Poster Paper: Identifying Magnet School Profiles in North Carolina: A Mixture Modeling Approach

Saturday, November 4, 2017
Regency Ballroom (Hyatt Regency Chicago)

*Names in bold indicate Presenter

Melinda Adnot and Savannah Williams, Davidson College


Magnet schools are an increasingly popular form of school choice, with nationwide enrollment doubling between 2000-01 and 2013-14 from 1.2 to 2.5 million students (Digest of Educational Statistics, 2015). These specialized schools are publicly funded and administered and have myriad goals which include fostering individual academic and artistic talents through thematic curricula, and reducing racial and socioeconomic isolation through a school assignment process other than traditional geographic boundaries (Magnet Schools of America, 2016). While it is well-recognized that magnet schools vary in their emphasis of specific goals (see, for instance Polikoff & Hardaway, 2017), these differences and their significance for students’ academic experience and outcomes are not addressed in educational policy research. Though important research has been done on the impact of magnets on racial and socioeconomic composition of schools (Bifulco, Ladd & Ross, 2008; Goldring & Smrekar, 2000; Saporito 2003) and student outcomes overall (Ballou 2009), most of the scant magnet schools literature is focused on single districts and concentrated on aggregate effects.

 This study employs unique longitudinal data from North Carolina to understand the heterogeneity in the magnet sector through the identification of different types, or “profiles” of magnet schools. For instance, we might imagine that schools with a strong thematic focus that are co-located in diverse schools may differ from magnet schools with low thematic emphasis and a separate building, and that these profiles have salience for the demographic composition of the school and the educational and social outcomes of their students. To identify these profiles, we employ latent profile analysis (LPA), an individual- (here, school-) centered latent variable measurement modelling technique that empirically identifies distinct subpopulations within a larger population (Vermunt & Magidson, 2002). LPA is not often seen in policy analysis but is well-suited to the task of identifying meaningful patterns in the clearly varied magnet school sector.

The importance of this work, in light of the demonstrated recent growth of magnet schools and federal interest in school choice policies, is clear. We contribute a first look at the types of magnet schools that exist in North Carolina, and suggestive evidence of the profiles of schools that foster the most diverse learning environments and where students are experience the most success.