The digital divide refers to persistent inequalities in access to ICT, patterns of engagement, and the benefits derived from technology (van Dijk, 2020). This divide is shaped by factors such as gender, socioeconomic status (SES), and country of origin (Acilar & Sæbø, 2023; van de Werfhorst et al., 2022). It is typically conceptualized across three levels: (L1) access to ICT, (L2) ICT use and digital skills and attitudes, and (L3) the outcomes gained from ICT use (van Dijk, 2020). As access (L1) has become increasingly widespread, recent research highlights the importance of examining disparities in L2 and L3 (Acilar & Sæbø, 2023; van de Werfhorst et al., 2022). For example, even when access is more equal, gender differences in ICT use and skills (L2) persist and their underlying causes remain unclear (Acilar & Sæbø, 2023). 

We examine the second level of the digital divide (L2) through the lens of a 21st century digital skill – computational thinking (CT). CT has gained substantial global attention in recent years, becoming embedded in school curricula worldwide (Fraillon, 2024) and amassing a growing body of research (Tang et al., 2020). Despite increased emphasis on developing CT, such advanced skills are still rarely developed at school in Finland (Oinas et al., 2023). Large-scale assessments, such as the International Computer and Information Literacy Study (ICILS), offer an opportunity to analyze students’ CT skills alongside students’ backgrounds, situating CT within the broader framework of the digital divide. In addition to evaluating school ICT infrastructure (L1), ICILS also evaluates how students use general and specialist applications (e.g., text editors, spreadsheets, programming software, etc.), how much they learn CT at school, their confidence in solving CT-related tasks, and their performance on practical CT tasks (L2). Finnish schools are well equipped technologically and there is low between-school variability in ICT resources (Fagerlund et al., 2024, p. 32). Thus, we have excluded L1, access to ICT, from our analysis, and focus on L2, CT-related self-efficacy, skills and use of ICT. Moreover, most secondary analyses of ICILS data focus on computer and information literacy (CIL), and only a small number of studies have thus far analyzed CT (Campos & Scherer, 2024; Fang et al., 2025). In addition, CT was introduced into the Finnish national core curriculum in 2016, and ICILS 2023 captures the first cohort of students educated under this curriculum (Fraillon, 2024).

We also propose and test a slightly different view of L2 of the digital divide. Previous studies have used CIL and/or CT scores as sole measures of digital skills, often including one or more influencing variables such as gender or self-efficacy while applying linear and parametric methods of analysis. We propose a more nested view which better reflects the circular connection between ICT use and skills (van Dijk, 2020) and utilizes a nonlinear and probabilistic method of analysis. This nested model combines variables in a way that allows detailed exploration of, e.g. how frequently students use different types of ICT, while also examining the influence of common predictors such as gender. Thus, using Factor Mixture Modelling, the current study examines what student subpopulations emerge based on student backgrounds and CT-related digital skills and engagement with technology, and whether these indicate a L2 digital divide in Finland. We ask the following research question.

To what extent do gender and SES differentiate student profiles that are based on CT-related ICT use, self-efficacy, and skills, indicating patterns of the digital divide in Finland?

We analyzed 4028 Finnish Grade 8 students’ responses (51.09% girls) from ICILS 2023. ICILS examines exogenous variables, such as age and gender, that influence how ICT is used and what digital skills students develop at school. CT-related ICT use is measured as the use of specialist applications in lessons. Specialist applications involve multimedia production tools, simulations and modelling software, computer programming environments, and more (Fraillon, 2024). CT skills are measured through practical tasks with real-world focus that require cross-curricular and diverse skills, as well as practicing responsible and ethical use of computers and the ability to create and execute algorithms to solve problems (Fraillon, 2024). The data for the present study were analyzed using Factor Mixture Modelling (FMM) in Mplus 8.11 and 9. FMM is a combination of latent class analysis and factor analysis, which models unobserved heterogeneity via categorical latent variables, and within-class covariation of the observed variables via continuous factors (Clark et al., 2013; Lubke & Muthén, 2005; Muthén, 2008). As ICILS is a cross-sectional study and the factor scores exhibited a non-normal distribution, we used non-parametric FMM which approximates the distribution via discrete classes (Clark et al., 2013; Muthén, 2008). Our FMM model is made up of classes, factors (ICT use, self-efficacy, and learning CT), factor items, covariates (gender and SES), and outcomes (CT scores). The outcomes are calculated via the Bolck, Croon, and Hagenaars (BCH) method. We regressed both the classes and the factors on the covariates, gender and SES, to investigate their effects on both class distribution and factor means. The model evaluation was based on statistics commonly used in mixture models such as the Bayesian Information Criterion (BIC) (Muthén, 2008). We used the following variables –– questions and items –– and scales, based on how they were grouped together in the ICILS 2023 Technical Report: • ICT use factor –– IS3G23:C-G,I,J,L,M –– 0–3: Never – In every or almost every lesson –– e.g., (I) Interactive digital learning resources • Self-efficacy factor –– IS3G24:D,K,L –– 0–3: I do not think I could do this – Very well –– e.g., (L) Use visual coding [...] • Learning CT factor –– IS3G26:E-I –– 0–3: Not at all – To a large extent –– e.g., (H) Systematically test computer programs [...] • Gender covariate –– SGENDER –– 1–2: Girl, Boy • SES covariate –– national index of socioeconomic background (NISB ) –– -3.209–1.783: low – high (mean 0.002)

