| Albert Beaton
Boston College
Examining changes in international multilevel variance and student correlates of mathematics achievement using data from TIMSS 1995 and TIMSS 1999.
FINAL REPORT:
Over the past four decades, spurred in part by global economic competition, researchers and policy makers in the United States have expressed an increased level of interest in international comparative studies. Publications such as A Nation at Risk (1983), A Nation Prepared (1985), Workforce 2000 (1987) and the Goals 2000: Educate America Act (1994) have heightened our awareness of the need to prepare our students to compete in a global economy. For the past 8 years, the Third International Mathematics and Science Studies (TIMSS 95 and TIMSS 99) have provided data about how students around the world compare on a common test of mathematics and science achievement.
Given the size, scope, and methodological consistency of TIMSS studies, the purpose of this research is to examine how nations around the world organize their formal education systems and how the United States' system compares, to examine whether the structure of the system is associated in any systematic way with student achievement, and to explore the diversity among the countries that participated in both TIMSS administrations in terms of the predictors that are associated with higher mathematics achievement. Eighth Grade mathematics achievement in 23 countries across two TIMSS administrations was examined.
Using multilevel regression techniques, this research shows that countries differ both in terms of the total variability in performance and in the way in that the variability is observed to exist between students within classrooms, among classrooms/schools. The structure of the formal education systems in TIMSS varies considerably among countries and as a result, students from around the world receive instruction in a variety of environments. When compared to other countries, this research shows that United States is certainly unusual in the way it differentiates instruction for its students. Students in the United States receive instruction in some of the most homogenous environments, and there is considerable variation between classrooms within the same school, and among schools. Unlike some European countries which operate technical and vocational tracks parallel to their academic tracks, here in the United States, although residential segregation may contribute to school-to-school differences, students are likely to be tracked within schools.
Conditional multilevel modeling shows that students' attitudes toward mathematics and students' SES seem to be the most consistent predictors of achievement both within and across administrations. School average SES is consistently predictive of school-level achievement in the majority of countries and in both administrations. The percent of variance explained by the multilevel models demonstrates that in all countries, the models are more powerful for predicting differences between classrooms/schools than for predicting differences within classrooms.
Back to Funded Research Grants Page |