were 3 times more likely to persist in STEM (Maltese & Tai, 2011).

Early entry into a STEM major has been proven to be an extremely influential factor in completing a STEM degree. Moreover, early interest may be a reason that students take more STEM classes (Heilbronner, 2011). Students who took more STEM classes during their first year of college had high levels of success in STEM (Maltese & Tai, 2011). Also, first-year students who enroll in entry-level (gatekeeper) mathematics and science courses were more likely to persist in STEM. Thus, interest, especially early interest, can be seen as an important factor of STEM degree attainment.

Even though students who switched into a STEM major later in their undergraduate career were less likely to complete a STEM degree, interest can be garnered at the collegiate level as well through educational experiences. For example, interaction with faculty members influences women’s interest in and commitment to STEM (Shapiro & Sax, 2011). It is clear that there are stark differences in STEM degree persistence between initially interested and not interested students.

Ability

One of the most prevalent hypotheses in STEM retention research is that success in STEM depends directly on the level of academic ability, especially in mathematics. Much of the literature concludes that ability indicators such as SAT math scores, high school GPA, high school rank, and grades during the first semester of college are predictors of aptitude and persistence in STEM fields (Crisp et al., 2009; Heilbronner, 2011; Thompson & Bolin, 2011; Whalen & Shelley, 2010). High school science grades had the strongest influence on STEM degree attainment in the Gaston Gayles and Ampaw longitudinal study (2011). Thompson and Bolin (2011) even go so far as to recommend that high ranked high school students should be encouraged to declare STEM majors.

Even though traditional measurements of achievement (test scores, grades, ranking, etc.) have been used to predict success in STEM attainment, high school course work has also been a major influence. Advanced placement (AP) work in high school is a predictor of success in obtaining a STEM degree (Kokkelenberg & Sinha, 2010). In addition, students who take more college preparatory courses have higher levels of achievement in the sciences (Crisp et al., 2009). However, many women do not take the necessary high school math courses to do well in STEM majors due to lower levels of math achievement in the eighth grade (Shapiro & Sax, 2011). Kokkelenberg and Sinha (2010) also show that engineering majors who have good math preparation before entering college are more likely to graduate with an engineering degree.

Self-Efficacy

Self-efficacy refers to an individual’s belief that he or she possesses the ability to accomplish a goal. Individuals with high self-efficacy are more likely to persist in the pursuit of expected outcomes (Bandura, 1977). According to Heilbronner (2011), students who believed that they could excel in STEM were more likely to graduate with a STEM degree. This is especially true for women; while men chose to declare STEM majors based on interest and personal use, women were more concerned with performance, ability, opinions of others, and effort (Sullins, Hernandez, Fuller, & Tashiro, 1995). In addition, Shapiro and Sax (2011) found that there are four factors that are likely to influence a woman’s decision to declare a STEM major: self-confidence, a sense of belonging in STEM culture, family influences and expectations, and peers and social groups. A lack of self-efficacy in women could also correlate with the lower math attainment of female eight graders and the deficiency of women with necessary high school math preparation mentioned above.

There exists some “sorting” of women and minorities into non-STEM paths before college that affects the persistence of underrepresented groups in STEM (Griffith, 2010). Hispanic students were found to have been placed in lower level science courses in high school, making it more difficult for them to succeed in STEM in college. In addition, women tend to opt out of math and science courses in middle school making them less prepared for further STEM courses (Burke, 2007). Women were also not taking the necessary high school math courses to do well in STEM majors due to lower levels of math achievement than their male counterparts in the eighth grade (Shapiro & Sax, 2011). However, White students were more likely to take college preparatory courses and have higher levels of achievement in college science courses (Crisp et al., 2009). It is unclear whether women and minorities leave STEM courses due to discrimination, either explicit or implicit, in their educational careers or whether they are more likely to self-select away from STEM. Even self-selection could be driven by discriminatory environments.

Educational Experiences

Ability, interest, and self-efficacy can be improved through educational experiences. Having classroom experiences in the sciences that are interactive and challenging in high school may lead students to declare a science major (Heilbronner, 2011). In addition, more interactive STEM learning experiences in high school correlate with success in first-year STEM courses (Maltese & Tai, 2011). This could be due to the fact that people who are interested in STEM are inclined to take pleasure in working with ideas and hands-on problem solving (Carnevale et al., 2011). However, high school and college STEM experiences tend to be passive. The sciences and other math-intensive disciplines tend to present content through lecture rather than more active learning teaching methods (Shapiro & Sax, 2011). Even though the literature has shown a need for STEM education reform, there has not been enough change in STEM departments who ultimately act as the gatekeepers to STEM degrees (Bayer Corporation, 2012).

