Differential effects of poverty on Hawai'i students' achievement in reading and mathematics

The Coleman Report of 1966, Equality of Educational Opportunity , provided a definitive picture of poverty's negative effect on educational outcomes. While numerous other large-scale studies have since further clarified our understanding of poverty's effect on achievement; there has been very little inquiry into poverty's differential effects in conjunction with demographic factors such as ethnicity and gender. This study examined the differential effects of poverty on achievement across ethnic and gender groups. The 2002 Hawaii State Assessment results for reading and mathematics served as the achievement outcomes. Third, fifth, eighth, and tenth grade Hawaii public school students (N = 42,704) participated in the statewide assessment. Differential effects were examined in three stages: (1) Student-level only, (2) school-level only, and (3) multi-level (student and school levels) achievement. The last stage examined Native Hawaiian and Part-Hawaiian students only, the most disadvantaged group in Hawaii's public education system. Multivariate analysis of variance, multivariate regression, and multi-level modeling were employed to investigate the effects of poverty, ethnicity and gender, and the interactions between these variables.;The student-level analysis was the only one to support a general model that included an interaction effect across grade levels and subjects. The student-level model comprised of the main effects of poverty, ethnicity, and gender, along with a poverty x ethnicity interaction. This multivariate model accounted for between 14% and 20% of the generalized multivariate variance. Further analyses revealed East Asian and Native Hawaiian students were more severely affected by poverty than other ethnic subgroups. Differential effects were most evident at lower grade levels for both reading and mathematics.;Results from the school-level analysis did not support the four effects model identified from the student-level data. Instead, a single-predictor model consisting of the main effect of poverty was found to be significant across the four grade levels. This single-predictor model accounted for between 35% and 56% of the generalized multivariate variance. Separate univariate analyses for reading and mathematics further supported the general multivariate results.;The multi-level analyses examined the effect of poverty at the student and school levels. The effect of poverty was found to be significant at both the student and school level; however, no consistent cross-level interaction was observed. Unlike poverty, gender was found to have no consistent significant effect at either the individual or contextual level. The overall proportion of variance accounted for by the general multi-level model ranged between 9% and 21%.;The poverty x ethnicity interaction identified at the student level is further discussed in the context of improved measures of differential effects. In addition, policy implications pertaining to the No Child Left Behind Act, educational accountability, and pedagogy are discussed.