Abstract
The more science and mathematics subjects that pupils in pre‐university education include in their final examination package, the more future academic routes are available to them. Equality of educational opportunity is thus threatened when groups of pupils, distinguished by sex and family background but otherwise of equal capacities and achievement, are found to differ in their choices. This proposition is examined using data from a large Dutch cohort. Multilevel analyses show that the choice of science and mathematics subjects by girls is influenced by their family background while the choice by boys is not. The influence of various pupil and family variables on the subject selection process is explored via path analyses. The results confirm the importance of viewing subject choice as a chronological process that progresses differently for boys and girls.
Notes
1. Along a scale of 2–6, the mothers of the selected pupils scored 4.03 (SD = 1.035) and the mothers of the non‐selected pupils scored 4.16 (SD = 1.087).
2. The variables at school level were omitted from further consideration because the software program Mplus (Muthén & Muthén, Citation2001) did not allow the hierarchical structure of a dataset to be taken into account in connection with a categorical outcome variable.
3. The measures of model fit were the root mean square error of approximation (RMSEA), the weighted root mean square residual (WRMR), and the p value for the chi‐square test of model fit. The fit was found to be reasonably good with RMSEA < 0.06, WRMR < 0.90, and p > 0.05 (Muthén & Muthén, Citation2001, appendix 5).
4. A grading committee involves all of the teachers for a particular group of students, and meets to discuss and determine the grades for each student.
5. In Mplus, it is possible to specify and estimate (non‐directional) correlational connections between endogenous variables. All of the correlations specified by us—whether between exogenous or endogenous variables—are represented with a straight bi‐directional arrow.
6. Given that an ordinal dependent variable is involved here, it is not possible to perform a so‐called “multi‐sample” test to directly determine whether the models for the boys and the girls differ from each other. For this reason, a different course of action was undertaken and the model for the boys was simply applied to the data for the girls, and vice versa.