STEM departments have been accused of “weeding out” lower performing students through unnecessarily challenging introductory courses and overly harsh grading practices (Olson et al., 2012). According to a survey of STEM department chairs, the majority of them believed that the practice of “weeding out” was harmful. However, the majority did not believe that they needed to change their teaching practices (Bayer Corporation, 2012). In reaction to weeding-out practices, some colleges have put in place programs to nurture rather than alienate their underrepresented and underachieving students. For example, the Howard Hughes Medical Institute (HHMI) Professors Program nearly doubled the retention rate of STEM majors at LSU by giving underrepresented and underachieving STEM majors additional opportunities to be mentored by faculty and peers, to receive academic support, and to participate in undergraduate research experiences (Wilson et al., 2012).

Also, instructors can affect persistence in the STEM pipeline. According to Takacs and Chambliss (2014), experiences with faculty members, especially in an introductory course, influence persistence in the major. Students tend to learn more from professors hired to teach rather than to do research (Figlio, Schapiro, & Soter, 2013). Black STEM majors who have Black instructors in their first year of study are more likely to persist in STEM after their first year (Price, 2010).

In contrast, female STEM majors who have female instructors are less likely to persist in STEM majors according to Price (2010). However, Shaprio and Sax call for more interaction between female students and women who are successful in the STEM field (2011). This conflict in opinions may be due to the fact that women are concerned with performance, ability, opinions of others, and effort (Sullins et al., 1995), but also need self-confidence and a sense of belonging in the STEM culture (Shapiro & Sax, 2011). Therefore, women might feel threatened by professors who measure their performance and ability, but welcomed by role models and peers.

Furthermore, the practice of grade discrimination could also influence minority students’ decisions to persist in STEM majors. There exists evidence that suggests that Republican professors give relatively lower grades to Black students after accounting for ability (Bar & Zussman, 2012). In addition, there tend to be more Republican professors in the sciences (Bar & Zussman, 2012) which could suggest that there exist higher levels of grade discrimination in STEM fields. Because grades communicate students’ strengths and abilities to graduate schools, possible employers as well as themselves (Bar & Zussman, 2012), students may rely upon them to enter or exit a major. Moreover, they could affect a student’s self-efficacy because his or her perceived performance is lower than the actual performance. Therefore, grade discrimination could cause potential successful STEM majors to exit STEM.

Educational experiences do not end in the classroom; campus life can also influence success in STEM. According to a study done at Spelman College, one of two historically Black women-only colleges in the United States, easy access to small class sizes, undergraduate research opportunities, faculty members, cooperative peers, and academic resources garner success for minorities in STEM fields (Perna, Gasman, Gary, Lundy-Wagner, & Drezner, 2010). Participation in campus activities as well as living on campus also had a small impact on STEM degree attainment (Gayles & Ampaw, 2011).

Multiple factors affect persistence and attainment in STEM fields, and given the growth in STEM jobs and attention on STEM from policy makers, analysis of the relative importance of various factors is vital for good policy. Following the breakdown of factors into categories of interest, ability, self-efficacy, and educational experience, this article analyzes the relative strength of various factors in STEM persistence and attainment through regression analysis.

Method

The BPS was completed from 2003 to 2009. The National Center of Education Statistics issued surveys to 16,100 beginning postsecondary students during the 2003-2004 school year and sent them follow-up surveys in 2006 and 2009 to measure their progress in attaining their degree. This article focuses on students pursuing a bachelor’s degree in 2003-2004. A STEM major is defined as any student who reported working toward a bachelor’s degree in biological/biomedical sciences, computer/information science/support, engineering, mathematics and statistics, physical sciences, science technologies/technicians, or engineering technologies/related fields. These majors are in alignment with the list of STEM majors created by the U.S. Department of Commerce (Langdon et al., 2011).

The analysis focuses on four questions related to STEM persistence:

What factors predict that incoming STEM majors who graduate will attain a STEM degree?

What elements affect incoming STEM majors’ persistence in college?

What variables influence non-STEM majors who graduate college to switch to and attain a degree in a STEM field?

What factors motivate undecided majors to declare and graduate with a STEM degree?

This study follows the framework set by Heilbronner (2011) and breaks up the independent variables into the categories of interest, ability, self-efficacy, and educational experience. The first two questions measure persistence and attainment of students who were initially interested in STEM. The latter two questions analyze STEM degree attainment in students who do not initially declare a STEM major and thus are not initially interested in STEM. The impact of interest in STEM attainment is thus considered indirectly by comparing results from samples of interested and noninterested students.

To determine what other factors (namely, self-efficacy, ability, and educational experiences) determine STEM persistence, four logistic regression models are estimated based on the questions above. Each of the models examined the same independent variables as similar elements influence persisting in, switching in, dropping out of, and declaring a STEM major.

Ability was measured by SAT/ACT scores, first semester GPA, and highest level of math taken in high school (Crisp et al., 2009; Heilbronner, 2011; Thompson & Bolin, 2011; Whalen & Shelley, 2010). While it is difficult to define and measure self-efficacy directly, it has been found that underrepresented groups in STEM majors, especially women, are heavily affected by factors surrounding self-efficacy including self-confidence and a sense of belonging (Shapiro & Sax, 2011). Thus, gender and race (White, African American, Asian/Pacific Islander, Hispanic, Other) are used as proxies of self-efficacy. In addition, other socioeconomic indicators including highest level of parental education, hours worked per week, and expected family contribution are considered.

Finally, educational experiences are measured by institutional characteristics in addition to various reported student experiences inside and outside the classroom. General institutional characteristics such as Carnegie classification (research, master’s, baccalaureate) and selectivity are included, as well as questions of whether students had graduate student instructors, lived on campus, had large class sizes, met with academic advisers, contacted faculty outside of class, joined study groups, and participated in on-campus organizations.

Four models are estimated based on the four questions listed previously. The first question addresses the factors that lead to completion of a STEM degree for incoming STEM majors who graduate college. The sample for this question is limited to incoming STEM majors who graduate; of those, 64% completed a STEM degree. This sample is made up entirely of students who are initially interested in STEM fields, as indicating by their initial choice of major. The dependent variable is thus an indicator for graduating in a STEM field, and the specification is as follows:

Y_{1} = β_{0} + X ’ β_{1} + W ’ β_{2} + Z ’ β_{3} + ε_{1},

where X is vector of variables measuring student ability, W is a vector of variables that serve as proxies for self-efficacy, and Z is a vector of variables reflecting educational experiences of the student. Logistic regressions of Equation 1 are estimated.

The second question addresses the factors that contribute to persistence in college of incoming STEM students. The sample is comprised of all incoming STEM majors, and the dependent variable is an indicator for whether an incoming STEM major graduated college or not, regardless of their graduating field. Overall, 64% of incoming STEM majors graduated college within 6 years. The specification is as follows:

Y_{2} = β_{4} + X ’ β_{5} + W ’ β_{6} + Z ’ β_{7} + ε_{2},

With X, W, and Z representing the same vectors as in Equation 1. Again, logistic regression of Equation 2 is estimated. The third question asks what variables influence switching in to a STEM field from a non-STEM major. The sample includes all incoming non-STEM majors who graduated. The dependent variable is an indicator for graduating in a STEM field. Overall, 5% of non-STEM incoming majors who graduate complete a STEM degree. The specification is parallel to Equations 1 and 2:

Y_{3} = β_{8} + X ’ β_{9} + W ’ β_{10} + Z ’ β_{11} + ε_{3} .

Finally, the fourth question considers what variables influence choosing and completing a STEM degree for undeclared majors. The sample for this model includes students who entered college without declaring a major and graduated. The dependent variable is an indicator for whether or not an undeclared major chooses and completes a STEM degree. The specification mirrors those presented before:

Y_{4} = β_{12} + X ’ β_{13} + W ’ β_{14} + Z ’ β_{15} + ε_{4} .

