We exploit a regression-discontinuity design to estimate peer effects on college students' attendance, using data from a classroom experiment, which required students who scored below a cut-off on the first midterm exam to attend subsequent classes. Since within a small interval around the cut-off, which side of the cut-off a student falls is randomly determined, so is the proportion of a student's classmates falling on one side of the cut-off in the same interval. Using this proportion to instrument peer attendance, we find that a one-point (out of ten) increase in classmates' average attendance score raises a student's attendance score by 0.7 points.
|Original language||English (US)|
|Number of pages||10|
|State||Published - Jan 1 2015|