Findings from Nonexperiemntal Analyses

Although the average effect sizes (comparing achievement of students of AC teachers to achievement of students of their TC counterparts) were not statistically different from zero, effect sizes varied across individual pairs of AC and TC teachers. In reading, the effect size was less than zero in half the pairs and greater than zero in the other half. For math, the effect was less than zero in 56 percent of the pairs and greater than zero in 44 percent. Separating the effects of characteristics of teachers from the influences of their training, however, requires nonexperimental analysis, as does examining the relationship between teacher characteristics and classroom practices and student achievement.

To estimate the relationship between teacher characteristics and training experiences and student achievement, we used ordinary least squares (OLS) regression equations to estimate the correlation between a student's posttest score and student-level characteristics (including pretest score), whether his or her teacher was from an AC program, differences between the characteristics of AC and TC teacher pair within a school and grade, and other unobservable effects. This model allows us to estimate the relationship between differences in student achievement and differences in AC teachers and their TC counterparts' characteristics, such as required coursework, whether a teacher is currently taking courses, undergraduate major, and SAT scores.

All together, the differences in AC teachers' characteristics and training experiences explained about 5 percent of the variation in effects on math test scores and less than 1 percent of the variation in effects on reading test scores.

Differences in teachers' demographic characteristics and coursework required for initial certification were not related to the effects of teachers on student achievement. Of the several aspects of teachers' education and training we examined, two were statistically significantly related to the effects of teachers on student achievement, and both relationships were negative. First, AC teachers with master's degrees were less effective in improving student achievement in reading than their TC counterparts without a master's degree (effect size was –0.12). Second, students of AC teachers who were taking coursework toward certification or a degree scored lower in reading (effect size –0.13) than did students of their TC counterparts who were not taking coursework.