This chapter discusses policies to promote completion rates once students arrive in higher education institutions. It highlights the value of traditional in-person tutoring and mentoring programmes to support academic achievement and student well-being before exploring the potential applications of emerging data-driven tools in supporting student success. It provides an overview of good practice for using analytical tools to support student success in Portugal and internationally and considers avenues for future policy action in Portugal specifically.
Higher Education in Portugal
Policies for Access and Success
Abstract
Lower-than-desired student completion rates in higher education are a challenge facing many OECD member countries, including Portugal. When students fail to progress in higher education, there are significant impacts on individuals, as well as institutions and countries (Bound and Turner, 2011[1]). At the individual level, students who drop out of their programmes limit their expected future earnings by curtailing their access to more qualified and better-paid job opportunities (Carneiro, Heckman and Vytlacil, 2011[2]; Gunderson and Oreopolous, 2020[3]).1 At the institutional level, negative consequences include challenges in planning revenue flows, allocating resources and the risk of experiencing reputational damage or falling short of fulfilling their social mission. At the national level, low levels of completion can limit the effect of any increased enrolment rates on overall human capital, which in turn can limit capacity for innovation and productivity growth in local and national economies.
The reasons students drop out of their programmes can be complex and influenced by students’ underlying academic ability, as well as student decisions about the time and resources they choose to invest in their education. From an equity perspective, high non-progression rates can be particularly concerning if they reinforce social disparities. This can happen if the most socio-economically disadvantaged groups are disproportionately likely to drop out of their degree programmes.
This chapter summarises a longer note on student success monitoring systems, delivered to the Portuguese authorities as part of the project. It first provides an overview of the available data on student success in Portugal. It subsequently discusses ongoing initiatives to combat drop-out in higher education. Finally, it explores developments in the field of student tracking methods and considers possible avenues for the implementation of successful data-driven approaches, along with necessary conditions for scaling these approaches to a national level.
Key findings
Drop-out rates in Portugal are broadly in line with the OECD average, but Portugal stands out slightly compared to other OECD member countries in the share of students who drop out after the first year, rather than during their first year. Data from Portugal indicate that most of the variation in completion rates by field of study appears after the first year.
Challenges related to social and academic integration can appear early but also differ and fluctuate along student careers and between study programmes. Indeed, almost one quarter (24%) of students in Portugal report feeling like they do not belong in higher education.
Several initiatives across Portugal aim to promote the social and academic integration of their students, primarily focusing on first-year students. Most higher education institutions in Portugal provide tutoring and mentoring services to their students to address academic, social and well-being challenges that students experience. These activities have been encouraged by the Programme for Promoting Success and Reducing Drop-out Rates, funded by the EU through via the Recovery and Resilience Plan. As the EU Recovery and Resilience Facility is set to end by 2026, institutions face the challenge of making sure that the best initiative can be maintained.
The national Programme for Promoting Success and Reducing Drop-out Rates has also fuelled a momentum for innovation in integrating advanced analytical tools to anticipate and customise student support at the institutional level. Some institutions are piloting emerging applications of predictive models in student support systems in Portugal, but challenges remain in identifying best practices and enabling more institutions to benefit from new digital tools.
Policy recommendations
1. Support higher education institutions to continue to provide and improve tutoring and remediation courses for students to develop subject-specific knowledge and transversal skills necessary for higher education programme completion.
2. Support institutions to strengthen their offering of mentoring and well-being programmes that have been shown to be effective in promoting student success.
3. Harmonise institutional-level collection and use of data on progress, drop-out and successful completion rates across programmes and higher education institutions, with the longer-term aim of consolidating the selection of active tracking models and rolling out good practice, along with identifying an alternative funding source when the EU’s Recovery and Resilience Facility ends in 2026.
4. Encourage the sharing of dedicated IT services across institutions in order to support institutions to develop and maintain their tracking tools, while reducing the duplication of work across institutions.
Many interlinked factors are relevant to understand why students drop out. International research has shown that factors like student characteristics and background, as well as their social and academic integration, the content and delivery of courses, and labour market conditions can be influential (Aina et al., 2022[4]). In Portugal, students are slightly more likely to drop out after the first year than during it, and there is considerable variation across courses. At the same time, a relatively large proportion of students feel that they do not belong in higher education. It is therefore promising that the government has funded a range of initiatives to support the academic and social integration of students through tutoring and mentoring.
6.2.1. When students drop out of higher education in Portugal, it can be a relatively long time after first enrolling
In Portugal, the drop-out rate after the first year of study is slightly lower than the average across other OECD jurisdictions for which data exist. Just 8% of full-time bachelor’s students were reported as no longer enrolled in tertiary education at the beginning of the second year in Portugal, compared to 12% on average across the OECD countries for which true cohort data ending in 2020 are available. The drop-out rate after the first year of studies was slightly higher in Portugal than its European neighbour Spain (7%) and on a par with that in Finland (8%) and Switzerland (8%) (Figure 6.1).
In the first year, students have a chance to find out whether the programme and institution they have chosen suits them. Before arriving in higher education, students often lack complete information about their chosen study programme, including the difficulty of courses, the effort needed to succeed, their fit with the academic environment and whether the career prospects associated with their chosen course align with their goals. On arrival, students may re-assess their initial decision to enrol in a particular programme or in higher education at all. Similarly, it is important for students to enjoy their subject area of study and to feel that their experiences during their studies correspond adequately with the expectations they had before applying. In support of this, Ferrão and Almeida (2018[5]) find that students in Portugal who enrol in a field of study that is not their first preference are more likely to leave in the first year of the programme.
Drop-out rates after the first year of study in Portugal remain higher than in peer countries. The proportion of students who drop out after the first year of studies, but before the theoretical duration of the programme plus three years in Portugal is higher than the drop-out rate in the first year, at 12%. This is similar to the average across OECD jurisdictions (11%), Austria (11%) and Colombia (12%), but higher than in Finland (9%) and Switzerland (4%) (Figure 6.1).
Share of full-time bachelor’s students no longer enrolled in tertiary education, by timeframe after entry, true cohort data only (2020)
Notes: See Education at a Glance, 2022. Students who repeat or transfer to a different course/level/programme type/mode of study are also treated as having progressed. Note that the year of reference for the data (2020) corresponds to a period three years after the theoretical end of the programme (2017). The reference year for students' entry to study may differ depending on the duration of their programme.