Results

Persistence to the STEM Degree

Equation 1 analyzes factors influencing persistence to a STEM degree among incoming STEM majors who graduate. The summary statistics for the first sample are presented in Table A1 in the online appendix. Overall, those who persisted in STEM had SAT scores 90 points higher, had first semester college GPAs 0.3 points higher, and were more likely to take precalculus and calculus than incoming STEM majors who switched out of STEM fields. In addition, those who graduated STEM tended to work fewer hours per week and have parents who were more educated and were expected to contribute more money to their education. However, the average educational experiences were similar for both groups. Logistic regression estimates are presented to determine the most important factors contributing to STEM persistence.

Table 1 presents logistic regression estimates of Equation 1.¹ While all variables in Table A1 were initially considered, Table 1 presents the preferred specification that includes statistically significant indicators (complete results and specification tests are available upon request). The ability indicators of SAT/ACT score and first semester GPA were statistically significant and positively related to STEM degree attainment across multiple regressions (only the preferred specification results are shown). In addition, results show a statistically significant negative relationship between females and STEM degree attainment and a statistically significant positive relationship between Asians and STEM attainment. Out of all the educational experience variables, study groups was the only one that had a statistically significant impact on persistence in the STEM major. These estimates point to ability and self-efficacy, not educational experience, as the most influential factors in persistence in STEM majors.

Table 1.

Persistence in STEM for Incoming STEM Majors Who Graduate.

Independent variable	Coefficient	p value
Admissions test scores (ACT or SAT)	0.0022***	.0028
Grade point average 2003-04	0.0077***	.0003
Gender
Female	−0.5846**	.0110
Race/ethnicity
Black or African American	0.0011	.9982
Hispanic or Latino	0.5209	.1885
Asian	1.1988**	.0104
Other	0.3154	.5613
Frequency 2004: Study groups
Sometimes	0.3519	.1930
Often	0.7148**	.0323
Observations^a	900
Likelihood ratio	0.1216

Note. STEM = science, engineering, technology, and math.

Per National Center of Education Statistics (NCES) Standards, the true sample size has been modified to minimize disclosure risk of individual survey responses.

Significance at 10% level. **Significance at 5% level. ***Significance at 1% level.

Persistence in College of Incoming STEM Majors

Equation 2 considers the factors that affect persistence in college of incoming STEM majors. Overall, 64% of incoming STEM majors graduated from college. Those who did graduate had SAT scores 119 points higher, first semester college GPAs 0.6 points higher, and were more likely to complete precalculus or calculus in high school than those who did not graduate college. In addition, the STEM students who did not graduate were less likely to attend selective institutions, have parents with a college education, or have families that could contribute more than US$10,000 to their schooling on average. Table A2 in the online appendix shows summary statistics for the second sample.

Table 2 shows the logistic regression results of the preferred specification. Test scores, gender, parental education, hours worked per week, selectivity of the institution, and participation in school clubs were all statistically significant. Again, the ability indicator of SAT score was a predictor of success. In contrast with Model 1, there was a direct relationship between females and graduation. The two models together suggest that women are more likely to graduate college but less likely to attain a STEM degree.

Table 2.

General Persistence in College for Incoming STEM Majors.

Independent variable	Coefficients	p values
Admissions test scores (ACT or SAT)	0.0019***	.0070
Gender
Female	0.7016***	.0004
Parents’ highest level of education
Some college	0.2493	.4790
Bachelor’s degree	0.8760***	.0025
Graduate degree	−0.0352	.9137
Job hours worked per week (incl. work study)	−0.0296***	.0003
First institution selectivity 2003-2004
Very selective	1.3281***	.0066
Moderately selective	0.8487*	.0525
Frequency 2004: School clubs
Sometimes	0.3884**	.0268
Often	1.0464***	.0017
Observations^a	1,100
Likelihood ratio	0.1931

Note. STEM = science, engineering, technology, and math.

Per National Center of Education Statistics (NCES) Standards, the true sample size has been modified to minimize disclosure risk of individual survey responses.

Significance at 10% level. **Significance at 5% level. ***Significance at 1% level.

In addition, the socioeconomic factors of hours worked per week and level of parental education affected the probability of graduating. This could be due to the heavier financial or cultural burden placed on these students. Participation in school clubs may have either kept students interested in school or was a product of amount of free time students had because they did not have a multitude of commitments outside of class. Overall, very few institutional factors predicted general persistence in college for STEM majors. The statistically significant indicators were very much in line with persistence in all majors. (Additional regression specifications [not shown] on degree attainment in all incoming majors revealed similar results.)

