Using propensity score methodology to study the effects of ability grouping on mathematics achievement: A hierarchical modeling approach.

摘要

The present study examined the effects of ability grouping (curriculum tracking) on middle school students' mathematics achievement using a national longitudinal survey database (the Longitudinal Study of American Youth [LSAY]). It applied innovative analytic methods (a combination of propensity score matching and hierarchical modeling) that addressed the two most pressing methodological issues involved-nonequivalence of comparison groups and school differentiation in ability grouping. It presented a comprehensive picture of grouping effects by comparing not only students between different ability groups, but also grouped and ungrouped students, with these two types of results validating each other. This study also examined structural, curricular and other school-specific factors that are related to the effects of ability grouping. Furthermore, it explored the fairness of group placement and school-specific factors that impacted group placement.;The results on group placement provided some evidence that students were not always assigned to different ability groups solely based on their achievement and potential to learn. In addition to previous achievement, mother's education was positively related to placement in the high ability group. However, other non-academic factors that have been suggested to be related to group placement, such as ethnicity and SES (Duncan Socioeconomic Index), were not significant predictors of either high or low group placement, holding student achievement variables constant. The results on grouping effects indicated that while students in the middle ability group progressed at an identical pace with comparable ungrouped students in mathematics learning, students in the high ability group fared substantially better than their ungrouped and middle-grouped counterparts, and students in the low ability group lagged behind both their ungrouped and middle-grouped counterparts. Furthermore, taking pre-algebra gave students in the high ability group an extra advantage in addition to the positive high-versus-middle grouping effect, and the achievement gap between the middle and the low ability groups enhanced the negative low-versus-middle effect. The results also suggested that two-level grouping might be more equitable than three-level grouping, yet more efficient than non-grouping. Further investigation on the effects of two-level grouping versus non-grouping is warranted.