Inequalities in Higher Education: Profile of Distance Learning Courses

Ângela Barbosa Reis Lúz; Elisa Morandé Sales

DOI: 10.22167/2675-6528-202602674

Article derived from a Course Conclusion Work (TCC), with content based on the student’s original work and adapted to the editorial format of the E&S Magazine with the support of the ResumeAI tool, an artificial intelligence solution developed by Instituto Pecege for textual synthesis and organization.

Summary

The study analyzed the landscape of Brazilian higher education, focusing on the expansion of Distance Education (DE) and its relationship with access inequalities and institutional profile. The objective was to investigate the profile of undergraduate students in Brazilian higher education institutions (HEIs) in 2024, considering demographic variables and institutional characteristics, to understand the influence of social origin on the DE modality. Data from the Higher Education Census were used, and Multiple Correspondence Analysis and Decision Tree models were applied, with analyses conducted in Python. The results revealed a strong segmentation of the system, where the private sector dominated the offer of DE (95.9%), especially in areas such as Services, Computing, and Licentiate degrees, while the public sector remained predominantly in-person. Inferential analysis indicated that structural variables, such as academic organization and the course’s age profile, were determinants in the choice of modality. It was concluded that the expansion of DE, while increasing enrollment numbers, may be deepening educational stratification by offering distinct formative trajectories, conditioned by the structure of institutions and the type of course offered.

Keywords: Multivariate Analysis; Machine learning; Educational Inequality; Higher Education; Distance Learning (EaD).

1. Introduction

The expansion of higher education in Brazil, observed in the last three decades, is characterized by a significant increase in the number of enrollments and educational institutions. This growth reflects a growing demand for places, driven by various public policies for inclusion and affirmative actions that have diversified the student profile. In this scenario, the debate on the democratization of access to higher education, aiming at expanding opportunities for broader sectors of society, has become central (Peixoto, 2021; Senkevics, 2021a).

The Distance Education (EaD) modality emerged as a relevant educational format in this context. Its adoption was strongly encouraged from the enactment of the Law of Directives and Bases of National Education (LDB), in 1996, with emphasis on the private sector (Borges, 2024; Neto, 2020). The regulation of EaD in the country progressed with the elaboration of laws and decrees, such as Decree No. 2,494 of 1998, which defined the modality, and Ordinance No. 2,253 of 2001, which allowed the inclusion of part of the workload of face-to-face courses in a non-face-to-face format. The discussion on pedagogical mediation and the quality of training in the blended learning model intensified with Decree No. 5,622 of 2005 (Minosso, 2025).

The goals for the expansion of higher education, established in plans such as Axis 4 of the National Education Plan (PNE) 2001-2010 and Goal 12 of the PNE 2014-2024, signaled a transition from an elite system to a mass system (Senkevics, 2021a). In 2017, Decree No. 9.057/2017 facilitated the opening of distance education centers, and the flexibility imposed by the pandemic further accelerated the expansion of this modality. More recently, Ordinance No. 378/2025 established a new regulatory framework, seeking to qualify Distance Education and align it with technological and social demands.

However, this expansion did not occur without challenges. An expressive increase in enrollments in private institutions was observed to the detriment of public ones, revealing disparities in access and quality. The continuous growth of Distance Education enrollments, accompanied by the decline in face-to-face enrollments, was authorized and accredited by the Ministry of Education, but without adequate monitoring of private institutions. This scenario resulted in deficient training, with impacts on the quality of Brazilian Basic Education (Minosso, 2025; Senkevics, 2021b). Faced with this reality, the weight of students’ social origin on access and the quality of the offered education is questioned.

The expansion of enrollments highlighted disparities in the profile of beneficiaries for each modality, type of institution, and course (Senkevics, 2021b). Although the demographic transition has indicated a reduction in the population aged 18 to 24 years and an increase in higher education enrollments, representing an opportunity for investment in human capital (Rigotti, 2012), it becomes crucial to evaluate the quality parameters of Distance Education. It is fundamental to investigate whether the expansion of distance learning, while promoting access democratization, contributes to intensifying educational stratification, that is, whether the student’s social origin influences the probability of admission to certain institutions and courses.

In this sense, the study of the impacts of distance learning on educational inequalities is essential to support effective management practices and public policies, allowing the mapping of students’ main difficulties based on their sociodemographic profile. Thus, the objective is to analyze the profile of undergraduate students in Brazilian Higher Education Institutions (HEIs) in 2024, considering demographic variables such as race/color, gender, and age, in addition to institutional characteristics, in order to investigate the weight of these students’ social origin in the Distance Education modality.

2. Material and Methods

This research was characterized as a quantitative study, with an exploratory and predictive approach. The objective was to analyze the profile of undergraduate students in Brazilian Higher Education Institutions (HEIs) in 2024, investigating the weight of the social origin of these students in Distance Learning (DL) modality. For this purpose, demographic variables such as race/color, gender, and age were considered, in addition to institutional characteristics. The data source used was the microdata from the 2024 Higher Education Census, made available by the National Institute for Educational Studies and Research Anísio Teixeira (INEP).

The data collection was based on microdata from the Higher Education Census course registry, the main source of information on Higher Education Institutions (HEIs), their courses, faculty, and students in Brazil (Brazil, 2025a). For this study, 18 items from the course module were used, out of a total of 226 available variables. In data processing, selection criteria were applied: only undergraduate courses nationwide with enrolled students were filtered, and variables not related to the topic or outside the target population were discarded.

Demographic variables, such as sex, race, and age, as well as the funding variable, were considered based on the number of enrollments in each course. For their use, specific treatment was carried out, transforming them into categorical variables. The definition of the created categories was based on the results presented in the Statistical Notes of the Higher Education Census (Brazil, 2025b). In the end, the dataset comprised 23 final variables, totaling 556,054 courses analyzed.

The analyses were conducted using Multiple Correspondence Analysis (MCA) and Decision Tree models. MCA, an unsupervised and exploratory technique, was employed to investigate the association between the qualitative variables of undergraduate course profiles and the teaching modality. This technique analyzes the relationship between variable categories through the decomposition of the chi-squared statistic (Bussab, 2017).

To verify the association, Pearson’s chi-squared test was applied, with a significance level of 5%. Adjusted standardized residuals were calculated, where values above 1.96 indicated a significant association between the categories. Total inertia was calculated for each pair of variables, measuring the variance and information provided. A perceptual map was created for the joint visualization of associations between variable categories (Greenacre, 2008).

Classification Decision Tree models were used to investigate patterns and make predictions about course characteristics. This supervised machine learning technique employed the teaching modality (in-person or distance) as the qualitative response variable, encoded in a binary format (0 for in-person and 1 for distance).

The explanatory variables for the decision tree model were selected based on exploratory analyses and the results of the ACM. Variables such as sex, race, and student funding were excluded due to low association. Variables related to academic organization, educational network, course area, academic degree, student age group, and administrative category remained. The explanatory variables were transformed into numerical (dummy) ones, and the database was divided into 70% for training and 30% for validation.

The decision tree estimation used the classification algorithm implemented in the Scikit-Learn library, which builds the tree from the iterative choice of splits that maximize node homogeneity. For this, impurity criteria such as the Gini index or entropy were used (Fávero; Belfiore, 2024). The cost-complexity pruning (ccp_alpha) technique was applied to control the tree size, balancing the fit quality with the model complexity (Therneau; Atkinson, 2022).

To improve performance, hyperparameter optimization was performed via Grid Search with Stratified K-Fold cross-validation in 10 folds (Géron, 2021). The selection criterion for the best model was the Area Under the ROC Curve (AUC). Model performance was evaluated by metrics calculated from the confusion matrix, including accuracy, sensitivity, specificity, and precision. All analysis and modeling procedures were performed using the Python programming language, version 3.12.7.

In accordance with Law No. 12,527, of November 18, 2011 (Law on Access to Information), and Law No. 13,709, of August 14, 2018 (General Law on Personal Data Protection), the data analysis was performed exclusively at the IES and course level. It was not possible to perform analyses at the student level, ensuring the privacy and protection of individual student data.

3. Results and Discussion

The analysis of the Brazilian higher education landscape in 2024 revealed a marked segmentation between face-to-face and distance learning (Distance Education) modalities, with a strong influence of institutional characteristics and course profiles. The findings indicate that the expansion of Distance Education, although it has broadened access, has also consolidated distinct academic trajectories, especially between the public and private sectors. The research sought to understand the influence of social origin on the Distance Education modality, examining demographic and institutional variables, and the results point to an educational stratification conditioned by the structure of institutions and the type of course offered, as discussed by Senkevics (2021b).

Initially, the bivariate descriptive analysis showed that distance learning is predominantly offered by institutions in the private administrative category, accounting for 95.9% of Distance Education courses. In contrast, the municipal and federal public categories presented a higher concentration of in-person courses, with 93.9% and 70.8%, respectively. The public education network, in general, concentrated its offerings in in-person courses (58.2%), while the private network exhibited a vast predominance of Distance Education courses, reaching 95.9% of the offerings. This disparity highlights the central role of the private sector in the expansion of distance learning in the country, as pointed out by Borges (2024).

Regarding the academic degree, technological and licenciatura courses stood out for the high prevalence of the Distance Education modality, with 98.2% and 95.2% of the offerings, respectively. Bachelor’s courses also presented a majority of distance offerings (88.4%), being the only category with a notable percentage of in-person courses, which corresponded to 11.6%. This concentration of Distance Education in technological and licenciatura degrees suggests the training of professionals who will largely work in Brazilian Basic Education, a fundamental stage in the school trajectory, according to Minosso’s (2025) discussion on the quality of training.

The academic organization of Higher Education Institutions (HEIs) also revealed distinct patterns. Universities (94.5%), university centers (97.0%), and colleges (55.6%) showed a predominance of Distance Education courses. In contrast, federal institutes (79.0%) and federal education centers (96.7%) concentrated most of their courses in the in-person modality. These data reinforce the idea that the organizational structure of HEIs is a determining factor in the offering of the teaching modality, with federal public institutions maintaining a more traditional and in-person profile.

Regarding the general areas of the courses, most of the offerings occurred in the distance learning modality. The highest percentages of Distance Education were observed in Services (98%), Computing/ICT (96.1%), Business, Administration and Law (95.8%), and Education (95.5%). On the other hand, the areas with a percentage of face-to-face courses above 10.0% included Basic Programs (17.1%), Natural Sciences, Mathematics and Statistics (11.6%), and Agriculture, Forestry, Fisheries and Veterinary (20.1%). This distribution suggests that certain areas of knowledge are more adaptable or preferably offered in the distance format, while others maintain a strong face-to-face tradition.

The age profile of the students also showed significant distinctions. For both age groups analyzed, the majority of students were enrolled in Distance Education courses. However, only 0.6% of enrollments in face-to-face courses corresponded to students over 25 years old. In contrast, courses with a predominantly young profile (18-24 years old) registered the highest rate of face-to-face teaching among the demographic profiles, with 11.2% of the courses. This indicates that the Distance Education modality attracts a more mature audience, while face-to-face teaching is still predominant among young people, which may influence the dynamics of the labor market for these groups (Rigotti, 2012).

Regarding race/color, both groups (Black, Brown, and Indigenous – BBI, and non-BBI) showed greater adherence to distance learning enrollments. However, among Black, Brown, and Indigenous students, a slightly higher percentage of in-person enrollments was observed, at 7.5%. The distribution by sex of students enrolled in distance learning courses was balanced for both genders. Regardless of the financing situation, distance learning courses were the majority, but in-person courses stood out among those predominantly without financing, with 19.1% of offerings. These data, although descriptive, begin to outline an access profile that may be related to the students’ social background.

The regional analysis revealed that Distance Education is hegemonic in all regions of Brazil, with little variation, oscillating between 93.0% in the Center-West and 95.1% in the South. The highest concentrations of Distance Education courses were identified in the South Region, in parts of the North Region, and in the state of Mato Grosso. This uniformity in the predominance of Distance Education throughout the national territory underscores the modality’s reach, but also raises questions about the quality and monitoring of private institutions that offer it on a large scale, as discussed by Senkevics (2021b).

Multiple Correspondence Analysis (MCA) was applied to investigate the association between the variables and the teaching modality. The results of the chi-square test indicated that all variables under analysis showed a significant association with the teaching modality, with a significance level of 5%. The variables that provided the most information, based on total inertia, were administrative category (0.18), educational network (0.17), and academic organization (0.14). In contrast, the race/color (0.00) and sex (0.00) variables showed the lowest inertia values, suggesting a lower discriminatory capacity compared to the other institutional variables.

The analysis of adjusted standardized residuals, with values above 1.96 indicating significant association, revealed that distance learning courses are strongly associated with the private administrative category, the private educational network, and academic organizations such as universities. Furthermore, Distance Education showed association with technological and licensure degree courses, and with areas such as arts and humanities, computing and ICT, education, business, administration and law, and services. The profile of students enrolled in Distance Education courses was associated with predominantly adult students (over 24 years old), non-PPIs, and males, corroborating the patterns observed in the descriptive analysis.

In contrast, in-person courses were associated with administrative categories and public education networks, as well as academic organizations such as federal institutes, federal education centers, and colleges. The areas of in-person courses included agriculture, forestry, fishing and veterinary, engineering, production and construction, and health and welfare. The profile of students enrolled in in-person courses was associated with predominantly young students (18 to 24 years old), PPIs, and females. These ACM findings reinforce the understanding that the teaching modality is intrinsically linked to the nature and structure of the institutions that offer it, as well as to the demographic profile of the students.

The perceptual map, which visualizes the association between variables, highlighted that the first dimension structures the organizational form of HEIs in Brazil. On the right side of the map, public institutions and face-to-face courses were grouped, especially those offered by federal and municipal institutions. On the left side, private institutions are found, associated with the distance learning modality and the predominance of funding. This main dimension shows that the differentiation structure of the higher education system is strongly associated with institutional characteristics, with a clear dichotomy between public face-to-face education and private distance education.

The second dimension of the perceptual map explored the organization of degree levels of the courses. The right side was composed of more technical areas, such as agriculture, natural sciences, and basic programs, associated with federal and municipal institutions, and a younger student profile. The left side, in turn, grouped undergraduate courses, such as education, with a profile of adult and predominantly female students. Sociodemographic variables, such as sex, race, and funding, were located near the center of the graph, indicating a lower discriminatory capacity to structure the main dimensions, in line with the low inertia values observed.

The application of classification decision tree models allowed for the investigation of patterns in the data and predictions about course characteristics. The final model, selected for its greater interpretative capacity, showed adequate predictive performance, with an Area Under the ROC Curve (AUC) of 0.92, evidencing strong discrimination capacity between in-person and distance learning courses. The classification rules were built through successive combinations of institutional characteristics and student profiles, confirming the relevance of these variables in determining the teaching modality.

The educational network was identified as the main discriminating variable, followed by academic organization and student age profile. The course area, academic degree, and administrative category complemented the classification, although with less weight. Distance Education courses were associated with a more adult age profile, with a higher participation of students over 24 years old, and with private institutions. In contrast, face-to-face courses showed a greater association with younger students. Regarding the course area, a greater association of courses in exact sciences and health areas with the face-to-face modality was observed, while courses in management and education areas showed a greater association with the Distance Education modality.

It is important to emphasize that the analysis of the importance of the variables confirmed that institutional factors, such as the educational network and the type of academic organization, and demographic factors, such as age profile, are the main determinants of the teaching modality. However, the interpretation of these results must consider the study’s limitation, where the unit of analysis was the course and not the individual student. Demographic characteristics were aggregated at the course level, which may reduce the ability to capture individual differences and inequalities associated with student profiles, as pointed out by Senkevics (2021b).

In summary, the research results revealed a strong segmentation in the Brazilian higher education system, where the expansion of Distance Education is predominantly driven by the private sector and associated with specific institutional and course profiles. The analysis demonstrated that structural variables, such as academic organization and the educational network, along with students’ age profile, are determinants in the choice of modality. This dynamic suggests that the expansion of distance learning, while promoting access, may be deepening educational stratification by offering distinct formative trajectories, conditioned by the structure of institutions and the type of course offered, impacting the democratization of access and the quality of training.

4. Conclusion

This study analyzed the profile of undergraduate students in Brazilian higher education institutions (HEIs) in 2024, considering demographic variables and institutional characteristics, with the aim of investigating the influence of social origin on Distance Education (DE) modality. A marked segmentation was found in the educational system, where the private sector dominated the offer of DE, accounting for 95.9% of courses, while the public network maintained its predominance in face-to-face courses. It was observed that the distance modality was concentrated in technological and licensure degrees, as well as in areas such as Services, Computing, and Education, and attracted a predominantly adult student profile. Inferential analysis, through Multiple Correspondence Analysis and Decision Tree models, confirmed that structural variables, such as the educational network and academic organization, along with the students’ age profile, were the main determinants of the teaching modality. These findings indicate that the expansion of DE, although it broadens access, may be deepening educational stratification by offering distinct formative trajectories, conditioned by the structure of institutions and the type of course offered. The study’s main contribution lies in supporting effective management practices and public policies, by mapping access dynamics and students’ main difficulties based on their sociodemographic profile, highlighting the need for careful attention to equity in the educational system.