Source: OECD (2022[6]), Table B5.2, Education at a Glance.
Academic challenges can differ along student careers and between study programmes
The previous section highlights that it is relevant to consider progression rates after the first year and later in student careers, even though much attention tends to be given to first-year drop-out rates. Over the course of students’ careers, there tend to be some critical elements that are required in order to complete a degree, such as particularly challenging exams or assignments. Identifying these elements and monitoring student success at these key moments in each academic year can help student support teams to pre-emptively address challenges. Therefore, support needs may be different depending on the stage of students’ progress through higher education programmes.
Data from Portugal also indicate that while completion rates differ somewhat across subjects in the first year of study, it is only in subsequent years that these differences become more pronounced (Figure 6.2). This may indicate that there are similarities across fields of study in the first year of study that support students’ initial integration processes. Subsequently, factors that differ across programmes, like contact with academic staff, the academic content itself and the organisation of programmes, may become more important.
Proportion of students remained enrolled in their programme after the first year of studies (Panel A) and the proportion of students who could have graduated who had graduated by 2021/22 by the theoretical duration of the programme plus one year of a typical bachelor’s programme (Panel B), by higher education programme (2018 cohort)
Note: Completion rates refer to the percentage of student entrants in 2018/19 who theoretically could have finished their programmes in 2021/22 (theoretical duration plus one year for a bachelor’s degree). The chart includes shortened study areas. The fully study area definitions are: Education; Health and welfare; Services; Social sciences, journalism and information; Natural sciences, mathematics and statistics; Arts and Humanities; Agriculture, forestry, fisheries and veterinary; Business, administration and law; Engineering, manufacturing and construction; Information and Communication Technologies.
Source: Bespoke tables from DGEEC, Prosseguimento de estudos no Ensino Superior 2015/2016 a 2021/2022.
There may also be differences in labour market returns to a completed degree compared to a partially completed degree. In certain fields – such as education, health and welfare – the labour market tends to rely on credentials rather than demonstrated skills, which would require students to graduate to be able to find a job in their field of studies. By comparison, in information and communication technologies, the labour market tends to value demonstrated skills. Thus, if students have acquired the necessary skillset, they may receive good job offers before completing their studies, raising the opportunity cost and lowering the expected marginal return of completing the programmes they enrolled in.
Social integration and good mental health can help students perform at their best
Looking beyond academic achievement, students’ sense of social integration, sense of belonging and mental health, can be key factors in student success (Zając et al., 2024[7]; Müller and Klein, 2023[8]). Stress and mental health issues, including emotional and mental burnout, can be major challenges for higher education students and negatively impact both their health and academic performance (Lipson and Eisenberg, 2018[9]; Wyatt and Oswalt, 2013[10]; Zając et al., 2024[7]; Turan et al., 2023[11]; Gómez-García et al., 2022[12]). High psychological distress has been associated with increased test anxiety, lower self-efficacy, poor time management and limited use of study resources (Brackney and Karabenick, 1995[13]).
Research by Casanova et al. (2021[14]) from Portugal finds that academic exhaustion is negatively correlated with satisfaction with education and positively correlated with intention to drop out. The authors also find that the intention to drop out was in turn negatively correlated with satisfaction with education. The authors highlight that these findings reflect a dynamic and continuous process of adjustment and persistence, where students struggling to manage challenges and adversity could experience lower levels of self-efficacy and more feelings of isolation and stress, affecting their well-being (Casanova et al., 2021[14]).
Issues related to mental health likely became even more pronounced during the challenging periods of self-isolation and social distancing enforced by many governments to combat the spread of COVID-19, including lockdowns, shutdown of student dormitories, social distancing, and online learning (Abraham et al., 2024[15]).
Participants in focus groups held within the scope of this project report that some students worry about whether higher education is “for them”. While many battle with fears of not fitting in, it may be especially challenging for students whose friends and family have not attended higher education. According to Eurostudent data, around one-in-four students (24%) in Portugal report totally agreeing or agreeing that they often have a feeling that they do not really belong in higher education (Figure 6.3). This proportion is in line with that in Czechia (24%), but higher than in countries like Germany (13%) and Finland (11%) (Eurostudent, 2024[16]). It is therefore promising that Portugal has invested in increased mentoring support in institutions across the country.
Share of students agreeing or disagreeing with the statement "I often have the feeling that I don't really belong in higher education", 2024
Note: The surveys were conducted over the time period 2020-2024. Data refer to 2023 in Spain, Portugal; 2022 in Czechia, Denmark, Estonia, Finland, Hungary, Ireland, Iceland, Lithuania, Latvia, the Netherlands, Norway, Poland, Sweden, Slovak Republic; 2021 in Germany, and 2020 in Switzerland.
Source: Eurostudent, (2024[16]), The eighth Eurostudent round (2021-2024).
6.2.2. National funding has encouraged institutions to keep investing in initiatives that aim to reduce higher education drop-out
Higher education institutions in Portugal have several interlinked policies and initiatives in place to support student success, which build on pre-existing initiatives by institutions to promote retention and stimulate the development of new programmes. The national Programme for Promoting Success and Reducing Drop-out Rates in Higher Education in Portugal has stimulated the development and strengthening of initiatives to support students (Box 6.1). The programme funding has been allocated to initiatives in four key general programme areas: welcome and integration programmes for new students; mentoring and tutoring activities for well-being and academic success; the use of digital tools (developing methods for monitoring student success and drop-out risk); and initiatives to promote pedagogical innovation, including teacher training sessions.
There is now a range of ongoing initiatives to provide tutoring and mentoring at higher education institutions in Portugal, delivered primarily via the social support teams within the institutions’ Social Action Services (Serviços de Ação Social, SAS) in collaboration with teaching staff. Most higher education institutions in Portugal have implemented welcome and integration initiatives led by alumni and staff (that go beyond the traditional but controversial integration activities involved in the student-led “praxe” initiations for first-year students), as well as peer-mentoring programmes for their first-year students.
High-quality tutoring to support students struggling to complete higher education programmes
International evidence from several countries suggests that prior achievement is a key predictor of first-year success in higher education, for instance in Estonia (Silva et al., 2022[17]), Italy (Aina, 2011[18]), and Germany (Danilowicz-Gösele et al., 2017[19]). To help incoming student cohorts with variations in their academic and practical preparedness, many institutions in Portugal have invested in peer- and staff-led academic support and tutoring, delivered through the institutional social support teams. The development of further support programmes has been encouraged by the national “Programme for Promoting Success and Reducing Drop-out Rates in Higher Education in Portugal”. Given the plurality and diversity of active programmes, there is a unique opportunity in Portugal to draw lessons from the initiatives that have worked.
Tutoring programmes target current students and aim to support the development of study skills and subject-specific knowledge (Mateus, 2023[20]). For instance, the Polytechnic Institute of Viana do Castelo has implemented the programme “Contigo” (“With you”), which includes tutoring with a student-centred focus. The tutoring promotes both study skills via the Study Skills Development Programme and online courses and e-books for subject-specific knowledge in physics and mathematics. The activities include tutorials, mediation, training in transversal skills, as well as participation in projects and networks. The institution also provides monitoring reports on these activities.
Mentoring programmes in higher education promote student well-being and sense of belonging
Most social support teams in higher education institutions in Portugal also run programmes to support student success through peer mentoring. Peer-mentoring programmes typically involve older student volunteers who take part in more informal activities with first-year students, for example showing them around campus and attending workshops together. Programmes tend to be run under the guidance and supervision of staff members and sometimes involve training and some incentives for student volunteer mentors (Mateus, 2023[20]). For example, the University of Porto has constructed a new portal for peer-mentoring activities within the scope of the Programme for Promoting Success and Reducing Drop-out Rates in Higher Education in Portugal, “Mentoria U.Porto”. The scheme produced guides and training references that are accessible via a dedicated portal. Going forward, the University of Porto aims to target integration initiatives especially to students who live independently by providing workshops on cooking, managing personal finances and other essential skills of managing an independent household.
The Programme for Promoting Success and Reducing Drop-out Rates (Programa de Promoção de Sucesso e Redução de Abandono) is a multi-year national programme that provides funding to support the implementation of initiatives in higher education institutions that support student completion rates. Higher education institutions apply for funding for specific initiatives and selection decisions are made by an evaluation panel.
Funding comes from the Human Capital Operational Programme (POCH) (from the European Social Fund (ESF)) for convergence areas, the Portuguese State Budget, and the More Digital Impulse Programme, funded by the EU Recovery and Resilience Plan (RRP) (Government of Portugal, 2024[21]; Diário da República, 2024[22]). Funding is granted to initiatives that explicitly address the dual aim of raising preparedness and improving academic skills, including peer-mentoring schemes. The scheme specifically targets students enrolled in higher education for the first time in their first year.
The programme has been implemented in three phases. The first phase of the programme is funded by the Human Capital Operational Programme for institutions in the convergence regions. A total of EUR 6.6 million has been allocated to projects in 24 higher education institutions. The second phase of the programme is funded by the State Budget. It supports 20 projects in institutions located in the regions of Lisbon, Algarve, Azores, and Madeira to the total amount of EUR 3.7 million (Government of Portugal, 2024[21]; Diário da República, 2024[22]).
In the third phase, an additional EUR 20 million are also planned to finance the Programme through investment in the programme More Digital Impulse, funded by the RRP. An objective for the investment in the More Digital Impulse Programme is to modernise pedagogical practices in higher education, with a view to promoting academic success, reducing drop-out rates and student well-being. Out of the EUR 20 million, four million are earmarked for the design or purchase of IT systems for predicting school drop-out risk (Government of Portugal, 2024[21]; Diário da República, 2024[22]).
There have been some initial efforts to draw lessons from ongoing initiatives
There is value in encouraging a plurality of approaches among higher education institutions, at least in an initial phase of development, since it allows institutions greater freedom to experiment with different initiatives and methods compared to a situation where there were strict guidelines to follow in place.
However, given that projects are largely funded with resources from temporary EU funding programmes, there are important questions about the possibilities to ensure that successful initiatives remain in place in the longer term. If it is deemed that there will be a need for consolidating projects and streamlining budgetary commitments in the medium term, it is essential to understand the success of individual programmes, and whether there are ways to cut costs in line with smaller budgets. Cross-institutional collaborations could be useful to streamline practices and minimise costs and roll out best practice to ensure equity in support across institutions. It is therefore in the interest of institutions to ensure that the pilot initiatives are thoroughly evaluated, and that evaluations are shared and understood in order to scale up good practice.
There have already been some efforts to draw lessons from institutions’ experiences. For example, throughout the Programme for Promoting Success and Reducing Drop-out Rates in Higher Education in Portugal, participating institutions have met during conferences to share information about their initiatives and their implementation process. Previously, an extensive review of ongoing initiatives was undertaken by Mateus (2023[20]). The Portuguese Foundation for Science and Technology (2015[23]) went further in that they aimed both to take stock of activities and also identify best practices for welcoming first-year students to higher education. The report finds evidence that welcome programmes are used by higher education institutions, and that many survey respondents participate in these activities and many report finding some activities helpful for their academic experience (Portuguese Foundation for Science and Technology, 2015[23]).
Further efforts to draw lessons from existing experience could be made. This could help to support institutions to improve their offering of remediation courses for students to develop subject-specific knowledge and transversal skills and to strengthen their offer of mentoring and well-being programmes that have been shown to be effective in promoting student success by adopting good practice when the EU’s Recovery and Resilience Facility ends in 2026.
Given the development of new initiatives and the end of the present round of funding, there may be scope for collating information on what has worked so far. Since these initiatives are institution-based, the first step could be to ensure that institutions conduct evaluations of their programmes and that they produce information about initiatives and findings from evaluations. It is therefore promising that many institutions in Portugal are reportedly already conducting evaluations of their programmes, although these are yet to be shared. Further inspiration could also be drawn from international examples of designing and delivering programmes with a clear evaluation strategy. For example, the EU-funded ENTRANTS project, delivered in Austria, Belgium, Germany and the United Kingdom between 2020 and 2023 took an approach that built in the possibility for evaluations by including the dissemination of a baseline survey to assess student needs and a self-assessment of overall resilience before delivering programmes aimed to improve the social integration of new student cohorts (European Commission, 2023[24]) (Box 6.2).
A national review could collect information about the existing programmes and analyse the results. Keeping in mind that the dissemination of this information among stakeholders is key, it could be useful to build an “Evidence Hub” that could be accessible online. The national review could also help identify good practice and play a role in drafting guidelines and requirements for future funding for tutoring programmes, while also ensuring that evaluations keep a good-practice standard, for example playing a role like TASO in the United Kingdom (TASO[25]) (Box 6.2). In Portugal, it is possible that the newly formed National Council for Pedagogical Innovation in Higher Education (Conselho Nacional para a Inovação Pedagógica no Ensino Superior, CNIPES) could play a role in the co-ordination and dissemination of evaluations, and the development and refinement of best practice (DGES, 2025[26]).
The EU-funded ENTRANTS project, delivered in the United Kingdom, Belgium, Austria and Germany between 2020 and 2023 focused on helping new higher education cohorts adapt to student life. The project targeted first-year students who encounter entry-level difficulties and mismatches between expectations and reality. The aim was to address these challenges before they become overwhelming and result in students dropping out. The project involved several activities, including disseminating a survey to assess student needs, delivering an online student experience platform which included a self-assessment of overall resilience, an obligatory community building course incorporating strength assessments and escape games to foster exchange, and staff training for academic and non-academic staff (European Commission, 2023[24]).
Transforming access and outcomes for students (TASO), in the United Kingdom, is an independent hub for the higher education sector founded in 2019 through a consortium of universities and the Behavioural Insights Team. An independent charity since 2021, it is publicly funded by the Office for Students and an affiliate of the What Works Centre, which is part of the United Kingdom Government’s What Works Movement that aims to drive evidence-based policy development by conducting evaluations and working with the sector to support and advice institutions to conduct their own evaluations. TASO provides evidence and resources to help ensure that everyone has the opportunity to access, succeed and thrive in higher education. Through the newly launched Higher Education Evaluation Library (HEEL), TASO aims to help in the dissemination of best practice by bringing together evaluations on access, participation, and student success interventions (TASO[25]).
Key finding:
Most higher education institutions in Portugal have used short-term funding from the EU’s Recovery and Resilience Facility (RRF) funding to increase the offer of tutoring and mentoring services to their students, but it is unclear whether these programmes will continue as the funding stream ends. Questions remain on which initiatives work well and warrant scaling up, which risks limiting the potential of these initiatives to promote student completion rates widely.
Recommendations:
1. Support institutions to improve remediation courses for students to develop subject-specific knowledge and transversal skills necessary for higher education programme completion by adopting good practice, identifying alternative funding source when the EU’s Recovery and Resilience Facility ends in 2026.
2. Support institutions to strengthen their offer of mentoring and well-being programmes that have been shown to be effective in promoting student success, identifying alternative funding source when the EU’s Recovery and Resilience Facility ends in 2026.
Academic researchers in Portugal have started to investigate the possibilities of using student tracking analytics to detect the risk of drop-out, and recent investment through the national Programme for Promoting Success and Reducing Drop-out Rates in Higher Education has spurred institutions to start applying findings from the research community in practice. This section reviews recent developments in the field in Portugal.
Before moving on to discuss the application of predictive models to promote student success, it is important to highlight that, although tracking systems can add to the toolbox of social support teams in higher education institutions to help identify students in need of support and personalise the kind of support offered, no such predictive system is fully accurate in all possible cases, as explored in detail in the accompanying student tracking report delivered to the Portuguese authorities as part of this project. They therefore work best as a complement to traditional support services, which should remain available to students who seek help through traditional means.
6.3.1. Institutions and researchers in Portugal have started leveraging detailed data to understand student success
Higher education institutions are increasingly implementing new ways to understand the reasons why certain students drop out, and what can be done to address their individual needs and support completion. Across OECD member countries, academic researchers and higher education institutions are working to develop predictive models to improve the understanding of student completion patterns and to support targeted interventions. This research has particularly benefited from the increased availability of data in education settings, not least institution-level data, to develop broader applications, including identifying students who are at risk of dropping out of study programmes (Sclater Niall, 2014[27]).
Institutions in Portugal have access to the detailed data required for predicting whether individual students are at risk of dropping out
Portugal is in a good position to develop targeted, data-led student support systems since higher education institutions can leverage the datasets already in use for providing services and administrative functions. These rich and nuanced administrative data help enable the development of prediction models capable of accurately identifying subsets of students at risk. The main data sources are detailed below (see also de Oliveira et al (2021[28])).
Information about incoming students
Attributes such as secondary education leaving grades, age, and other demographic factors provide immediate predictors for a general assessment of the risk of dropping out. These can then be integrated with academic performance measures from higher education to adapt to students’ study trajectories over time. Information about background and socio-economic factors is collected upon enrolment in Portugal and serves as a foundation for providing initial feedback on student risk of dropping out, which can be used by social support teams or academic staff. While it can be useful for predicting both early academic success and the risk of drop out, this information is static and does not evolve with a student's career trajectory.
Higher education course performance
Students’ progress in successfully passing study credits offers reliable predictive power regarding drop-out rates and exam success. It acts as a real-time indicator for ongoing proficiency monitoring, adapting to each student's behaviour and providing insights into their specific engagement with their study programme. Subjective information, like post-matriculation follow up surveys about students’ social and academic engagement, and mental health conditions, can complete the data. Real-time data, such as logs from access to online services, can also serve as a valuable resource for identifying at-risk behaviours during teaching sessions.
Records of successful graduates
Datasets using records of successful graduates can illustrate how incoming and current students align with profiles that achieved study success in the past. It helps identify potential graduation challenges, such as differences in background, study habits, and proficiency levels, as well as key challenging elements, such as certain exams.
Academics in Portugal have been experimenting with different models of understanding student drop out and predicting student success
The understanding of learning analytics within academia has expanded significantly over time (Prieto et al., 2019[29]), and models used by academics internationally now incorporate a range of methodologies, such as descriptive, predictive and prescriptive analytics (Berland, Baker and Blikstein, 2014[30]). Academics and institutions in Portugal have been experimenting with different models to understand the factors that are important in student drop out, and to predict the chance of student success.
For example, a study from the University of Porto focuses on understanding the key factors that are the most important in predicting student success. The authors use institution-level data from enrolments between 2012 to 2019 and a relatively large sample of 50 000 students from the University of Porto. The study finds that the top five attributes that were the best at predicting drop-out risk were the number of delayed courses, percentage of programme completion, number of courses already enrolled, number of delayed years and number of ECTS credits to which the student has committed for the current semester (Belokurows, 2021[31]).
Many studies to date focus on predicting the study success of individual students. For example, a Portuguese study on students who transfer courses, also at the University of Porto, uses static data on their incoming cohort to estimates the length of time students are expected to take to complete their degree. The authors use information about prior achievement by considering the selectivity of the previous programme and machine learning models to obtain estimates of the time to degree (Pêgo, Miguéis and Soeiro, 2024[32]).
Several studies use institution-level data to identify the machine learning techniques that can identify students at risk of dropping out with the highest accuracy. Examples of research teams developing predictive models include a Portuguese study based on a sample of 2 934 students from the University of Évora. The researchers propose a method to identify the drop-out risk of students based on academic performance. The model exploits student academic record data from four different programmes, includes a wide range of institution-level information and evaluates four different machine learning techniques. They find that the best model reaches an accuracy of around 96% when distinguishing between students at risk of dropping out and students not at risk of dropping out (Prite, Gonçalves and Rato, 2020[33]).
6.3.2. Emerging applications of predictive models to promote student success in Portugal
Building on the academic research on identifying the chances of student success and the risk of drop out, efforts are increasingly being made to apply the findings in the implementation of student support systems. Several institutions in Portugal are in the process of piloting integrated analytical tools in their support activities to promote student success, drawing on the funding from the national Programme for Promoting Success and Reducing Drop-out Rates in Higher Education.
Analytical models can help identify students who may benefit from better integration in academic or social life and address these issues through traditional interventions tailored to the nature of the difficulties and the type of at-risk student, such as tutoring, mentoring and psychological support. Interventions based on findings from the models could be tailored to individual needs and identified in at least three key phases of student trajectories: a) at the time of enrolment and during the first term; b) after the first term and within the first academic year; and c) at year-end in all years following the first. It is possible to implement individual support policies to reduce drop-out rates based on critical elements in each phase, considering the available financial resources and human capital at each higher education institution.
In Portugal, there have been at least three different forms of applications of advanced analytics in institutions so far: integration of findings into traditional support activities; integration of findings into virtual support activities; and integration of findings into evaluations of support activities.
Integration of findings from analytical models into traditional (human-based) support activities
Based on findings from a risk assessment, students may be personally invited to participate in the existing offer of counselling, mentoring, and/or tutoring activities, depending on the identified needs. These may include targeted courses to address specific knowledge gaps, mentoring activities with students enrolled in subsequent years, and counselling services aimed at promoting engagement and social and academic integration.
For example, the Polytechnic Institute of Bragança has developed a system of monitoring student success using an online platform that includes both descriptive statistical elements and an analysis of predicted risk of dropping out which uses machine learning tools that generate real-time feedback for immediate action. Students who are flagged by the algorithm as being at risk of dropping out have access to a dedicated telephone-based support. Student profiles that are identified as needing pedagogical support also receive targeted support from teachers and students in the curricular units that have been identified as challenging. Such targeted tutoring is available in all areas of basic training, including mathematics, physics, chemistry, biology and Portuguese and is complemented by online courses in certain subjects (mathematics, chemistry and biology) (Pacheco, 2024[34]).
Similarly, the Polytechnic Institute of Portalegre links information from different data sources including demographic, socio-economic, macro-economic, and academic data on enrolment, and academic performance at the end of the first and second terms. The dataset is used to build machine learning models for predicting academic performance and drop-out risk (Cabezuelo et al., 2022[35]). The models are integrated into a Learning Analytic tool that is applied to students enrolled for the first time and provides information to the tutoring team and is used as a tool to help select incoming students invited to participate as mentees in mentoring activities. The institution foresees a continual development of the models, with annual updates and validation either with the new information of incoming students or with the information of the final situation of the already enrolled students (Martins et al., 2023[36]).
Integration of findings from analytical models into virtual support activities
Risk assessments using analytical models can also help direct students to targeted online or virtual support. This could allow some of the students who would have contacted the staff to find the help they need more quickly. This way, staff support teams can spend more time working with the students who need more complex forms of support. This idea of sharing the case load between the staff and online tools can help make the existing support offer more targeted and efficient.
For example, the University of Coimbra has implemented and tested a solution of adaptive learning and virtual tutoring using a chatbot (Albuquerque, 2024[37]). Similarly, the private university Autonomous University of Lisbon (UAL) platform will integrate a chatbot including a virtual tutoring solution that uses natural language processing and artificial intelligence. Through this system, students who are flagged as being at risk of dropping out will be issued with either mentoring to support students who lack psychological guidance or tutoring to students who need support in technical-scientific learning (Donário, 2024[38]).
Integration of findings from analytical models into evaluations of support activities
The higher the quality of the services and facilities available to students, the more effective the results of drop-out prevention efforts will be. Monitoring the individuals involved in these activities can be useful for measuring the effectiveness of the implemented measures, particularly regarding those who, despite being identified as at risk, did not participate. Evaluation can be readily conducted by analysing the outcomes achieved at the end of the first examination session.
For example, Nova University of Lisbon has also developed systems that identify students at risk of dropping out using predictive elements in the analysis of data on academic, personal, socio-economic and behavioural factors, partly with the aim to inform holistic policy development (Henriques and Xufre, 2024[39]). In a similar vein, the platform for monitoring student success and predicting risks of dropping out developed by UAL is foreseen as a tool to measure the impact of actions taken to address elevated drop-out risks in students by estimating the effects from adopting different responses. As such, it can help identify interventions with the highest effects (Donário, 2024[38]).
6.3.3. Scaling up successful initiatives related to student progress and success in Portugal
A predictive model could be built upon a core set of common features while remaining adaptable to specific static and dynamic information relevant to individual study programmes. There are therefore opportunities to scale up successful pilots across multiple institutions. As expected during a piloting phase, there are currently considerable differences in the development and use of advanced analytical models across institutions in Portugal. This suggests that now is a good time for cross-institutional learning and sharing of knowledge, models and applications.
While this chapter mentions a few studies and initiatives drawing on published papers and presentations made by institutions as part of the Programme for Promoting Success and Reducing Drop-out Rates in Higher Education in the event series “Sucesso académico e prevenção do abandono no ensino superior”, there has seemingly not (yet) been a systematic collection of evidence on institution-based initiatives undertaken in this field, nor of evaluations of these initiatives.
While it is valuable to fund a number of institution-based initiatives in a relatively new policy field to stimulate innovation, the long-term value of innovation hinges on the ability to harness results and identify good practice that can be rolled out widely. A long-term aim could therefore be to consolidate the selection of active tracking models and encourage the adoption of good practice when the EU’s Recovery and Resilience Facility ends in 2026.
To this aim, information from ongoing pilot initiatives would need to be shared to understand what works what policies could be adopted across institutions. This would be important for institutions that have not yet implemented a student tracking system. These institutions could learn from others and, for example, harmonise institutional-level data collection and use on progress, drop-out and successful completion across programmes and higher education institutions in order to better be able to adopt existing tracking models. Such a harmonised use of tracking models could enhance the overall cost-efficiency at the national level and help ensure equity of access to digital tools across institutions.
While institutions are already able to spontaneously collaborate in this field, for example by the use of the forums created by the higher education consortia and the centres of excellence in pedagogical innovation, it appears that cross-institutional sharing of good practice is limited. As the funding for the Programme for Promoting Success and Reducing Drop-out Rates in Higher Education is due to end, the Ministry convened the National Council for Pedagogical Innovation in Higher Education (CNIPES) in 2024. As an independent consultative body on matters of pedagogical innovation and training, it could play a role in the co-ordination of a stocktaking exercise of initiatives and collection of results from evaluations (DGES, 2025[26]). It is also possible that Portugal’s National Research and Education Network, FCCN, as well as private digital services providers have roles to play in the scaling up of successful initiatives. International examples of where governments have played a role in simplifying the procurement of digital services in the education sector, including in Ireland and the Netherlands (Box 6.3).
The SURF organisation in the Netherlands is an IT co-operative that connects institutions and promotes a collaborative organisation for network and computer services in education and research. It runs a host of ICT services, including identity and access management, procurement and delivery of IT services and content, online security and network and connectivity services. SURF helps its members to deliver ICT and data solutions by managing software licencing, content services and tenders on behalf of its members. It is also a meeting place where members work together on innovative solutions in a range of ICT areas such as cybersecurity, study data and artificial intelligence (SURF[40]).
Ireland offers an example of how to facilitate the procurement of digital services from the private market. Given the autonomy afforded to educational institutions in Ireland, procurement of digital educational tools and resources can be performed directly by the central government, by education and training boards managing multiple schools, or by schools themselves. Procurement processes for the government are performed through the Office of Government Procurement (OGP), a division under the Department of Public Expenditure and Reform. Framework contracts negotiated by the OGP, by the Department of Education, and by the Higher Education Authority Network (HEAnet) are in place with ICT equipment suppliers, resulting in an approved selection of suppliers that are recommended for schools to procure. In addition, schools and boards can profit from brokerage services in order to negotiate prices with suppliers available by partnering with organisations with expertise on the ICT procurement market and processes such as the government funded HEAnet. Furthermore, the Department of Education has also established single provider framework contracts to provide schools with a variety of ICT equipment (OECD, 2023[41]).
To facilitate the adoption of good practice, there could be benefits in encouraging the sharing of dedicated IT services across institutions in order to support institutions to develop and maintain their tracking systems, while reducing the duplication of work. This will likely be essential as the funding available for such initiatives risk shrinking significantly by 2026 when the EU’s Recovery and Resilience Facility ends. Institutions may benefit from opportunities to harmonise data collection on progress and exit across institutions, particularly as institutions may be in the process of investing in their data infrastructure and governance processes. This could contribute to facilitating the sharing of models and systems across institutions.
Many countries are working towards streamlining and centralising data to facilitate data linking and research access and creating infrastructures that ensure the protection of personal data. This can help cutting-edge innovation and also facilitate the implementation of systems and models across institutions, including those with fewer resources to develop tools internally. For example, Ireland has developed a data plan for equity that encourages the harnessing of new opportunities to link data from different sources. Elsewhere – including in Finland and Lithuania – the government has acted on these opportunities to invest in data infrastructure and management systems that aim to support data-driven research and policy implementation in educational institutions (Box 6.4)
The Data Plan for Equity of Access to Higher Education in Ireland, developed by Trutz Haase and Jonathan Pratschke, aims to improve the capacity of the Higher Education Authority (HEA) to measure and monitor equity of access. The data plan exploits the possibilities created by new technologies and administrative databases and expanded the scope for research on educational inequalities in Ireland. It relies on linking and geocoding datasets, whereby information from different sources is brought together either at the level of the individual or for Small Areas of residence (Haase and Pratschke, 2017[42]). Building on this, the HEA has committed to further closing data gaps by developing a new Data Plan in their Strategic Action Plan for Equity of Access, Participation and Success in Higher Education 2022-28 (HEA, 2022[43]).
For instance, the Research Information Hub is a national service of the Finnish Ministry of Education and Culture that gathers and shares information on scientific research carried out in Finland in an easily accessible way and format. The Research Information Hub includes the Virta higher education achievement register, launched in the early 2010s and developed from the study information systems of Finnish higher education institutions. The development of the Virta register created a natural platform for harmonising and improving the quality of data in higher education study information systems and has several additional purposes. The Virta register handles several data transfers to authorities and other actors that higher education institutions are obliged to carry out and allows the Ministry of Education and Culture the flexibility to produce a variety of statistics on higher education. It is also used for student selection in higher education institutions, where, from 2014 onwards, applicants could be allocated to quotas according to whether they had a previous higher education entrance qualification (Haapamäki, 2024[44]).
In Lithuania, the central government’s student information system, ŠVIS (Švietimo Valdymo Informacinė), is the cornerstone of the public digital infrastructure for system management. Although ŠVIS is built from IBM Cognus, a commercial tool, it is publicly owned by the government and the data are stored on the ministry’s servers. ŠVIS exchanges data with databases from the student register system that contain statistical information related to all levels of education, including higher education, with data about schools, teachers, and students, but can also be linked with data from other central registers, including the health system register and the social insurance system register. Teacher and student data are pseudonymised: individuals are linked to their national personal ID number, which is unique, longitudinal, and confidential, but different from their personal educational ID. The system stores students’ standardised assessment results, as well as teacher-given grades in upper secondary education and VET (from non-standardised exams). ŠVIS is updated in real time so that authorised users, be they administrators, school principals, or teachers, have access to analytics dashboards quickly after the information is collected (OECD, 2023[41]).