Attainment of STEM Degree Among Incoming Non-STEM Majors

Equation 3 considers the factors influencing attainment of a STEM degree by students who did not initially select a STEM major. Only 5% of incoming non-STEM majors switched into and graduated in a STEM field. Those who graduated with a STEM degree had SAT scores that were 44 points higher and were more likely to have taken precalculus or calculus in high school. Otherwise, the characteristics of students who switched into STEM were similar to the broader population. Table A3 in the online appendix shows the summary statistics for this sample.

Table 3 shows the significant variables from logistic regression estimates of Equation 3. Highest level of high school mathematics taken, gender, race, the institution’s Carnegie classification, and study group attendance were all significant factors in determining whether or not incoming non-STEM majors would switch into and graduate in STEM field. Whether or not a student had taken calculus in high school was one of the variables that best predicted STEM degree attainment. Again, females were less likely to graduate with a STEM degree than their male counterparts while Asians were more likely to switch into a STEM field and graduate than any other ethnic group.

Table 3.

STEM Degree Attainment in Incoming Non-STEM Majors.

Independent variable	Coefficients	p values
Highest level of high school mathematics
Precalculus	0.2654	.4178
Calculus	1.1543***	.0000
Gender
Female	−0.5005**	.0250
Race/ethnicity
Black or African American	0.1733	.7235
Hispanic or Latino	−0.2251	.8193
Asian	0.7466**	.0234
Other	0.2812	.7064
Carnegie classification
Master’s	−0.3060	.2781
Baccalaureate	0.6488***	.0077
Frequency 2004: Study groups
Sometimes	0.4551	.1034
Often	0.9723***	.0005
Observations^a	2,600
Likelihood ratio	0.0343

Note. STEM = science, engineering, technology, and math.

Per National Center of Education Statistics (NCES) Standards, the true sample size has been modified to minimize disclosure risk of individual survey responses.

Significance at 10% level. **Significance at 5% level. ***Significance at 1% level.

There were also educational experiences that affected STEM degree attainment in incoming non-STEM majors. Student who attended a school that only granted baccalaureate degrees were more likely to graduate with a STEM degree. In addition, students who often attended study groups had a higher rate of STEM degree attainment. The statistically significant variables imply that ability and self-efficacy were important to STEM degree attainment in incoming non-STEM majors. Furthermore, type of institution may be important because professors may be more focused on teaching than research at a nondoctoral granting institution.

Attainment of a STEM Degree by Incoming Undeclared Students

Equation 4 considers the factors that influence the attainment of a STEM degree by students who entered college as undeclared. Twenty percent of the undeclared majors in the sample attained a STEM degree. On average, those who did earn a STEM degree had an SAT score 102 points higher than those who graduated in a non-STEM major. In addition, students who came to college as undecided majors were more likely to have parents who completed some graduate school course work. Table A4 in the online appendix presents the summary statistics.

Table 4 shows the statistically significant variables from logistic regression estimates of Equation 4. SAT/ACT score, high school GPA, highest level of high school mathematics taken, gender, race, and Carnegie classification were all statistically significant. Again, the various indicators pointed to ability being an important factor in STEM degree attainment, especially taking calculus in high school. Similar to Models 1 and 3, there was an inverse relationship between females and STEM degree attainment as well as a direct relationship between Asians and persistence in STEM fields. Also, students who attended universities that only grant bachelor’s degrees were more likely to declare and complete a STEM degree, implying that institutions that are more focused on teaching than research may better nurture students who are not initially interested in STEM.

Table 4.

STEM Degree Attainment for Incoming Undeclared Majors.

Independent variable	Coefficients	p values
Admissions test scores (ACT or SAT)	0.0019***	.0044
Highest level of high school mathematics
Precalculus	0.7951**	.0351
Calculus	1.1337***	.0087
Gender
Female	−0.7309***	.0001
Black or African American	0.1142	.9265
Hispanic or Latino	−0.1267	.8122
Asian	0.9761***	.0006
Other	0.1847	.6936
Carnegie classification
Master’s	0.3994	.1934
Baccalaureate	0.8879**	.0185
Graduate student instructors
Sometimes	0.4691*	.0912
Often	0.1395	.6928
Academic 2004: large classes
Sometimes	0.0543	.8585
Often	0.7414**	.0304
Frequency 2004: Faculty talk outside class
Sometimes	−0.2105	.4753
Often	−0.7317**	.0357
Observations^a	1,500
Likelihood ratio	0.1036

Note. STEM = science, engineering, technology, and math.

