Children’s environmental self-reports: dimensionality, response-style sensitivity, and cross-national comparability in TIMSS 2023

Introduction The 2023 cycle of the Trends in International Mathematics and Science Study (TIMSS) is the first international assessment to ask Grade-4 children (around age 10) directly about their environmental values and self-reported pro-environmental action. We examined the association between values and reported action, and how far that association and cross-national comparisons of means were sensitive to measured proxies for response style. Methods We analyzed data from 58 education systems and 359,098 students, after excluding benchmarking files that duplicate national samples. Measurement models assessed dimensionality and comparability. Within-system associations were pooled across systems, and sensitivity analyses adjusted for acquiescence and extreme responding indexed by unrelated mathematics and science items. Results Stronger environmental values were associated with more self-reported action across all systems ( β = 0.45). Adjustment for response style attenuated the value–action and knowledge–action associations by 9% and 17%, respectively, but left both intact. At the system level, the two proxies accounted for 25% of the between-system variance in mean values (95% bootstrap CI [0.09, 0.47]; leave-one-system-out range 0.22–0.30). Girls reported stronger environmental values than boys (d = 0.20; d = 0.18 in systems with scalar gender invariance). Cross-system metric and scalar invariance were not supported. Discussion The findings support within-system analyses and gender comparisons where the required invariance holds, but do not support ranking education systems by mean environmental values. The association between system means and response style indicates potential confounding with scale use; it does not estimate the proportion of observed differences that is artefactual.