Abstract
The purpose of this paper is to identify school effects on student performance for tertiary entrance in Australia, taking into account student-level predictors using longitudinal data from the 2003 Programme for International Student Assessment (PISA) study. It finds that aspects of schooling, such as positive attitudes to school and disciplinary climate, affect student performance at the student level but not generally at the school level. The socioeconomic context of schools has no effect on student performance when taking into account schools' academic context. Apart from academic context, teacher shortage, academic press, and teacher efficacy were the only school factors that had positive significant effects on student performance. The policy implications are that school-based policies are unlikely to improve performance or promote equity, but, instead, policies should focus more on students falling behind, who are found across the school system, not limited to a small proportion of schools with particular characteristics.
Notes
1. In the data analyzed for the present paper, often the school-level variables have intercorrelations above 0.5. For example, the correlations between the school-level PISA score with school-level Economic, Social, and Cultural status (ESCS), Academic Press, and Disciplinary Climate are 0.8, 0.6, and 0.5. Such correlations are much larger than the respective student-level correlations of 0.43, 0.25, and 0.14. This is because the aggregated school-level measures have, by definition, removed individual variation within schools.
2. Other countries include Canada and Denmark.
3. The PISA variables used for the attrition weights are family structure, the higher level of parents' education, country of birth, year level, intended occupational level, education program orientation, Indigenous background, sex, and home location.
4. According to Bond and Fox (2001, p. 259), the intercorrelations at the student level are 0.82 between mathematics and reading, 0.89 between science and reading, and 0.85 between mathematics and science.
5. The intercept term is the predicted value (in this case ENTER score) for cases that scored zero on each predictor variable.
6. The formula for BIC = −2l+dlog(n), where l is the likelihood ratio for the model; d is the dimensions of the analysis, and n is the number of cases (see Schwarz, 1978).
7. The bivariate correlation of PISA test score with ENTER score was about 0.6 compared to about 0.3 between ESCS and ENTER score.