Per National Center of Education Statistics (NCES) Standards, the true sample size has been modified to minimize disclosure risk of individual survey responses.

Significance at 10% level. **Significance at 5% level. ***Significance at 1% level.

Interestingly, large class sizes were statistically significantly positively correlated with STEM degree attainment in incoming undeclared majors, as was an indicator for sometimes having graduate student instructors. Also, talking to faculty members outside of class was statistically significantly negatively correlated with STEM degree attainment. These surprising estimates may result from STEM classes being larger than humanities classes because they tend to be lecture-style classes rather than discussion based. It is thus possible that these three variables are simply capturing an impact of taking more STEM courses.

Discussion

The goal of this research is to better understand the determinants of persistence and attainment in STEM fields by analyzing the impacts of ability, self-efficacy, and educational experiences on students who were initially interested or not interested in STEM fields.

As expected, measures of ability had consistent and significant estimated impacts on STEM persistence and attainment. For example, an incoming STEM student scoring 100 points higher on the SAT is estimated to be 5.5% more likely to persist in their STEM field and is also 5% more likely to graduate. An incoming undeclared major with a 100-point higher SAT score is estimated to be 5% more likely to choose and complete a STEM major. Incoming STEM majors with first-year GPAs one half of a point higher are 9% more likely to complete their STEM degree. It is clear that ability matters, but it is not the only or even the largest determinant of STEM persistence and attainment.

However, there is evidence that for noninterested students, high school math experiences have a large impact on the likelihood that they will end up in STEM fields. For incoming non-STEM majors, having taken calculus in high school is estimated to increase the probability that they switch to a STEM field by 29% compared with students who have less than precalculus. For incoming undeclared majors, the number is 28% and even having precalculus increases the likelihood of declaring a STEM major by 20% over students with less than precalculus. These results may be driven by math ability—students who are more competent and confident in math would naturally be more inclined to consider STEM after getting to college. They also may capture an ease of transition: Students without strong math backgrounds might have to take remedial math courses in college to even consider a STEM major. The policy ramifications are clear: Increasing the number of high school students who take higher math courses is likely to lead to more STEM majors in college.

The results here show large estimated impacts of the measures of self-efficacy on STEM persistence. Incoming female STEM majors are estimated to be 15% less likely than males to complete a STEM degree, despite the fact that they are 17.5% more likely to graduate college than men. Estimates thus indicate that a female incoming STEM major needs an SAT score 300 points higher than a male counterpart to have even odds of completing a STEM degree. Female non-STEM majors are estimated to be 12% less likely to switch to STEM, and female undeclared students are 18% less likely to choose a STEM major.

The result that female students are more likely to graduate but less likely to complete STEM degrees suggests that competent female students leave STEM fields. The regressions here control for ability and math preparation, so these results suggest that there is significant bias against women in STEM. The literature review suggests that this might have to do with women’s concern with performance, ability, opinions of others, and effort (Sullins et al., 1995), or their family influences and a lack of a sense of belonging in the STEM culture (Shapiro & Sax, 2011). However, further research on ways to identify and combat such bias is needed.

Interestingly, the results here do not indicate bias in STEM against traditionally underserved minorities (Blacks and Hispanics). There are, however, strong and significant impacts of being Asian. Incoming Asian STEM majors are estimated to be 30% more likely to complete their STEM degree, although they are not more likely to graduate in general. The estimates suggest that an Asian STEM major with an SAT score 400 points lower and a GPA half a point lower than a non-Asian counterpart has an equal chance of completing a STEM degree. These estimates are large and consistent, indicating that Asian students simply do better in STEM. Asian non-STEM majors are 19% more likely to switch to STEM and Asian undeclared majors are 24% more likely to declare a STEM field. It is likely that these strong results are driven by cultural differences such as the value placed on science and mathematics by different people. Once again, the results indicate that further research is needed to better understand how and why Asian students are more successful in STEM.

