Whose School Is Safe? Informant-Specific Validity in School Safety and Student Well-Being Assessment Across 63...

Schools constitute the principal setting for the identification of children's psychosocial risk and the delivery of preventive support, and school-health surveillance therefore depends on informant reports. Selection of the reporting informant is constrained by feasibility, cost, and questionnaire length, which frequently preclude multi-informant assessments, and the question of whether adult and child reports index a common property has predominantly been examined in single-country designs. The present study estimated the criterion-related validity of child, teacher, and principal reports of school safety across multiple education systems, against outcomes spanning children's experience, institutional functioning, and externally assessed achievements; the estimand was thus informant-specific validity rather than cross-informant agreement.

TIMSS 2023 Grade 4 data from 12,955 schools in 63 education systems (395,961 children, 27,100 teacher records) were analyzed, with child, teacher, and principal reports modeled as separate informants. Students were randomly split within schools so that no student contributed both a predictor and an outcome; the split was replicated 50 times and averaged. Convergence was evaluated with a multitrait-multi-informant model, with ceiling effects and differential unreliability specified as rival explanations; informant contributions were estimated by an exact relative-importance (LMG) variance decomposition, with response surface analysis reported as a secondary characterization of directional discrepancy.

Median school-level cross-informant correlations were below 0.10 and were not attributable to ceiling effects or to unreliability. With all three informants retained, the child report accounted for 95% (95% CI 92-97) of the informant-explained variance in school belonging and 75% (65-82) in peer victimization, but 44% (28-59) in mathematics, where the three contributions were approximately equal. These proportions were computed on a modest total: the three informant scores jointly explained between 1% and 13% of between-school variance. The pattern was robust to informant, item, covariate, and weighting specifications, and to correction for differential predictor unreliability.

Criterion-related informant validity in school safety measurement was domain-specific. For outcomes describing children's experience, adult reports contributed substantially lesser than child reports and cannot be regarded as interchangeable substitutes for children's reports of experienced safety. Because school-health surveillance, risk identification, preventive targeting, and resource allocation depend on these inputs, valid monitoring requires student self-reports rather than adult proxy reports alone.