- Evidence of stronger peer influences within reference groups than across reference groups, the magnitude of which varies substantially across the percentiles of the achievement distribution.
- Diversifying peer groups leads to small but fairly uniform improvements in the achievement gap across the achievement distribution.
- Whites appear to conform only to the behavior of their White peers, while there is evidence that nonwhites receive positive spillovers from both their White and non-White peers.
- At least within the same race, detracking classrooms would be more efficient for white students and slightly less efficient for non-Whites, while creating more equitable outcomes relative to tracking students into high and low-achieving classes.
- Desegregating classrooms does help to narrow the achievement gap, but only by .06 of a standard deviation relative to the observed assignment policy. White achievement does not appear to be affected much by desegregation on average, and the overall improvement is driven almost entirely by gains to nonwhite achievement. Therefore, random assignment is marginally more efficient and more equitable than the observed assignment policy.
- A higher percentage of nonwhites in the classroom is negatively correlated with achievement.
- Contemporaneous peer effects are very important to achievement and that in ignoring them, the literature to date has severely underestimated the impact of peers in education production.