| Reallocation of instructional resources is among common responses to educational accountability under NCLB. To "game" the system, schools may reallocate resources among subjects, skills, formats, and students. In this thesis, I examined state test results to evaluate whether score patterns were consistent with reallocation to focus on standards heavily featured on state tests, open-response items, or students near the threshold of a performance level (i.e., "bubble students").;In the first study, I explored whether gains were larger on standards heavily sampled on the Massachusetts grade 10 English language arts (ELA) test, overall and in disadvantaged schools and schools facing strong accountability pressure. Using five years of data, I found no statistically significant evidence that gains on standards heavily sampled were larger than expected for schools with relatively large gains on these standards, given the schools' gains on standards lightly sampled. In contrast, disadvantaged schools and schools under accountability pressure showed differentially large gains on heavily sampled standards.;In the second study, I examined whether gains on open-response (OR) items were larger than gains on multiple-choice items and writing on the Massachusetts grade 10 ELA test over the same five years, overall and in disadvantaged schools. I found no statistically significant evidence that OR gains were larger than expected for schools with relatively large OR gains, given these schools' gains on multiple-choice items or prompt writing. In addition, contrary to what might have been expected, large percentages of low-income and non-Asian minority students were associated with differentially small gains on OR (and differentially large gains on English conventions in prompt writing).;In the third study, I analyzed student-level data on the New York state grades 3-8 ELA tests and found that the differentially large gains for bubble students were larger where incentives based on cut scores were stronger. Moreover, the New York City Progress Report, an accountability system based partially on achievement and partially on growth, did not mitigate the differential gains for bubble students at the basic level but did at the proficient level two years after the implementation of the system. |