However, the study presented the limitation of having the unit of analysis at the course level and not at the individual student level, which may have reduced the capacity to capture the differences and inequalities associated with the students’ sociodemographic profile. Variables such as sex and race, for example, showed less discriminatory capacity in the models, although the literature recognizes their relevance for understanding educational inequalities. For future studies, the inclusion of performance indicators, such as those from ENADE, and institutional quality assessments is suggested. This approach would allow verifying whether the current expansion and democratization of access effectively translate into equity of opportunities and consistent pedagogical quality, deepening the understanding of the impacts of distance learning on the training of professionals and on the quality of Brazilian Basic Education.

Bibliographic References

Borges, C.; Souza, M.R.; Peixoto, R.T.O.; Paiva, L.S.; Almeida, P.C.; Oliveira, D.S. 2024. Maior inclusão social foi evidenciada na educação a distância? Análise da democratização de acesso ao ensino superior em saúde. Revista Educação Online 19(45): 1-18.

Brasil. Instituto Nacional de Estudos e Pesquisas Educacionais Anísio Teixeira [INEP]. 2025a. Censo da Educação Superior. Disponível em: https://www.gov.br/inep/pt-br/areas-de-atuacao/pesquisas-estatisticas-e-indicadores/censo-da-educacao-superior. Acesso em: 20 jan. 2026.

Brasil. Instituto Nacional de Estudos e Pesquisas Educacionais Anísio Teixeira [INEP]. 2025b. Censo da Educação Superior 2024: notas estatísticas. Disponível em: https://download.inep.gov.br/publicacoes/institucionais/estatisticas e indicadores/notas est atisticas censo escolar da educacao superior 2024.pdf. Acesso em: 20 jan. 2026.

Bussab, W.O.; Morettin, P.A. 2017. Estatística Básica. 9ed. Saraiva, São Paulo, SP, Brasil.

Fávero, Luiz Paulo; Belfiore, Patrícia. Manual de análise de dados: estatística e machine learning com Excel®, SPSS®, Stata®, R® e Python®. 2. ed. Rio de Janeiro: LTC, 2024.

Greenacre, M. 2008. La Práctica del Análisis de Correspondencias. Fundación BBVA, Bilbao, Espanha. Disponível em: http://www.carme-n.org. Acesso em: 06 jan. 2026.

Géron, Aurélien. Mão à obra: aprendizado de máquina com Scikit-Learn, Keras & TensorFlow. 2. ed. Rio de Janeiro: Alta Books, 2021.

Minosso, A.; Silva, G.T.F.; Veloso Júnior, C.M.; Amaral Filho, F.S. 2025. Educação superior a distância: entre fundamentos históricos e políticas públicas no cenário do marco regulatório atual. Contribuciones a las Ciencias Sociales 18(10): e21536. Disponível em: https://ojs.revistacontribuciones.com/ojs/index.php/clcs/article/view/21536. Acesso em: 06 jan. 2026.

Neto, V.B.S.; Borges, M.C. 2020. Educação a distância no Brasil: a regulamentação como falácia da democratização e acesso ao ensino superior de qualidade. Revista Educação e Políticas em Debate 9(1): 53-72.

Peixoto, M.C.L. 2015. Políticas de avaliação e regulação na educação superior brasileira: a difícil relação entre expansão, melhoria da qualidade e controle de resultados. Propuesta Educativa (43).

Rigotti, J.I.R. 2012. Transição demográfica. Educação & Realidade 37(2): 467-490. Disponível em: http://www.ufrgs.br/edu_realidade. Acesso em: 06 jan. 2026.

Senkevics, A.S. 2021. O acesso, ao inverso: desigualdades à sombra da expansão do ensino superior brasileiro, 1991-2020. Tese de Doutorado em Educação. Faculdade de Educação, Universidade de São Paulo, São Paulo, SP, Brasil. Disponível em: https://www.teses.usp.br/teses/disponiveis/48/48137/tde-11012022-103758/publico/ADRIANO_SOUZA_SENKEVICS_rev.pdf. Acesso em: 21 set. 2025.

Senkevics, A.S. 2021a. A Expansão Recente do Ensino Superior. Instituto Nacional de Estudos e Pesquisas Educacionais Anísio Teixeira, Brasília, DF, Brasil.

Therneau, Terry M.; Atkinson, Elizabeth J. An Introduction to Recursive Partitioning Using the RPART Routines. Mayo Foundation, 2022.

Article originating from the Final Course Project of the Specialization in Data Science and Analytics of the MBA USP/Esalq

To learn more about the course, click here and access the MBX Academy platform