Critical reading and the structure of institutional academic performance: evidence from Colombia's Saber 11 examination

Large-scale assessments provide important evidence on educational performance, but institution-level results may provide an incomplete picture when academic domains are interpreted as independent indicators or simple ordinal comparisons. In a secondary analysis of the 2024 Icfes institutional classification file, this study examined the structure of the five Saber 11 institutional classification indices in Colombia, with particular attention to the position of Critical Reading. Analyses included 10,043 educational establishments and used confirmatory factor analysis, symmetric one-domain-out structural parameterizations, and multigroup measurement-invariance models across official and non-official sectors. The five indices were extremely highly correlated, and the first dimension accounted for 96.45% of the total standardized variance. A five-domain common-factor model yielded standardized loadings from .952 to .991, although global fit was mixed, indicating strong commonality without strict unidimensionality. Critical Reading was strongly associated with the latent dimension defined by the remaining four domains ( β = .988), but analogous one-domain-out specifications for Mathematics, Natural Sciences, Social Sciences and Citizenship, and English reproduced essentially the same covariance structure. Thus, the large Critical Reading coefficient did not identify a uniquely privileged directional role. Full metric invariance across institutional sectors was not supported; a limited partial metric solution was obtained after allowing Mathematics and English loadings to vary. dMACS effect sizes indicated non-negligible practical noninvariance for Mathematics (0.283, 95% bootstrap CI [0.266, 0.299]) and especially English (0.609, 95% bootstrap CI [0.584, 0.635]). Standardized Critical Reading associations remained very similar across sectors, although exact equality of the unstandardized coefficient was rejected. These findings indicate that Saber 11 institutional performance is characterized by a dominant but not perfectly unidimensional common structure. Critical Reading is a substantively meaningful marker of this multidomain organization, but its cross-domain relevance should not be interpreted as statistical primacy or causal influence.