Multilevel Analysis of Programme for International Student Assessment (PISA) 2018 in Indonesia

One goal of education is to produce competent and competitive human beings to realize the nation’s welfare. An assessment is essential to determine whether educational goals are achieved. Effective evaluation needs to be carried out sustainably to obtain overview information on student achievement through a global assessment called the Program for International Student Assessment (PISA) survey. In PISA, the student data nested in schools, resulting in the observed responses being taken tending to be not independent and having variations between schools that are not homogeneous. If modeling is done using linear regression analysis, it can lead to deviations in assumptions, and information related to groups will be neglected. Thus, multilevel modeling is needed to overcome this drawback. This study analyzes the 2018 PISA data in Indonesia, which consists of 3 test domains (Mathematics, Reading, and Science) using multilevel regression analysis. Modeling was carried out using the same predictor variables at the student and school levels with various modeling scenarios for each domain. In PISA, educational achievement is measured through the plausible value so that the modeling will be related to the formulated plausible value. In line with one of the goals of PISA, to encourage the equality of education, several variables related to educational equity are selected. The empirical results show that the factors that affect student achievement in three domains (Reading, Mathematics, and Science) are ESCS (index of economic, social, and culture status), grade repetition, mean ESCS in school, and type of school for significant level 5%. Not all categories of language at home, immigration status, and grade significantly affect student achievement, for the science domain, all category of immigration status has a significant effect. For the mathematics and science domain, the gender factor has no significant effect, but the interaction of ESCS and mean ESCS has a significant effect.