Key finding:
The EU’s Recovery and Resilience Facility (RRF) funding has fuelled a growing momentum for innovation in integrating advanced analytical tools to anticipate and customise improve student support at the institutional level, but challenges remain in identifying best practices and enabling more institutions to benefit from new digital tools, in order to capitalise on the ongoing innovation to promote completion rates.
Recommendations:
3. Harmonise the collection and use of data on progress, drop-out, and successful completion rates across programmes and higher education institutions at an institutional level, with the longer-term aim of consolidating the selection of active tracking models and encouraging the adoption of good practice, along with identifying alternative funding source when the EU’s Recovery and Resilience Facility ends in 2026.
4. Encourage the sharing of dedicated IT services across institutions in order to support institutions to develop and maintain their tracking tools, while reducing the duplication of work.
References
[15] Abraham, A. et al. (2024), “Burnout increased among university students during the COVID-19 pandemic: a systematic review and meta-analysis”, Scientific Reports, Vol. 14/1, https://doi.org/10.1038/s41598-024-52923-6.
[18] Aina, C. (2011), “The determinants of success and failure of italian university students. Evidence from administrative data”, Rivista Internazionale di Scienze Sociali 2.
[4] Aina, C. et al. (2022), “The determinants of university dropout: A review of the socio-economic literature”, Socio-Economic Planning Sciences, Vol. 79, https://doi.org/10.1016/j.seps.2021.101102.
[37] Albuquerque, C. (2024), ON-BOARD - Programa integrado de tutotia para a prevenção do abandono e insucesso académico, https://www.dges.gov.pt/pt/noticia/iii-seminario-sucesso-academico-e-prevencao-do-abandono-no-ensino-superior-25-de-marco-2024.
[31] Belokurows, R. (2021), A Predictive Analysis of Academic Success at Universidade do Porto Orientador, https://repositorio-aberto.up.pt/bitstream/10216/137753/2/515061.pdf (accessed on 5 February 2025).
[30] Berland, M., R. Baker and P. Blikstein (2014), “Educational Data Mining and Learning Analytics: Applications to Constructionist Research”, Technology, Knowledge and Learning, Vol. 19/1-2, pp. 205-220, https://doi.org/10.1007/s10758-014-9223-7.
[1] Bound, J. and S. Turner (2011), “Dropouts and Diplomas”, https://doi.org/10.1016/B978-0-444-53444-6.00008-0.
[13] Brackney, B. and S. Karabenick (1995), “Psychopathology and Academic Performance: The Role of Motivation and Learning Strategies”, Journal of Counseling Psychology, Vol. 42/4, pp. 456-465, https://doi.org/10.1037/0022-0167.42.4.456.
[35] Cabezuelo, S. et al. (2022), “Predicting Student Dropout and Academic Success”, Data Mining and Computational Intelligence for E-learning and Education), Vol. 7/11, https://doi.org/10.3390/DATA7110146.
[2] Carneiro, P., J. Heckman and E. Vytlacil (2011), “Estimating Marginal Returns to Education”, American Economic Review, Vol. 101/6, pp. 2754-2781, https://doi.org/10.1257/aer.101.6.2754.
[14] Casanova, J. et al. (2021), “Dimensionality and reliability of a screening instrument for students at-risk of dropping out from Higher Education”, Studies in Educational Evaluation, Vol. 68, https://doi.org/10.1016/J.STUEDUC.2020.100957.
[19] Danilowicz-Gösele, K. et al. (2017), “Determinants of students’ success at university”, Education Economics, Vol. 25/5, pp. 513-532, https://doi.org/10.1080/09645292.2017.1305329.
[28] de Oliveira, C. et al. (2021), “How Does Learning Analytics Contribute to Prevent Students’ Dropout in Higher Education: A Systematic Literature Review”, Big Data and Cognitive Computing, Vol. 5/4:64, https://doi.org/10.3390/bdcc5040064.
[26] DGES (2025), MECI Dec.24 | Order No. 15261-H/2024 - Designation of CNIPES members, https://www.dges.gov.pt/pt/noticia/meci-dez24-despacho-no-15261-h2024-designacao-de-membros-do-conselho-nacional-para-inovacao.
[22] Diário da República (2024), Despacho n.º 676/2024, de 22 de janeiro, https://diariodarepublica.pt/dr/detalhe/despacho/676-2024-838325383.
[38] Donário, G. (2024), Projeto ambição, https://www.dges.gov.pt/pt/noticia/iii-seminario-sucesso-academico-e-prevencao-do-abandono-no-ensino-superior-25-de-marco-2024.
[24] European Commission (2023), Enhancing the transition of non-traditional students, https://erasmus-plus.ec.europa.eu/projects/search/details/2020-1-AT01-KA203-078052.
[16] Eurostudent (2024), Eurostudent database, https://database.eurostudent.eu/drm/.
[5] Ferrão, M. and L. Almeida (2018), “Multilevel Modelling of Persistence in Higher Education”, Avaliação e Políticas Públicas em Educação, Vol. 26/100, pp. 664-683, https://doi.org/10.1590/S0104-40362018002601610.
[45] Giani, M., P. Attewell and D. Walling (2020), “The Value of an Incomplete Degree: Heterogeneity in the Labor Market Benefits of College Non-Completion”, The Journal of Higher Education, Vol. 91/4, pp. 514-539, https://doi.org/10.1080/00221546.2019.1653122.
[12] Gómez-García, G. et al. (2022), “Impact of COVID-19 on University Students: An Analysis of Its Influence on Psychological and Academic Factors”, International Journal of Environmental Research and Public Health, Vol. 19/16, p. 10433, https://doi.org/10.3390/ijerph191610433.
[21] Government of Portugal (2024), 10 milhões de euros para reduzir insucesso escolar no Ensino Superior, https://www.portugal.gov.pt/pt/gc23/comunicacao/noticia?i=10-milhoes-de-euros-para-reduzir-insucesso-escolar-no-ensino-superior.
[3] Gunderson, M. and P. Oreopolous (2020), “Returns to education in developed countries”, in The Economics of Education, Elsevier, https://doi.org/10.1016/B978-0-12-815391-8.00003-3.
