| Standardized-test scores are increasingly important indicators of school success. But how robust are school-performance ratings when they are based on measures derived from these scores? In my thesis, using data from Houston Independent School District (HISD) and New York State (NYS), I examined the robustness of school-performance ratings across three methodological factors: (1) different achievement tests in the same academic subject; (2) different methods of transforming raw scores into scale scores (i.e., scaling methods); and (3) the phenomenon of students' scores being higher than their true achievement levels (i.e., score inflation).;I find that, in both the HISD and NYS datasets, school-performance ratings depend substantially on the test used. This applies to a variety of status and value-added measures with different model specifications. Further, in the HISD dataset, there is some evidence that the observed test effects were associated with differences in consequences for schools attached to results from different tests (i.e., stakes). Similarly, based on pilot data collected in NYS using two subtests designed specially to detect score inflation, the between-subtest inconsistencies in school ratings are consistent with the hypothesis that schools' ratings on NYS's high-stakes state tests are likely to reflect in part their relative amounts of inappropriate test preparation.;I also find that school ratings are less dependent on scaling methods than on subjects, grades, or years in the NYS dataset. However, there are usually substantive explanations for inconsistencies in schools' ratings associated with these latter factors. In contrast, it is particularly difficult to explain the dependency of the ratings on the scale used to stakeholders, especially schools whose ratings became worse with a switch in scale.;It is important that policymakers and researchers recognize these sources of variations in score-based school-performance measures, and adopt appropriate systems to prevent, detect, and correct them. This is especially when educators' inappropriate responses to high-stakes pressures could have distorted the initial construct the test was designed to measure. When they rate schools using scores that were indeed distorted by inappropriate responses, they risk incentivizing and propagating behaviors that run counter to the educational goals of accountability-based reforms and school-improvement efforts. |
