This presentation describes a prediction model of the inner workings of a small high-equity education system in Europe, Iceland. Here high equity referes to low school variance in student achievement and little influence of student background such as SES or parent education. Results are presented from the second year of a 3-year research project that evaluates school-level trends in student achievement and the explanatory power of contextual factors at student and school level. Linear and non-linear trends are considered for both short- and longterm. A unique feature of this analysis is that it is census based, not sample based, and covers all schools in the system. A decade of near perfect system-wide records are available of all schools in Iceland for both achievement and contextual indicators from four cycles of PISA and the TALIS survey 2007. Methodological and compatibility issues in international surveys are considered.
Achievement is affected by an interplay of factors relating to the student, teacher and school and multilevel and multivariate explanatory models can be used to disentangle relative importance of these interrelated factors for preformance by controlling for and conditioning effects. Even in a Nordic low-variance education system like Iceland, factors influencing students‘ academic performance in compulsory education interact on student, teacher and school level. At school level these include school resources, school size and teacher training policy, at teacher level for example disciplinary climate, classroom practices and teacher-student relations, and at student level external factors like socio-economic status, behavioral factors like effort and perserverence and internal factors like enjoyment of reading.
The theoretical framework for this analysis is the field of Educational Effectiveness. In the past decades education researchers have developed analysis for explaining academic peformance in context of background indicators identified in parametric statistical designs (see Teddlie and Reynolds, 2000 and McInerney et.al. 2006). The context is seen as multilevel, with researchers identifying the most important factors at each level, for example at student, class, school and system level. This presentation will focus on specific factors, drawing on work by Creemers, Kyriakides et.al. (1994, 2000, 2005 and 2010) in several sample based studies in the Netherlands and Cyprus.
This kind of intensive system wide analysis is rare on a European scale in terms population coverage and extensive time period, virtually without sampling error. With large scale longitudinal datasets interrelationships of contextual factors and relative explanatory value for performance variance can be modelled and critically evaluated in terms of causal interaction to provide contextual results and realistic grounds for policy and improvement programs.
