Estimating the causal effect of class size on the achievement of Moroccan students: Evidence from TIMSS 2023

Background .: Administrative caps on class size are the main source of quasi-experimental evidence on class-size effects, but whether a cap generates usable variation depends on the institutional setting. In Morocco, lower secondary classes average 36.5 pupils and a quarter of pupils sit in classes above the legal maximum of 40, yet no study has used the national cap to estimate the causal effect of class size, and whether the cap can identify it is unknown. Data and methods .: We use the TIMSS 2023 Grade 8 sample for Morocco (8,032 pupils, 240 classes, 229 schools) and five subjects: mathematics, physics, biology, chemistry and earth science. Teacher-reported class size is instrumented by the class size predicted from the maximum of 40 pupils and Grade 8 enrollment. A pupil-level design (weighted two-stage least squares with covariates and enrollment controls, and fuzzy regression discontinuity) is combined with a class-level design based on class mean scores. Inference uses the plausible values, school-clustered standard errors and Anderson--Rubin confidence sets robust to weak instruments. Results .: Least squares yields a negative association in every subject, significant in mathematics (-1.76 points per pupil, SE=0.47). In the pupil-level design, the first-stage F statistic falls from 87.4 to 3.9 under a cubic enrollment control, the Anderson--Rubin set is unbounded, and no discontinuity estimate is significant. In the class-level design, the Anderson--Rubin sets exclude zero under linear and piecewise-linear enrollment controls (-4.33 points per pupil in mathematics, set [-9.8,-0.5]) but not under a cubic control, and only 12 classes lie near the first three cut-offs. The evidence leans toward a negative effect of class size, but its magnitude is not identified: with a high cap and large cohorts, the rule generates too little local variation. JEL Classification: I21 , I28 , C26 , C21

Authors

Institutions

Publication Details

Journal
Research Square
Published
2026-11-18
DOI
https://doi.org/10.21203/rs.3.rs-11517111/v1
Primary Topic
School Choice and Performance
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Estimating the causal effect of class size on the achievement of Moroccan students: Evidence from TIMSS 2023

Abdellatif Chatri, Said Maizzou, Khalid Boulahris
1 links
Research Square
School Choice and Performance
preprint

Estimating the causal effect of class size on the achievement of Moroccan students: Evidence from TIMSS 2023

Abdellatif Chatri, Said Maizzou, Khalid Boulahris
preprint en

Abstract

Background .: Administrative caps on class size are the main source of quasi-experimental evidence on class-size effects, but whether a cap generates usable variation depends on the institutional setting. In Morocco, lower secondary classes average 36.5 pupils and a quarter of pupils sit in classes above the legal maximum of 40, yet no study has used the national cap to estimate the causal effect of class size, and whether the cap can identify it is unknown. Data and methods .: We use the TIMSS 2023 Grade 8 sample for Morocco (8,032 pupils, 240 classes, 229 schools) and five subjects: mathematics, physics, biology, chemistry and earth science. Teacher-reported class size is instrumented by the class size predicted from the maximum of 40 pupils and Grade 8 enrollment. A pupil-level design (weighted two-stage least squares with covariates and enrollment controls, and fuzzy regression discontinuity) is combined with a class-level design based on class mean scores. Inference uses the plausible values, school-clustered standard errors and Anderson--Rubin confidence sets robust to weak instruments. Results .: Least squares yields a negative association in every subject, significant in mathematics (-1.76 points per pupil, SE=0.47). In the pupil-level design, the first-stage F statistic falls from 87.4 to 3.9 under a cubic enrollment control, the Anderson--Rubin set is unbounded, and no discontinuity estimate is significant. In the class-level design, the Anderson--Rubin sets exclude zero under linear and piecewise-linear enrollment controls (-4.33 points per pupil in mathematics, set [-9.8,-0.5]) but not under a cubic control, and only 12 classes lie near the first three cut-offs. The evidence leans toward a negative effect of class size, but its magnitude is not identified: with a high cap and large cohorts, the rule generates too little local variation. JEL Classification: I21 , I28 , C26 , C21

Research Square
Mohammed V University (MA), Université Hassan 1er (MA)
Quality education
School Choice and Performance
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon