Monday, Nov 3
Melvin Stephens, Estimating Program Benefits
a PSC Research Project [ARCHIVE DISPLAY]
As policymakers search for innovative approaches to improve public education in the U.S., charter schools have become increasingly popular. Part of the broader movement toward parental choice and school autonomy that includes magnet schools and private school vouchers, charters are the fastest growing avenue for school choice. In 2009, 4,963 charter schools served over 1.6 million students . The prominence of charter schools in current education reform is partially attributable to President Obama?s commitment to charter schools. The federal government?s Race to the Top competition, which awarded over $4 billion to states, gave extra qualification points to states ?ensuring successful conditions for high-performing charters and other innovative schools.?
Given the growth of charter schools, there has been increasing interest in assessing their effectiveness. A number of studies have compared achievement among students at charter and traditional schools, with mixed results. However, selection bias in an overwhelming problem in this research. A recent strand of research overcomes selection bias by exploiting the randomization in charter school admissions via lotteries. Lottery studies of charter schools in high-poverty, urban areas (NY, Boston, Chicago) have estimated extremely large, positive achievement effects. A national study by Mathematica concurred with these urban results, but also uncovered zero to negative effects in low-poverty, suburban areas.
Understanding the settings, practices and populations that produce positive charter effects is the central goal of our proposed study. To link charter practices to charter effects, we need to first estimate causal effects for a large and heterogeneous set of charter schools. Previous studies, limited to single cities, have been hampered in this effort. By contrast, Michigan is an ideal setting to undertake this research agenda, with 242 charter schools enrolling more than 110,000 students in urban, rural and suburban areas.
We propose to conduct a rigorous, Michigan-wide analysis of the effect of charter schools on student performance and postsecondary schooling decisions, including college entry, choice and completion. Our research will exploit the randomized lotteries conducted by over-subscribed schools as a source of identifying variation in charter school attendance. We will leverage the fact that, as part of an ongoing IES-funded project, we are collecting student-level, longitudinal data for all charter and traditional public school students in Michigan, which will include student achievement scores, high school graduation and college enrollment. The study will not only estimate the overall effect of charter schools but also will investigate heterogeneity in the effect by student type, age of the charter school, charter practices and type of charter.
This project will serve as a model for researchers and policymakers of a cost-effective approach to rigorously evaluating the effectiveness of charter schools. One of the greatest challenges in conducting education research is the difficulty (both ethical and practical) of randomly assigning students, teachers or schools to interventions. The fact that over-subscribed charter schools are legally required to admit students on the basis of random lotteries offers the opportunity to conduct rigorous analysis on a scale rarely seen in education research. By the standards of randomized trials, the cost of conducting research using the lotteries is quite low: states already maintain the longitudinal datasets that contain the outcomes
of interest and random assignment has already occurred. All that is required is that the lottery data be collected and matched to the outcome data. Our hope is that the lessons from this project will persuade state education agencies to provide the institutional resources and incentives (or even mandates) to allow academic researchers or charter autho
|Funding:||Smith Richardson Foundation, Inc.|
Funding Period: 03/01/2011 to 07/01/2015
This PSC Archive record is displayed for historical reference.