Principal-reported school absenteeism in Morocco: contextual correlates and same-respondent attendance information in TIMSS 2023

Background Principal questionnaires in international large-scale assessments provide school-level information about perceived attendance problems, but their interpretation depends on distinguishing broad contextual characteristics from measures that are conceptually close to the outcome and reported by the same respondent. This study examined principal-reported student absenteeism in Morocco and assessed how the evidence changed when a same-principal teacher-attendance indicator was added to a contextual school specification. Methods We analysed the official Moroccan Grade 8 school file from TIMSS 2023. Of 250 schools, 243 had valid absenteeism responses and 232 formed the common complete-case sample. Population-oriented estimates used the TIMSS school weight and 150 full jackknife repeated-replication estimates. Survey-weighted logistic regression evaluated five contextual school indicators, followed by a specification that additionally included a two-item principal report of teacher lateness and absenteeism. Out-of-sample discrimination was assessed with 10 repetitions of nested, zone-grouped cross-validation using Elastic Net regularization. Results The weighted proportion of schools whose principals rated student absenteeism as a serious problem was 43.1% (JRR 95% CI 29.2–57.0%). No contextual indicator remained significant after false-discovery-rate correction. After the teacher-attendance indicator was added, principal-reported teacher-attendance problems (OR = 8.24, 95% CI 3.60–18.83) and economic disadvantage (OR = 2.91, 95% CI 1.25–6.79) remained associated with the outcome. Out-of-sample discrimination was near chance for the contextual specification (balanced accuracy = 0.511; ROC AUC = 0.499) and was higher for the specification including teacher attendance (balanced accuracy = 0.728; ROC AUC = 0.774). Conclusions Broad contextual indicators alone contained little information for discriminating the contemporaneous principal-reported absenteeism outcome. The large increment associated with the same-respondent teacher-attendance measure may reflect shared organizational attendance conditions, common-source measurement covariance, or both. The findings concern cross-sectional association and internal out-of-sample discrimination; they do not establish prospective forecasting or criterion validity.