Using data from the 2018 and 2023 International Computer and Information Literacy Study, this study investigates inequalities in digital skills among eighth-grade students in Italy through an innovative integration of multilevel modeling and quantile regression, which provides deeper insights into the distributional dynamics of digital competencies. The study examines the influence of socio-demographic factors, ICT usage habits, self-efficacy and school environment on digital literacy. A multilevel quantile regression is used to a) examine the gaps in the whole distribution of digital literacy, not just at the mean, and b) take into account the hierarchical structure of the data, where students are nested within schools. The results show significant territorial disparities, with Northern and Central regions outperforming the South. Gender differences favour females at average levels of digital literacy but diminish at the extremes of the distribution. Socio-economic background and educational aspirations emerge as key predictors of digital literacy. Classroom use of general digital applications has a positive effect on low-achieving students, while specific applications show a negative association that warrants further investigation. The school effects are more pronounced for students with low to moderate levels of competence. Between 2018 and 2023, average performance improved; however, underlying inequalities, particularly territorial disparities, persisted and in some cases even worsened. The pandemic-induced digitalization of education did not lead to uniform gains in digital competence but rather exacerbated existing gap. These findings highlight the urgent need for tailored policies to address regional disparities, promote digital inclusion and optimise ICT education.

Keywords: digital literacy; ICILS; gender gap; territorial gaps; multilevel models; quantile regression.

JEL codes: C21, I24.