Do AP Courses Make a Difference in STEM College and Career Choices?

High school student with STEM teacher

Many students take Advanced Placement (AP) courses in science, technology, engineering, and mathematics (STEM) subjects to tackle college-level work in high school—and earn college credit and placement in higher-level courses. But some research has questioned whether taking STEM AP courses makes any difference when it comes to pursuing STEM majors in college and STEM occupations.

New research by AIR Senior Vice President and Institute Fellow George Bohrnstedt with AIR colleagues Burhan Ogut, Darrick Yee, and Yifan Bai strongly contradicts those prior findings. In this Q&A, Bohrnstedt explains the findings and reflects on his career as a researcher.

Q. Some prominent researchers found no causal relationship between taking STEM AP courses and then choosing a STEM major in college or expecting to work in a STEM occupation. Your research counters these prior studies. Can you explain?

Bohrnstedt: Researchers have long studied the role of AP coursetaking in students’ pursuit of STEM majors in college. Some early studies found positive results. More recently, a couple such studies received a fair amount of attention. These studies indicated that if you considered students’ previous interest and experience in STEM-related activities, the relationship between taking STEM AP courses in high school and majoring in STEM in college virtually disappeared.

This finding struck me as unlikely, and after looking into it more deeply, I saw what I thought were some limitations in the samples and methodology of those studies. In particular, some of the samples used were limited to a set of colleges that were supported by the National Science Foundation, rather than a broader range of colleges. The researchers asked students in freshman English classes whether they planned to major in STEM, but they didn’t have measures of whether students actually did major in STEM.

We found students were between 1.6 and 2.4 times more likely to major in STEM if they took a STEM AP course in high school.

AIR’s researchers used a more recent longitudinal dataset, the National Center for Education Statistics’ High School Longitudinal Study of 2009, a nationally representative study of ninth graders in 2009, with a first follow-up in 2012 and a second follow-up in 2016. With this sample, we could actually measure in the junior year of college whether students who had taken STEM AP courses had, in fact, majored in STEM. We used all the control variables that other researchers did, plus more. We found students were between 1.6 and 2.4 times more likely to major in STEM if they took a STEM AP course in high school.

Similarly, we found that students who had taken STEM AP courses were more likely to expect to be in a STEM occupation at age 30. Of course, what people say they are going to do and what they do don’t always align. But there is a plan to question this cohort of students again at age 30, so we can find out where they ended up. 

Q. STEM education has been a priority for years, and demand for STEM talent in the workforce is strong and growing. How should policymakers and educators respond to your research findings?

Bohrnstedt: We need to be careful about generalizations because this was not an experimental study. These students weren’t randomly assigned to take STEM AP courses in high school. I also want to be cautious because we know that the STEM pipeline is quite leaky. Also, there are plenty of examples of students who don’t take STEM AP courses in high school but end up in STEM. And not everyone who takes STEM AP courses in high school ends up in STEM. I wouldn’t want anyone to think that if students don’t take STEM AP courses in high school that there’s no chance they can end up in STEM. That’s not the case.

But for me, the most important takeaway is to correct the impression left by prior studies that taking STEM AP courses is not important. The truth is there are many high schools that do not provide access to these courses, especially for students who come from less advantaged backgrounds. We should be looking to expand those opportunities in high schools that have large numbers of such students.

Q. One of your research interests is understanding the role of academic identities on STEM educational outcomes. What are you learning?

Bohrnstedt: We’ve done two studies within the past couple of years that looked at whether a math identity is correlated with mathematics performance. Students who reported that they thought they were good in math and that other people thought they were good in math—which is one definition of how sociologists define identity—did substantially better on the National Assessment of Educational Progress (NAEP) 12th-grade mathematics assessment.

We’ve also done some preliminary analyses with the High School Longitudinal Study on how math and science identities relate to the choice of a STEM major in college. We’re finding significant relationships, even when we take into account the math and science courses students took in high school and how well they did in them, as well as other background variables. We plan to explore this issue further. In particular, we want to find out if these relationships among STEM identities, performance, and college major apply equally well to gender, race, and ethnicity.

Q. You’ve had a long career as a scholar and expert in research methods and statistics. Which ones stand out, and why?

Bohrnstedt: This is pretty esoteric, but in 1969, a colleague at the University of Wisconsin challenged me to develop a statistical formula to improve the quality of the measurement of a particular kind of complex variable.

Let me explain the problem with an example: If you look at the relationship between the amount of education someone’s had and their income, does that relationship vary by, say, their race or socioeconomic status? That is, is everybody getting the same value for education in terms of their eventual income or not? If not, then race or socioeconomic status, or both, have effects on the relationship of education to income. You could measure these effects, but the reliability of the measurement was unknown.

I consulted with Art Goldberger, a noted econometrician at the University of Wisconsin. Then I spent my entire Christmas vacation doing the algebra necessary to derive the statistic need to figure out the reliability of this type of variable. Since then, the paper I published with Art on that statistic has been cited in almost 600 additional articles in virtually every field of science. Even though most people would not associate this with me, it’s the article that I’m most proud of because it has had the largest impact in terms of where it’s been applied and used.

Being careful not to overgeneralize one’s findings is important to me. It’s so easy to let one’s own prejudices get in the way of the work, and to find what you want to find and not consider rival hypotheses that could explain your findings. I was an academic for 20 years and I taught my students to bring a sense of caution to everything we do.

Q. I’m sensing a theme that you are very attentive to research methodology, correct?

Bohrnstedt: Being careful not to overgeneralize one’s findings is important to me. It’s so easy to let one’s own prejudices get in the way of the work, and to find what you want to find and not consider rival hypotheses that could explain your findings. I was an academic for 20 years and I taught my students to bring a sense of caution to everything we do.

Q. What trends in education are you paying attention to?

Bohrnstedt: I am interested in understanding more about the effects of COVID-19. For over 25 years, I chaired the NAEP Validity Studies Panel, which AIR maintains. I’ve been interested in trends in NAEP over time, especially in cases where we found increases in performance for minority students.

Right now, one of the big issues is the significant drops in NAEP performance that occurred with COVID. In one study being led by AIR colleague Martin Hooper, we’re looking at the degree to which changes we observed were due to the performance of students versus changes in the population of students that NAEP measured, given that over a million students didn’t come back to public schools after COVID. We have found that the vast majority of the change in NAEP scores is due to a change in student performance, not to changes in the composition of students taking NAEP.

I’m also interested in understanding the recovery from COVID. One promising intervention is intense tutoring. But when you look at the research on this, it’s amazing how much difficulty districts are having in implementing tutoring. I recently finished reading a book called The Tutoring Revolution, which is a nice example of what the research shows is needed for tutoring to be effective. You have to have a knowledgeable tutor, do a lot of diagnostic testing to make sure students are making progress on a day-to-day basis, and correct things in real time where they’re not. Tutoring also appears to be more effective when it is done in the home with parents involved than in school. Tutoring can be effective, but it’s important to implement it correctly. That requires a major investment, both financially and in terms of human resources needed to do it well.