Burke, Mary, & Sass, Tim
Classroom Peer Effects and Student Achievement
Federal Reserve Bank; Florida State University
Compare the strength of peer effects across different groupings of peers, and compare effects of fixed vs. time-varying peer characteristics.
Journal Name or Institutional Affiliation:
- Some grouping ability may create Pareto improvements over uniformly mixed classrooms.
- Contemporaneous behaviors wields stronger influence than peers' fixed characteristics.
- Obtain very few significant effects of the gender and racial composition variables. But are reluctant to make definitive pronouncements on race and gender effects based on findings.
- Unobserved teacher inputs induce correlated effects on student achievement gains.
- Peer effects operate differently for reading achievement than for math. Peer performance generally has stronger effects on individual achievement in math than in reading, particularly at the elementary school levels.
- Mean exogenous characteristics as well as current average peer performance are found to have significant effects on individual student achievement in a number of circumstances, though results are not very robust.
- Find large and highly significant effects of mean peer achievement gains, at all schooling levels and for both math and reading, among students in the lowest quintile.
- Even in cases where the highest-ranked students experience negative spillovers, as in elementary and high school math, these effects of shifting some high-performance peers to low-performing classrooms not be zero-sum.
- Estimation of peer effects may be sensitive to the inclusion of teacher controls even if individual fixed effects are already included.
- Peer effects are not "one-size-fits-all", differences in mathematics and reading achievement across schooling levels, depending on whether we allow spillovers from contemporaneous peer outcomes or only from predetermined peer traits, and depending on the ability of the individual student.
- Strong effects from contemporaneous behaviors that cannot be predicted on the basis of fixed peer factors.
- Some externalities on achievement from peer disruption, although they may be either positive or negative.
- Authors are reluctant to make definitive pronouncements on race and gender based on their current findings.
Academic Achievement, Classroom Composition, Math, Peer Effects, Reading
Secondary Survey Data
Method of Analysis:
Fixed Effects Regression Models
Florida Public Schools
Unit of Analysis:
- School administrative records from Florida.
- Data covers five school years, 1999-2000 through 2003-2004
- Test available for both math and reading in each of grades 3-10.
- Randomly selected samples of 100 elementary, 100 middle and 100 high schools in Florida.
- Eliminate charter schools.
- Include "endogenous" (student achievement gains and number of discipline referrals) and exogenous peer characteristics (race and gender).
- Instrumental variables for peer achievement : the proportion of classroom peers that moved to the school during the current year, the mean age of the current peer group (expressed in months), and the percentage of classroom peers with disabilities.
- DV: Mean classroom peer math and reading achievement gains in Florida 1999/2000-2003/2004
- IV: Fraction of peers who are female, fraction of peers who are Black, fraction of peers who changed schools, mean age of peers (in months), fraction of peers who are special ed.students, class size, "structural" mover, "non-structural mover"