Even though many educational experiences were studied, there were few significant effects. Results show large estimated impacts of participation in study groups. STEM majors who often participate in study groups are 26% more likely to graduate and 18% more likely to graduate STEM. Non-STEM majors who participate often in study groups are 24% more likely to switch to and complete a STEM degree. This result is intuitive as study groups both foster connections among students and help them learn the material. However, it is also likely due to better performance of assertive and responsible students. Thus, official encouragement of study groups may have some improvement in STEM persistence, but the estimates here may also be driven by high self-efficacy in general.

Attending a baccalaureate-only institution is the only postsecondary educational experience variable that had a significant impact as expected. Non-STEM majors at baccalaureate-only institutions are estimated to be 16% more likely to switch to STEM fields, and undeclared majors at baccalaureate-only institutions are 22% more likely to declare STEM majors. This result may come from broader general education requirements at small liberal arts colleges that allow students to take more introductory STEM courses. There is, thus, some indication here that requiring students to take introductory STEM courses may lead to more STEM majors. This question deserves further attention and research as it implies a simple policy remedy to increase the number of STEM majors.

Three variables measuring educational experiences were estimated to have a significant impact on the probability that undeclared majors would choose STEM fields. However, all three coefficients, on indicators for having graduate student instructors, having large classes often, and talking to faculty often, had the opposite sign of what was expected. Students who reported often having graduate student instructors were 12% more likely to choose a STEM major, and those who had large classes often were 19% more likely. Students who reported talking to faculty often were 18% less likely to declare a STEM major. It is likely that these counterintuitive results are driven by some unobserved factors. Introductory STEM classes may be more likely to be large lectures with graduate instructors for some or all classes and little interaction with faculty. Taken together, these variables may simply indicate that students who took more introductory STEM courses were more likely to declare STEM majors. Unfortunately, we lack the data to investigate this further and determine whether that result is driven by interest or whether students took those courses due to requirements and the course sparked an interest.

The results here do not indicate that educational experiences at the postsecondary level have a large impact on STEM persistence and attainment. Indeed, one of the two educational experience measures that significantly affects persistence of STEM majors is often attending study groups, which, as discussed previously, likely indicates high self-efficacy. In addition, there is evidence that greater participation in school clubs along with less time working improves the likelihood that students will graduate. However, these results are not limited to STEM fields.

Conclusion

There is high demand for STEM jobs and large societal benefits to growth in STEM production and research throughout the economy. There is thus strong interest in effective policy to improve STEM education in the United States, as evidenced by the focus of the White House on STEM policy initiatives. However, the recommendations of the “Engage to Excel” report (Olson et al., 2012) focus on postsecondary educational experiences and recommends changing science instruction and experimenting with postsecondary math courses. The evidence here indicates that postsecondary educational experiences have little impact on STEM persistence or attainment. Instead, we find consistent impacts of ability, consistent with the literature on STEM attainment.

There is evidence of a strong gender gap in STEM fields, which should be examined further. Women are more likely to graduate college but much less likely to complete STEM degrees, even after controlling for ability and math background. There thus seems to be a strong bias against women in STEM that needs to be addressed.

Finally, results here indicate that simple policies may have strong effects. Encouraging or requiring more students to take precalculus and calculus in high school is likely to increase the number of students who switch in to STEM majors in college. Furthermore, requiring more introductory STEM courses, as many small baccalaureate institutions whose broader degree requirements do, may increase the likelihood that students will switch into STEM fields. A simpler and more direct policy focus may therefore be most effective.

Footnotes

Authors’ Note

This article began as a senior research project at Stetson University.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

Author Biographies

Alan Green is Associate Professor of Economics and department chair at Stetson University in DeLand, Florida. His research focus is economic development, including the role of education in development, and his horizons are continually broadened through close interactions with exceptional students, particularly as they work on senior research projects at Stetson. He also conducts research on effective pedagogy.

Danielle Sanderson is a doctoral student of educational policy at Michigan State University. She specializes in the economics of education with focuses on teacher labor markets, school choice, accountability, and student transportation. Before enrolling in her doctoral program, she taught high school mathematics at a charter school in New Orleans.

References

Bandura

(1977). Self-efficacy: Toward a unifying theory of behavioral change. Psychological Review, 84, 191-215.

Bar

Zussman

(2012). Partisan grading. American Economic Journal: Applied Economics, 4, 30-48.

Bayer Corporation. (2012). Bayer facts of science education XV: A view from the gatekeepers—STEM department chairs at America’s top 200 research universities on female and underrepresented minority undergraduate STEM students. Journal of Science Education and Technology, 21, 317-324.

