- Cross-sectional methods reveal a substantial negative association between exposures to high-poverty classrooms and test scores; this association grows with grade level, becoming especially large for middle school students. Over time, cumulative effects produce even larger differences between students exposed to higher versus lower poverty classrooms, growing from a -0.082 standardized effect size to -0.429.
- Growth models, however, produce much smaller effects. Including grade level, student characteristics, and interactions reduces the effect of classroom poverty on reading and math achievement.
- Even smaller effects emerge from student fixed-effects models that control for fixed baseline unobservables such as innate ability, early childhood experiences, and mother’s IQ. The effect of high-poverty classrooms on math is not statistically significant and the effect on reading is significant only at p < 0.05. Both math and reading effect sizes have been reduced to less than 0.01 standard deviations. Cumulative effects and continuous poverty effects were also smaller in these models.