80 likes | 189 Views
Choices for Studying Choice Dev Davis Macke Raymond. Study Approach. One large multi-location dataset 14 states plus two large districts Student-level results 2004-2005 through 2007-2008 Two methods Student fixed effects Virtual control records Two comparisons
E N D
Study Approach • One large multi-location dataset • 14 states plus two large districts • Student-level results • 2004-2005 through 2007-2008 • Two methods • Student fixed effects • Virtual control records • Two comparisons • Head-to-head comparison of the methods • Policy analysis
Virtual Control Record Process Charter School Student Feeder School(s) Students Provide List of Potential Match Schools • MATCHING VARIABLES: • Race/ethnicity • Gender • English proficiency • Lunch status • Special education status • Grade level Find Matches Based on Demographic Variables Eliminate Matches Who Attend Charter Schools • MATCHING VARIABLE: • Test scores from t0 Match Test Scores Virtual Control Records Match rate is 86%
Head-to-Head Comparison * Significant at p ≤ 0.05 ** Significant at p ≤ 0.01
Policy AnalysisNational Charter Performance * Significant at p ≤ 0.05 ** Significant at p ≤ 0.01
Summary of Findings • Charter school results vary by student subgroup. • Positive for low income, English learners • Other groups mixed or negative • Charter school results vary by location. • Positive for 6 locations in math & reading • Negative for 4 locations in math & reading • Charter school results vary by school. • Positive for 19% of schools in math; 19% in reading • Negative for 33% of schools in math; 21% in reading
Policy Implications • Closure rate comparison • Small businesses: 50 percent • Charter schools: 15 percent • Accountability side of flexibility-for-accountability equation is important for quality. • Authorizing & renewal • Publicly available information & comparisons • Other possibilities • School age • Affiliation with charter networks • State charter school policies