Are Environmental Attitudes and Behaviors Comparable Across Countries? A Measurement Invariance Analysis of TIMSS 2023

Background Environmental attitudes and behaviors are critical constructs in global education research, yet most existing studies in this area are conducted within single countries or regions, limiting opportunities for cross-national comparison. The 2023 TIMSS dataset provides a rare opportunity to examine adolescent environmental attitudes and behaviors at scale, encompassing nearly 300,000 eighth-grade students across 44 countries. Methods This study used confirmatory factor analysis (CFA) with complex survey design corrections and multiple imputation to evaluate the factor structure and model fit of the TIMSS Environmental Attitudes and Behaviors Framework, which assesses students' views on environmental preservation, resource utilization, and engagement in pro-environmental actions. Multi-group CFA (MGCFA) was employed to test measurement invariance by country at the configural, metric, and scalar levels. Results A three-factor model—environmental preservation, environmental utilization, and environmentally-friendly behaviors—demonstrated reasonable overall fit (CFI = 0.9414, RMSEA = 0.0532) when evaluated with established psychometric standards. However, fit indices for the individual factors varied, with environmentally-friendly behaviors showing marginal fit (RMSEA = 0.0916). Despite these fit concerns, metric invariance was supported across countries, enabling meaningful comparisons of structural relationships. Scalar invariance was not supported across countries, limiting the interpretability of latent mean comparisons. Conclusions The scope and scale of the TIMSS 2023 data make it one of the most powerful resources available for cross-national research on environmental attitudes and behaviors among adolescents. While metric invariance supports many comparative cross-national analyses, scalar invariance limitations underscore the need for cautious interpretation of latent means. Importantly, variation in model fit may reflect genuine cultural and ecological differences, not merely measurement error. As such, measurement invariance should be approached not only as a statistical requirement but also as an analytic lens for understanding how environmental constructs manifest differently across global contexts.