Abstract:

This report attempts to explain the conceptual and technical base when linking two international large-scale assessment studies set up by the Organization for Economic Co-operation and Development (OECD): the Teaching and Learning International Survey 2013 (TALIS) and the Programme for International Student Assessment 2012 (PISA).

The linking procedure facilitates the investigation of mathematics performance at the student, teacher and school level and shed light on the effectiveness and efficiency of school systems when it comes to mathematics education. Data from eight national contexts: Australia, Finland, Latvia, Mexico, Portugal, Romania, Singapore and Spain are used to link TALIS and PISA.

International large-scale assessments have the potential to boost multi-level research that fits state-of-the-art educational effectiveness models. But - as tackled in the present technical paper - this potential is often flawed by the methodological difference in the study design. The present technical report presented solutions, procedures, strategies to develop overarching databases that link datasets from earlier studies; more specifically PISA 2012 and TALIS 2013. The report analyzed characteristics about the instruments, the variable frameworks, and especially the specific sampling strategies. The report emphasized four critical aspects in the context of merging datasets:
- Looking for a shared student variables value: TALIS-PISA Link 2013 and PISA 2012 can be linked by looking at data from 15-year-old students and their mathematics teachers.
- Looking for a shared anchor variable to link the data sets: the link between the TALIS and PISA data set depends on a shared school ID. This implies that - in view of the analyses - data should be aggregated at the school level.
- A time gap was observed in the administration of the two performance indicator studies; that again differed on the hemisphere where the study was set up: teacher work experience was used to look for a valid choice to sample teachers that were actually teaching to the students in the PISA sample.
- Not all teachers in both studies were relevant to be linked to students’ PISA mathematics performance: a five-step sample selection procedure was adopted to identify the relevant mathematics teachers whose data can be used in the merged database to be linked to student achievement.

Taking into account four critical issues and applying five selection criteria, 3473 teachers, teaching to 31548 students in 1115 schools have been selected.
At a technical level, further checks were needed to screen the original dataset and to apply corrections. The newly available dataset forms a promising basis to study theoretical models about the relationship between student, teacher and school-level variables.The report can potentially stimulate future research studies endeavours to advance the understanding of multilevel perspectives of PISA students’ mathematics learning outcome in various national contexts building on the educational effectiveness dynamic model.

Keywords:

TALIS and PISA, mathematics achievement, educational effectiveness, school climate, multi-level perspective.