ABSTRACT
In the 2009 Programme for International Student Assessment, the Flemish community of Belgium outscored its French community in reading, with low achievers accounting for a large proportion of the score gaps. In this study, between-community comparisons based on the Blinder-Oaxaca decomposition method showed that the Flemish community benefits largely from its policies and practices of giving schools relative autonomy in resource allocation, promoting good student conduct, and decreasing grade retention. Between-community comparisons using the Juhn-Murphy-Pierce decomposition method showed that compared to their Flemish peers in the same percentile, low achievers in the French community are hindered more by their relatively low grade level; disadvantageous school economic, social, and cultural composition; and negative school supportive climate while medium and high achievers benefit more from their advantageous school economic, social, and cultural composition; positive supportive climate; and high proportion of qualified teachers in schools.
Acknowledgements
The authors would like to thank Christian Monseur and Ariane Baye for their help in preparing the data analysis.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Notes
1. In this paper, when we talk about regional differences in Belgium, we refer to the community differences. The Flemish community (58% of Belgium’s population) includes the Dutch-speaking schools in Brussels. The French community (41% of Belgium’s population) includes the French-speaking schools in Brussels. We do not consider the small German community. For a full overview of the educational system in Belgium, we refer to Geyer (Citation2009).
2. In multilevel linear analysis, a basic assumption is that the dependent and independent variables are multi-normally distributed. Due to the deviation of the observed from the theoretical distribution, the estimated uncorrected/corrected mean scores of reading achievement in these models might deviate from the observed ones.
3. Due to the use of five plausible values for the dependent variables, SAS 9.3 did not calculate the standard error for each residual estimate in the multilevel linear models. Therefore, the significant levels for the percentile score gaps could not be calculated.
Additional information
Funding
Notes on contributors
Bo Ning
Bo Ning is a researcher in the Research Institute for International and Comparative Education, Shanghai Normal University, China. His research focuses on cross-country comparison in school effectiveness. He was a doctoral student in the Centre for Educational Effectiveness and Evaluation, KU Leuven. This article is based on his doctoral research in KU Leuven.
Jan Van Damme
Jan Van Damme is a professor in the Centre for Educational Effectiveness and Evaluation, KU Leuven, Belgium. His major research interest is school effectiveness.
Wim Van Den Noortgate
Wim Van den Noortgate is an associate professor in the Methodology of Educational Sciences Research Group, KU Leuven. His major research interest is research methodology.
Sarah Gielen
Sarah Gielen is an assistant professor in the Centre for Educational Effectiveness and Evaluation, KU Leuven. Her major research interest is comparative education.
Kim Bellens
Kim Bellens is a doctoral student in the Centre for Educational Effectiveness and Evaluation, KU Leuven. Her major research interest is comparative education.
Vincent Dupriez
Vincent Dupriez is a professor in the Interdisciplinary Research Group in Socialisation, Education and Training, Université Catholique de Louvain. His research interests are comparative education, educational policies, and school administration.
Xavier Dumay
Xavier Dumay is a professor in the Interdisciplinary Research Group in Socialisation, Education and Training, Université Catholique de Louvain. His research interests lie in educational policies, educational administration, and research methodology.