Are PISA 2022 Creativity-Related Self-Reports Comparable Across Education Systems? Measurement Invariance, Alignment, and Associations with Creative-Thinking Performance

PISA 2022 creativity-related self-reports are used to study differences among education systems, but their interpretation depends on measurement structure, scoring and estimator choice. We audit six scales using item-level responses from up to 73 systems, the OECD item-response scaling record and associations with creative-thinking performance. Because the rotated questionnaire leaves almost no complete records, the primary models read weighted pairwise-present moments; the parameter estimates are inferential, but the fit indices and dispersion-reduction R2 computed from them are descriptive summaries of that approximation rather than survey-robust tests, and we mark which conclusions are which. The continuous models show substantial unidimensional misfit for four published indices. Loading constraints produce small fit decrements under the retained structures, scalar constraints larger ones. Alignment for four constructs concerns a selected item subset and cannot establish comparability of the published scale. Re-estimating all six by full-information marginal maximum likelihood on the ordinal responses, which handles the planned missingness and the response format exactly, lowers every loading R2 and rearranges their ordering (ρ = 0.14), leaves the intercept ordering essentially intact (ρ = 0.94), and reproduces the system means at r = 0.77 to 0.97 while moving 18% to 53% of systems by more than five rank positions. Within systems, observed-score associations are small and heterogeneous. Construct-level summaries are reasonably robust to these choices; individual system positions are not.