[44] Haapamäki, J. (2024), “Developing Research Information Hub in Finland, a view from the Ministry”, CRIS2024: 16th International Conference on Current Research Information Systems (Vienna, Austria, May 15-17, 2024), https://doi.org/10.1016/j.procs.2024.11.044.
[42] Haase, T. and J. Pratschke (2017), A Data Plan for Equity of Access to Higher Education | Publications | Higher Education Authority, HEA, https://hea.ie/resources/publications/a-data-plan-for-equity-of-access-to-higher-education/ (accessed on 2 June 2025).
[43] HEA (2022), National Access Plan: A Strategic Action Plan for Equity of Access, Participation and Success in Higher Education 2022-2028, https://hea.ie/policy/access-policy/national-access-plan-2022-2028/.
[39] Henriques, R. and P. Xufre (2024), STEPS@NOVA, https://www.dges.gov.pt/pt/noticia/iii-seminario-sucesso-academico-e-prevencao-do-abandono-no-ensino-superior-25-de-marco-2024.
[46] Jacobson, L., R. LaLonde and D. G. Sullivan (2005), “Estimating the returns to community college schooling for displaced workers”, Journal of Econometrics, Vol. 125/1-2, pp. 271-304, https://doi.org/10.1016/j.jeconom.2004.04.010.
[47] Jepsen, C., K. Troske and P. Coomes (2014), “The Labor-Market Returns to Community College Degrees, Diplomas, and Certificates”, Journal of Labor Economics, Vol. 32/1, pp. 95-121, https://doi.org/10.1086/671809.
[48] Kane, T. and C. Rouse (1999), “The Community College: Educating Students at the Margin Between College and Work”, Journal of Economic Perspectives, Vol. 13/1, pp. 63-84, https://doi.org/10.1257/jep.13.1.63.
[9] Lipson, S. and D. Eisenberg (2018), “Mental health and academic attitudes and expectations in university populations: results from the healthy minds study”, Journal of Mental Health, Vol. 27/3, pp. 205-213, https://doi.org/10.1080/09638237.2017.1417567.
[36] Martins, M. et al. (2023), “Multi-Class Phased Prediction of Academic Performance and Dropout in Higher Education”, Applied Sciences, Vol. 13/8, p. 4702, https://doi.org/10.3390/APP13084702.
[20] Mateus, A. (2023), Sucesso e Inclusão: Medidas e Iniciativas de Apoio aos estudantes do ensino superior, DGES.
[8] Müller, L. and D. Klein (2023), “Social Inequality in Dropout from Higher Education in Germany. Towards Combining the Student Integration Model and Rational Choice Theory”, Research in Higher Education, Vol. 64/2, pp. 300-330, https://doi.org/10.1007/s11162-022-09703-w.
[41] OECD (2023), OECD Digital Education Outlook 2023: Towards an Effective Digital Education Ecosystem, OECD Publishing, Paris, https://doi.org/10.1787/c74f03de-en.
[6] OECD (2022), Education at a Glance 2022: OECD Indicators, OECD Publishing, Paris, https://doi.org/10.1787/3197152b-en.
[34] Pacheco, M. (2024), Drop-In@IPB, https://www.dges.gov.pt/pt/noticia/iii-seminario-sucesso-academico-e-prevencao-do-abandono-no-ensino-superior-25-de-marco-2024.
[32] Pêgo, J., V. Miguéis and A. Soeiro (2024), “Students’ complex trajectories: exploring degree change and time to degree”, International Journal of Educational Technology in Higher Education, Vol. 21/1, pp. 1-29, https://doi.org/10.1186/S41239-024-00438-5/FIGURES/13.
[23] Portuguese Foundation for Science and Technology (2015), Acolhimento nas Instituições de Ensino Superior Portuguesas, Fundação para a Ciência e Tecnologia, https://aepq.tecnico.ulisboa.pt/files/sites/22/relatorio-final-fct.pdf.
[29] Prieto, L. et al. (2019), “Orchestrating learning analytics (OrLA): Supporting inter-stakeholder communication about adoption of learning analytics at the classroom level”, Australasian Journal of Educational Technology, Vol. 35/4, https://doi.org/10.14742/ajet.4314.
[33] Prite, S., T. Gonçalves and L. Rato (2020), “Identifying Risky Dropout Student Profiles using Machine Learning Models”, Proceedings of the 26th Portuguese Conference on Pattern Recognition, RECPAD 2020, http://hdl.handle.net/10174/33879 (accessed on 5 February 2025).
[49] Schnepf, S. (2015), “University dropouts and labor market success”, IZA World of Labor, https://doi.org/10.15185/izawol.182.
[27] Sclater Niall (2014), Code of practice “essential” for learning analytics, https://analytics.jiscinvolve.org/wp/2014/09/18/code-of-practice-essential-for-learning-analytics/.
[17] Silva, P. et al. (2022), “High School and Exam Scores: Does Their Predictive Validity for Academic Performance Vary with Programme Selectivity?”, SSRN Electronic Journal, https://doi.org/10.2139/SSRN.4134133.
[40] SURF (n.d.), SURF, https://www.surf.nl/en.
[25] TASO (n.d.), Transforming access and outcomes for students, https://taso.org.uk/.
[11] Turan, J. et al. (2023), “Global health reciprocal innovation to address mental health and well-being: strategies used and lessons learnt”, BMJ Global Health, Vol. 8/Suppl 7, p. e013572, https://doi.org/10.1136/bmjgh-2023-013572.
[10] Wyatt, T. and S. Oswalt (2013), “Comparing Mental Health Issues Among Undergraduate and Graduate Students”, American Journal of Health Education, Vol. 44/2, pp. 96-107, https://doi.org/10.1080/19325037.2013.764248.
[7] Zając, T. et al. (2024), “Student mental health and dropout from higher education: an analysis of Australian administrative data”, Higher Education, Vol. 87/2, pp. 325-343, https://doi.org/10.1007/s10734-023-01009-9.
← 1. While expected returns are higher for higher education graduates, findings also indicate that some tertiary education may be better than none at all. Some evidence shows that students who drop out have higher earnings and better employment outcomes compared to secondary education graduates who never attended higher education (Giani, Attewell and Walling, 2020[45]; Jacobson, LaLonde and G. Sullivan, 2005[46]; Jepsen, Troske and Coomes, 2014[47]; Kane and Rouse, 1999[48]; Schnepf, 2015[49]).