Abstract
Understanding which features of classroom instruction are associated with mathematics learning is a central question for education systems worldwide, yet distinguishing instructional associations from differences in family background and prior schooling remains challenging. This study examines the relationship between instructional quality and mathematics achievement using an international assessment framework. Colombian adolescents are among the lowest performers in the OECD mathematics assessment, and substantial variation in achievement occurs across schools, making Colombia a relevant context for examining these relationships. Drawing on PISA 2022, we analyzed 7,367 students nested within 260 schools using a Bayesian two-level model with a random school intercept. Three instructional constructs aligned with the ICALT framework were examined at the school level: classroom orderliness, perceived teacher support, and two facets of cognitive activation. Socioeconomic status, sex, grade of enrolment, and grade repetition were included at the student level. Approximately one-third of the variance in mathematics achievement occurred between schools before predictors were included (ICC = 0.36), declining to approximately one-quarter after their inclusion. Classroom orderliness and reasoning-oriented cognitive activation were positively associated with mathematics achievement, whereas perceived teacher support and thinking-oriented cognitive activation showed negative conditional associations when the other instructional indices were held constant. At the student level, socioeconomic advantage, male sex, higher grade placement, and not having repeated a grade were associated higher achievement. The findings indicate that orderly classroom environments and reasoning-oriented instruction are positively associated with mathematics achievement, while the negative conditional associations highlight the importance of considering overlap among correlated, perception-based instructional measures when interpreting school-level estimates.
Funding Information
  • No funding

This publication has 26 references indexed in Scilit: