This research empirically investigates the size and nature of peer effects in North Carolina public schools and how parents choose schools (and thus how peer groups within schools endogenously form) both through their choices of residential location and their choices of school type (private or public). A new and unique dataset from the North Carolina Education Research (NCER) Data Center contains detailed information on all North Carolina public school students in grades 3 through 8 over a five-year period. Students are linked to teachers and this allows for a variety of definitions of peer groups, including at the classroom rather than the grade or school level. Student addresses, available for all public school students, will be linked to addresses in privately available mortgage/renter datasets that include a wealth of household information not typically available in education datasets of this kind. This data will be the basis for the research.
In addition to permitting researchers to use more extensive household controls, this dataset permits the separate identification of students who (i) remain within North Carolina public schools over the sample period regardless of whether they move and switch schools, (ii) switch between public and private schools, and (iii) enter or exit the state and thus enter or drop out of the school dataset. The NCER dataset is currently available and can be used in pursuit of the first of the identification of peer effects and the evaluation of the precise nature of peer effects within schools. Merging the private mortgage/renter data for the Raleigh-Durham metropolitan area with the NCER data permits the researchers to investigate the second objective of the proposal. Policy experiments and frequent school neighborhood boundary changes within the Raleigh Durham area will be used to identify determinants of household choices over residential location and school type. This lays the groundwork for a more ambitious effort to merge similar mortgage/renter data with the entire NCER dataset in order to create a single dataset with extensive information on all students, teachers, schools, peers and households within North Carolina. This would be the only dataset of its kind.
The research results will be useful in several ways: Because the research allows the testing the importance of different types of peer groups, it is likely to make substantial advances in identifying the causal nature of peer effects. Second, the research will yield insights for policymakers whose efforts are often frustrated by household choices that respond to policy changes. Finally, the project lays the groundwork for the construction of a very comprehensive education dataset that combines detailed information on students, households, peers and teachers.
