Based on the PISA data in 2003 and 2006, this paper develops a hierarchical linear model to identify multi-level explanatory variables of the divide of adolescents' self-reported digital skills. At the country-level, the study finds a generally negative relationship between the ICT penetration rate of a country and adolescents' digital skills, implying that increased ICT penetration rate does not guarantee that adolescents have more chance to learn and use ICTs. Educational expenditure on secondary education is positively related to digital skills in 2006, but not in 2003. At the school-level, whether the school is public or private does not elicit impact on adolescents' self-reported digital skills; school ICT access is positively related to students' self-reported digital skills. At the individual-level, self-reported digital skills is affected by home ICT access, adolescents' Socio-Economic Status, gender and their history of using ICTs. The cross-level interaction effect of ICT penetration rate and home ICT access on self-reported digital skills is significant in the 2003 model, but not significant in the 2006 model. The cross-level interaction effect of ICT penetration rate and school ICT access is significant in the 2006 model. The interaction effect of school ICT access and home ICT access is not supported by the dataset of either year. (Contains 5 tables.)