What an Air Force Academy experiment teaches us about how to help all students shine
By KATHERINE L. MILKMAN | Special To The Washington Post | Published: September 25, 2017
At the U.S. Air Force Academy, freshmen spend the vast majority of their time with a group of other students who aren't selected based on shared physical fitness, choice of college major, comfort with mixed-sex bathrooms, preferred bedtime, neatness or any other attribute. They are chosen by a random lottery.
For the first seven months of freshman year, the 30 or so peers assigned to an Air Force Academy cadet's freshman squadron constitute the cadet's social universe. Even entering the premises of other squadrons is forbidden during this time.
Lotteries are, of course, a fair way to choose squadrons, but their random nature has an added benefit for social scientists: It creates a near-perfect experiment for testing how students' peers affect their performance — a question nearly every parent, teacher and student has pondered.
The answer, based on a team of economists' analysis of three years of data on roughly 3,500 students, is quite a lot. For every 100-point increase in the average verbal SAT scores of a freshman's squadron cohort, that freshman's GPA goes up by 0.4 grade points on a 4.0 scale. That's the difference between getting all A-'s and being a B/B+ student.
Exactly why and how freshmen's squadron peers influence their performance is a little tricky to pin down, and there are probably multiple factors at play. For instance, it could be that immersion in a cohort of outstanding peers creates pressure on less-dedicated students to hit the books. Past research has shown that merely informing students that the majority of their peers are engaging in some behavior (for instance, consuming only moderate amounts of alcohol) is enough to influence them to shift their behavior toward the social norm. (This influence tactic is called social norms marketing, and it has been shown to affect a wide range of decisions ranging from accepting a job at Teach for America to reusing hotel towels.)
However, the Air Force Academy data hints that it may not be social norms but instead the opportunity to form useful study partnerships that primarily drives positive squadron peer effects. In plain language, it's probably not primarily that students see their peers studying that boosts their grades, it's that they have more successful peers to study with.
What suggests this? First, the benefits of having squadron peers with higher SAT verbal scores are largest in math and science courses (relative to foreign language courses and physical education courses, for instance). Emulating the social norm established by hard-working peers would be expected to yield benefits in all subjects, but math and science are arguably subjects where working in study groups with dedicated, smart peers is particularly useful.
Second, although the average SAT verbal scores of a cadet's freshman squadron predict grades in math and science classes quite well, the average SAT math scores of a cadet's freshman squadron don't. This surprising finding offers yet another clue about what drives beneficial peer effects at the Air Force Academy. If SAT verbal scores are associated with strong communication skills and the capacity to form useful study partnerships, this could explain their benefits in math and science classes and again suggest that the opportunity to form valuable study partnerships may be a key driver of positive peer spillover effects in Air Force Academy squadrons.
Now for teachers and parents, the other big question: How can we use all we now know about how peers affect students to help the children in our own lives?
Leading scholars are exploring when and how peer effects can be leveraged to do just that. And this objective led to a bold experiment a few years ago at the U.S. Air Force Academy. The experiment allocated half of students to squadrons using the usual lottery method, and half using an optimized method intended to leverage peer influence to maximize the GPAs of students in the bottom third of the predicted grade distribution. Specifically, this "optimized" sorting method placed low-ability students in squadrons with a large proportion of peers who received high verbal SAT scores, and middle-ability students were sorted into squadrons together.
It did not turn out as hoped, but it taught social scientists a lot.
Interestingly, the low-SAT students actually performed worse when they were surrounded by high-SAT score peers. Within these supposedly optimally designed squadrons, low-performing students avoided interacting with brainy cadets and instead formed smaller homogenous cliques, such that their peers were bad, rather than good, role models.
So what's the takeaway for parents, teachers and policymakers? While students' peers do really affect them, proceed with caution when trying to exert influence over who they hang out with — it can backfire because like attracts like and, besides that, adolescents hate meddling.
Still, there is hope: We can take a lesson from research on social norms marketing and gently point out when the majority of a student's peers are doing something we'd like to see them do more of. And when we do introduce low-achieving students to high-achieving peers, it's clearly important to help them feel like they belong, stave off discouragement by doing all we can to boost their self-confidence and encourage friendships rather than isolation.