ABSTRACT
The objective of this study was to identify trajectories of school improvement experienced by Chilean elementary schools over the last decade. Using econometric analysis and controlling for potential confounding factors, we created an index of school performance combining outcome indicators focused on different school dimensions, and estimated the 2002/2010 evolution of school improvement for all Chilean schools with available data. Broadly, we estimated an average increase in the school performance of about 0.19 SD; nevertheless, while 41% of the schools increased their educational performance by at least 0.1 SD, 25% of the schools decreased it during the same period; we also found that improving school effectiveness was more probable among schools with lower student socioeconomic status (SES). Finally, we found that the observed school improvement trajectories tended to be non-linear; thus, we estimated that only 13.4% of the schools improved their performance in a systematic way during the entire decade.
Notes
1. SIMCE is a national test administered by the Ministry of Education, which evaluates academic achievement in language and mathematics. It is applied every year to 4th-grade students and alternately to 8th and 10th graders.
2. A critical challenge for conducting this kind of research in Chile and most of the countries is that there are no data available to estimate “value-added” models on students’ academic achievement, which is the recommended methodology to control for confounding factors (Harris & Bennett, Citation2001).
3. There are no reliable and longitudinal data available on other relevant dimensions of students’ learning objectives in Chile, like interpersonal and intrapersonal skills.
4. Since we were interested in measuring final intended school goals, we restricted our performance indicators to measures of students’ attainment (i.e., excluding school intermedia objectives like increasing school resources or teachers’ capacities).
5. In cases where information was missing for one of the two considered periods, the one with information was used as the moving average; this imputation method allowed us to increase the number of observations without incorporating a large amount of bias to the analysis carried out. We replicated the estimates by using each year separately, and the results were broadly consistent with those reported in the paper.
6. We included those school-level variables to capture the compositional (or “contextual”) effects of students’ family characteristics, which are not fully captured for individual-level variables. This kind of compositional effects is especially relevant in highly segregated school systems like Chile. Those variables may be interpreted as indirect measures of peer effects.
7. Tests are available upon request.
8. We also estimated two alternative models, one without the estimated school effect and a second one not using the standardized scores from the SIMCE test and its proficiency levels. The selection was made based on the Kaiser-Meyer-Olkin (KMO) index.
9. Alternatively, we implemented a statistical technique of non-hierarchical clusters of K-median to define the number of clusters endogenously (i.e., taking into account the observed distribution of the data). This method led to the generation of up to 17 different groups. Nevertheless, since clusters varied significantly in the size of ISP range (some extremely limited and some very large) and they were not consistent over time, the substantive explanation of the findings became diffuse, especially in making intertemporal comparisons, which was the main purpose of defining ISP clusters.
10. An alternative approach used in the literature would be to compare school’s changes in a ranking order; nevertheless, this alternative indicator would only capture changes in the relative performance of schools compared to the rest of the schools. Our ISP allowed us to estimate absolute changes relative to the own initial school performance in the baseline year.
11. SES groups are defined by the Ministry of Education based on the average education of the mothers and fathers of students at the school and the per capita income of the families of the students. The schools are classified into five categories: Low, Middle-Low, Middle, Middle-High, and High SES.
12. We considered 0.4 SD as the parameter to identify relevant improvement at the school level during the last decade based on the observed national average increase in language academic achievement (measured by SIMCE) in the same period; this value is about two times the national average increase in the ISP.
Additional information
Funding
Notes on contributors
Juan Pablo Valenzuela
Juan Pablo Valenzuela is an associate researcher of the Center for Advanced Research in Education and associate professor in the Economics Department, both at the University of Chile. His main research areas are economics of education and social inequality.
Cristián Bellei
Cristián Bellei is an associate researcher of the Center for Advanced Research in Education and assistant professor in the Sociology Department, both at the University of Chile. His main research areas are educational policy, school effectiveness, and school improvement; he has published extensively about quality and equity in Chilean education.
Claudio Allende
Claudio Allende is an economist and a young researcher of the Center for Advanced Research in Education at the Universidad de Chile. His main research areas are economics of education and violence and security.