Burke

(2007). Women and minorities in STEM: A primer. In Burke

Mattis

(Eds.), Women and minorities in science, technology, engineering, and mathematics: Upping the numbers (pp. 3-28). Northampton, MA: Edward Elgar Publishing.

Carnevale

Smith

Melton

(2011). STEM: Science, technology, engineering, and mathematics. Washington, DC: Georgetown Center on Education and the Workforce.

Chambliss

Takacs

(2014). How college works. Cambridge, MA: Harvard University Press.

Chen

Soldner

(2013). STEM attrition: College students’ paths in and out of STEM fields. Washington, DC: U.S. Department of Education.

Crisp

Nora

Taggart

(2009). Student characteristics, pre-college, college, and environmental factors as predictors of majoring in and earning a STEM degree: An analysis of students attending a Hispanic serving institution. American Educational Research Journal, 46, 924-942.

Figlio

D. N.

Schapiro

M. O.

Soter

K. B.

(2013). Are tenure-track professors better teachers? Cambridge, MA: National Bureau of Economic Research.

10.

Gayles

J. G.

Ampaw

F. D.

(2011). Gender matters: An examination of differential effects of the college experience on degree attainment in STEM. New Directions for Institutional Research, 2011(152), 19-25. doi:10.1002/ir.405

11.

Griffith

A. L.

(2010). Persistence of women and minorities in STEM field majors: Is it the school that matters? Economics of Education Review, 29, 911-922.

12.

Heilbronner

N. N.

(2011). Stepping onto the STEM pathway: Factors affecting talented students’ declaration of STEM majors in college. Journal for the Education of the Gifted, 34, 876-899.

13.

Kokkelenberg

E. C.

Sinha

(2010). Who succeeds in STEM studies? An analysis of Binghamton university undergraduate students. Economics of Education Review, 29, 935-946.

14.

Langdon

McKittrick

Beede

Khan

Doms

(2011). STEM: Goods jobs now and for the future (No. #03-11). Washington, DC: U.S. Department of Commerce Economics and Statistics Administration.

15.

Maltese

Tai

(2011). Pipeline persistence: Examining the association of educational experiences with earned degrees in STEM among U.S. students. Science Education Policy, 95, 877-907.

16.

Olson

Riordan

D. G.

, & Executive Office of the President. (2012). Engage to excel: Producing one million additional college graduates with degrees in science, technology, engineering, and mathematics. Report to the president. Washington, D.C.: Executive Office of the President.

17.

Perna

L. W.

Gasman

Gary

Lundy-Wagner

Drezner

N. D.

(2010). Identifying strategies for increasing degree attainment in STEM: Lessons from minority-serving institutions. New Directions for Institutional Research, 2010(148), 41-51. doi:10.1002/ir.360

18.

Price

(2010). The effect of instructor race and gender on student persistence in STEM fields. Economics of Education Review, 29, 901-910.

19.

Shapiro

C. A.

Sax

L. J.

(2011). Major selection and persistence for women in STEM. New Directions for Institutional Research, 2011, 5-18.

20.

Sullins

E. S.

Hernandez

Fuller

Tashiro

J. S.

(1995). Predicting who will major in a science discipline: Expectancy? Value theory as part of an ecological model for studying academic communities. Journal of Research in Science Teaching, 32, 99-119.

21.

Thompson

Bolin

(2011). Indicators of success in STEM majors: A cohort study. Journal of College Admission, 212, 18-24.

22.

Whalen

D. F.

Shelley

II (2010). Academic success for STEM and non-STEM majors. Journal of STEM Education: Innovations and Research, 11, 45-60.

23.

The White House Office of the Press Secretary. (2009). President Obama launches “educate to innovate” campaign for excellence in Science, Technology, Engineering & Math (STEM) education [Press release]. Retrieved from http://www.whitehouse.gov/the-press-office/president-obama-launches-educate-innovate-campaign-excellence-science-technology-en

24.

Wilson

Z. S.

Holmes

deGravelles

Sylvain

M. R.

Batiste

Johnson

. . . Warner

I. M.

(2012). Hierarchical mentoring: A transformative strategy for improving diversity and retention in undergraduate STEM disciplines. Journal of Science Education and Technology, 21, 148-156.