Peer effects are positive in task-oriented groups, but the opposite is true in broad social settings
The effects of high-performing peers on student performance may actually be negative in a broad social situation, according to research by Ahmed Rahman, to be presented at the Royal Economic Society''s annual conference at the University of Bristol in April 2017.
His study tracks the performance of 20,000 students in the United States Naval Academy on their compulsory courses. Because the students had no control over whom they lived with or whom they studied with, differences in peer effects could not be explained by students choosing particular friends or courses.
Students who lived alongside high performers had slightly worse grades than would be expected: a 100-point increase in the average verbal SAT score of company-mates reduces a student''s STEM course grade, on average, by 0.16 grade points. But when their classmates were the high performers, the peer effect became positive.
The author comments: ''In broad social settings, students of like characteristics tend to interact more with each other, so higher ability students can drive average students away from them and worsen academic performance. But in more task-oriented groups students are more willing to overcome differences and work together. Policy-makers should take note: just throwing smarter students in the group can backfire and make overall performance worse.''
Higher quality peers in broad social settings can actually hurt student performance in the classroom. That is the central finding of our paper ''Bad Company: Reconciling Negative Peer Effects in College Achievement.''
Most social policies are predicated on the assumption of significant positive peer effects – a person of mediocre ability should perform better if compelled to interact with a person of high ability. But in the area of education, the jury is still out.
Studies from a range of colleges have suggested peer effects of various sizes, both positive and negative, on a range of academic outcomes. The wildly inconsistent findings from the literature suggest that policy-makers who seek to affect the composition of students, with the intent to improve overall student performance, are as of yet still peering into a black box.
Part of the problem is what is commonly referred to as ''selection bias''. This occurs when an individual chooses, or is chosen, to join a social group based on characteristics that correlate with group outcomes. A hard working student may choose to befriend other hard workers, or ambitious parents may choose a classroom for their child with a highly reputable teacher. In each case the group was formed in part based on educational factors that themselves affect educational outcomes.
Another consideration is the fact that each study measures a different social setting, where the nature of social interactions may differ. Students in College X or College Y will have different study habits, different types of social functions, different cultural norms. Differences across studies in signs and magnitudes of peer effects stem in part from these different settings, and this limits what we can learn about policy prescriptions.
The study here attempts to solve these problems by looking at a college with no student discretion over living arrangements or classes, and by observing groups for the same school in different social settings. Specifically, we examine data on college freshmen at the United State Naval Academy, using a dataset that includes fall semester grades from over 20,000 freshmen.
All students live in dormitories called ''companies'', and all freshmen take a set of standard courses. They have no discretion over their dorm-mates or their classmates. This gives us an ideal environment to study the academic peer effects between those living in the same social environment (''company-mates'') and between those living in the same environment AND engaged in the same tasks (''course-company-mates'').
What we find is startling. Across many types of classes, average peer ability across all freshman company-mates negatively affects own grades. The magnitude of the peer effect is small but consistent – for example, a 100-point increase in company-mates'' average verbal SAT score reduces a student''s STEM course grade, on average, by 0.16 grade points.
But when we look at peer effects from course-company-mates, the results flip – for example, a 100-point increase in course-company-mates'' average verbal SAT score increases a humanities or social science course grade by 0.11 grade points.
Why the different results? In short the nature of peer interaction matters a lot. In broad social settings, students of like characteristics tend to interact more with each other, so higher ability students can drive average students away from them and worsen academic performance.
But in more task-oriented groups (such as classmates), students are more willing to overcome differences and work together. Policy-makers should take note – just throwing smarter students in the group can backfire and make overall performance worse.
Bad Company: Reconciling Negative Peer Effects in College Achievement – Ahmed Rahman