We extracted 5 classes (C1–C5) based on the used variables. The results indicate a gender- and SES-based digital divide among Finnish students. The two highest-performing classes, C1 and C3 (24% & 34.9% of the population), have the highest odds of high SES students, and the highest odds of girls. The worst-performing class, C2 (4.3%), has the highest odds of boys, the highest odds of low SES, and shows CT proficiency at a whole level below the other 4 classes. In the two middle-performing classes, C4 and C5 (19.5% & 17.3%), SES is significantly lower than C1 and C3. C4 has higher odds of boys, while C5 is more balanced in terms of gender. Even though C4 and C5 are the middle-performing classes, their CT scores are still below the standardized mean of 500 (Fraillon et al., 2025). It is noteworthy that the lowest-performing class (C2) reports using ICT in most lessons, alongside moderate self-efficacy and moderate-high learning of CT. In contrast, the high- and middle-performing classes report that they use ICT either never or only in some lessons. Further, although the middle-performing classes (C4 & C5) report moderate learning of CT, their CT scores remain significantly lower than those of the high-performing classes (C1 & C3), which report learning CT only to a small extent. While these findings are consistent with prior research, they were reached through an analytical approach rarely applied in educational research. The nested model offers a detailed description of student groups, revealing clear mismatches between students’ self-reported ICT use, attitudes, and learning, and their actual CT performance, underscoring a need for future research. Our findings have clear implications for educational policy and teaching practice. They suggest that simply including or emphasizing CT in the curriculum is insufficient to address persistent divides in students’ CT skills.

Acilar, A., & Sæbø, Ø. (2023). Towards understanding the gender digital divide: a systematic literature review. Global Knowledge, Memory and Communication, 72(3), 233–249. https://doi.org/10.1108/GKMC-09-2021-0147 Campos, D., & Scherer, R. (2024). Digital gender gaps in students’ knowledge, attitudes and skills: an integrative data analysis across 32 Countries. Education and Information Technologies, 29, 655–693. https://doi.org/10.1007/s10639-023-12272-9 Clark, S., Muthén, B., Kaprio, J., D'Onofrio, B., Viken, R., & Rose, R. (2013). Models and Strategies for Factor Mixture Analysis: An Example Concerning the Structure Underlying Psychological Disorders. Structural Equation Modeling: A Multidisciplinary Journal, 20(4), 681-703. https://doi.org/10.1080/10705511.2013.824786 Fagerlund, J., Leino, K., Niilo-Rämä, M., Puhakka, E., & Markkanen, I. (2024). Kohti digiosaamisen strategista kehittämistä: Kansainvälinen monilukutaidon ja ohjelmoinnillisen ajattelun tutkimus (ICILS 2023) [Towards the strategic development of digital competence: International research on multiliteracy and computational thinking]. Koulutuksen tutkimuslaitos. https://doi.org/10.17011/ktl-t/40 Fang, G., Wang, J., Chan, P., & Kalogeropoulos, P. (2025). Understanding the gender divide in digital literacy in four European countries: A comprehensive decomposition analysis using unconditional quantile regression. Computers & Education, 229, 105264. https://doi.org/10.1016/j.compedu.2025.105264 Fraillon, J. (2024). An international perspective on digital literacy: Results from ICILS 2023. IEA. Fraillon, J., Rožman, M., Meyer, S., Musu, L., Liaw, Y.-L., Christiansen, A., & Tieck, S. (2025). ICILS 2023 Technical Report. IEA. Lubke, G. H., & Muthén, B. (2005). Investigating Population Heterogeneity with Factor Mixture Models. Psychological Methods, 10(1), 21-39. https://doi.org/10.1037/1082-989X.10.1.21 Muthén, B. (2008). Latent Variable Hybrids: Overview of Old and New Models. In G. R. Hancock, & K. M. Samuelsen, Advances in Latent Variable Mixture Models. IAP Information Age Publishing, Inc. Oinas, S., Vainikainen, M.-P., Asikainen, M., Gustavson, N., Halinen, J., Hienonen, N., Kiili, C., Kilpi, N., Koivuhovi, S., Kortesoja, L., Kupiainen, R., Lintuvuori, M., Mergianian, C., Merikanto, I., Mäkihonko, M., Nazeri, F., Nyman, L., Polso, K.-M., Schöning, O., Svedholm-Häkkinen, A. M., . . . Hotulainen, R. (2023). The impact of digitalisation on learning situations, learning and learning outcomes in lower secondary schools: Initial results and recommendations of a national research project. Tampere University and University of Helsinki. https://urn.fi/URN:ISBN:978-952-03-2782-8 van de Werfhorst, H. G., Kessenich, E., & Geven, S. (2022). The digital divide in online education: Inequality in digital readiness of students and schools. Computers and Education Open, 3, 100100. https://doi.org/10.1016/j.caeo.2022.100100 Tang, X., Yin, Y., Lin, Q., Hadad, R., & Zhai, X. (2020). Assessing computational thinking: A systematic review of empirical studies. Computers & Education, 148, 103798. https://doi.org/https://doi.org/10.1016/j.compedu.2019.103798 van Dijk, J. (2020). The Digital Divide. Polity Press.