Rewarding Schools Based on Gains: It's All in How You Calculate the Index and Set the Target

摘要

The purpose of this paper is to resent some empirical evidence about how the allocation of rewards depends on the methods used to construct the accountability formulas. As far as we know, there is little published research on this topic, and states have limited information on which to base choices. In particular, changes in three features of California's accountability system: the way student test scores are combined into a school index, the way targets are set, and whether the system has additional targets for particular subgroups of students. In the first case, we compare an index defined in terms of a linear metric mean National Curve Equivalent gains to one that is more 'progressive' (i.e., one that allocates more points for gains among low-scoring students than for gains among high-scoring students). In the second case, we compare three methods for defining gain targets a fixed amount of improvement, a fixed percentage improvement, and a percentage of distance to target. Finally, we examine the effects of imposing similar conditions on subgroups of students as well as the school as a whole. We use 1998-99 and 1999-2000 data from California and from the Los Angeles Unified School District as the basis for our analyses. This study suggests that small differences in computational formulas can have large effects in terms of consequences for schools.