Comparison of grades by assessment modality in an Ecuadorian public university during the pandemic transition (2019–2020)

The COVID-19 pandemic produced an abrupt transition from in-person to virtual teaching in Latin American public universities, with effects on the comparative validity of summative assessments that remain insufficiently documented. The present study examines whether the average grades obtained by undergraduate students at an Ecuadorian public university differ systematically between the 100% in-person modality (2019) and the 100% virtual modality (2020), and explores the institutional mechanisms that may have produced such differences. A non-experimental, cross-sectional convergent mixed-methods design was adopted, drawing on two complementary sources of evidence: (i) a census sample of more than 19,000 academic records corresponding to three evaluative instances (first partial, second partial, and recovery examinations); and (ii) a structured survey administered to 35 instructors who supervised assessment processes during the virtual period. The quantitative analysis included descriptive statistics, the Welch t-test, the Welch ANOVA with Games–Howell post-hoc correction, and non-parametric robustness tests (Mann–Whitney and Kruskal–Wallis). The perceptual component was modeled by means of Partial Least Squares Structural Equation Modeling (PLS-SEM) with bootstrap (500 replications). The results show grades that are significantly higher and less dispersed in the virtual modality (Cohen's d between −0.53 and −0.69; p < 0.001 in all instances), a pattern corroborated by the non-parametric tests. The structural model identifies an associative pattern consistent with full mediation: digital gaps are associated with the weakening of evaluative-control mechanisms ( β = 0.658, p < 0.001), and this weakening predicts perceived grade inflation ( β = 0.878, p < 0.001), with a significant indirect effect [ β = 0.578; 95% CI (0.545, 1.306)]. The convergence between census-based evidence and perceptual evidence suggests that the increase observed in the virtual modality may reflect, at least in part, the erosion of institutional control mechanisms during the health emergency rather than an intrinsic improvement in learning. The direct comparison between both modalities therefore demands interpretive caution, in view of the confound between modality and instrument format and of the absence of sociodemographic control variables. These findings have implications for the validity of academic-quality indicators, for the design of post-pandemic hybrid assessment models, and for quality-assurance policies in Latin American higher education.