This award funds a research project that will develop new statistical methods to evaluate the effect of attending a specific public school on outcomes such as test scores, attendance, and graduation rates. The new method can be used to evaluate school outcomes for public school districts that use centralized assignment methods to assign students to schools. The PIs plan to employ the method to analyze data on schools located in Boston, Denver, and New Orleans. The results will shed new insight into how to best compare the results of different kinds of public schools, including comparisons between charter schools and traditional public schools. The results will also be useful to other school districts that are considering whether or not to use a centralized assignment method.
Centralized matching mechanisms have been used in a number of large public school districts for some time. Experience has shown that school priorities are often coarse and many students have identical preferences. Therefore the assignment mechanism used must include a tie-breaking feature. Some districts employ algorithms that use random assignment to break ties, while others use a regression discontinuity style running variable. In either case, the tie-breaking method makes quasi-experimental impact evaluation possible. The PIs will develop and demonstrate econometric methods that fully exploit the potential of these quasi-experimental designs for causal inference. They will develop theoretical results that show how to isolate randomness in market designs of increasing complexity, and will apply these methods to data from three large urban public school districts.
