The present study uses the method of propensity score matching and multilevel data to assess the causal effects of tracking on junior high students in Taiwan. Our research questions are: Would students who took part in high-achievement tracks have performed less well if they had not been tracked? Would students of low-achievement tracks have performed better if they had not been tracked? Who would benefit or suffer most from tracking?
With data gathered by Taiwan Education Panel Survey (TEPS) in 2001 and 2003, we focus on estimating the average treatment effect on the treated (ATT) by comparing separately the math performance of 9th of students who were assigned to high-achievement (N=1,216) or middle/low-achievement tracking (N=1,383) to students who attended schools with no tracking system (N=3,266).
After obtaining each student’s propensity scores of participation in either high-achievement tracks or low-achievement tracks, we further stratify samples into five strata based on their math ability scores in the 7th grade to explore differential treatment effects. The results of PSM analysis shows that tracking in general has positive effects on students of high-achievement tracks. The ATT of this track type is between 4 to 5 points. For students of middle/low tracks, the impact of tracking is inconclusive depending on the matching model. With the inclusion of school-level variables in the matching model, the result points to the direction that this track type would have a small negative effect on students’ math performance.
The findings further show that students with lower prior math ability would benefit more by being assigned to the high-achievement tracks. On the other hand, students with higher prior math ability would lose the most if they were assigned to the middle/low achievement tracks.