The institutional organization of the Italian educational system is based on strong assumptions about its equality purposes, among which a key role is assigned to the presumption that all schools provide similar educational standards. Commonly, families perceive that quality of different schools is quite homogenous, especially at lower levels, e.g. primary and junior secondary. Therefore, recent aggregate data provided by the Italian Institute for the Evaluation of Educational System (hereafter, Invalsi) show that it is not the case, and that a significant portion of variance in students’ test scores is attributable to between-schools differences.
This evidence about gaps in achievement across schools is accompanied by a specific feature of the Italian educational system, namely a strong difference in educational attainment and results in different geographical macro-areas and a debated difference between public and private schools’ performances.
A study of school effects on achievement for the different areas of the country seems worthy of specific attention, also because there are still few studies analysing specifically the determinants of Italian students’ achievement and performance. The most part of research efforts in this Project has been devoted to this aim, also with the idea of supporting more informed decision-making at system level.
Additionally, it is important to investigate the relationship between achievement and variables measuring students and schools’ characteristics, together with estimates of the relative weights of the different groups of covariates. Indeed, it is also likely that the various elements contribute differently to students’ achievement in the different groups. In both the contexts we deal with a huge amount of hierarchical structured data. So an effort in construct new statistical methodologies to use new waves of administrative data is of primary importance for assessing the effectiveness and efficiency of schools.
In this research project, we used a variety of methods for the purpose of assessing the ‘value added’ of schooling in the students’ educational purposes, ranging from multilevel linear models to the specification of bivariate models to take into account the multiple-output nature of the educational production processes. We also employed machine learning techniques, which include regression trees, random forests and boosting. Using a novel panel dataset, we also tested the stability over time of school-effects’ estimations, and we propose the integration of multiple methods to check the sensitivity of results to different specifications.
Administrative data employed in this research cover information about schools of various grades, from primary to secondary education, in Italy and in others European countries.
Within the HEAD project, people involved in the Education Research are: