Huge Study Reveals What Really Drives Students to Drop Out or Graduate

Every year, roughly one in three university students across OECD countries walks away from higher education without a degree. The personal, institutional and economic costs of that exodus are enormous, and nowhere in the developed world is the problem more acute than in Italy, which combines one of the lowest tertiary attainment rates among OECD members with unusually high dropout and delayed-completion rates. A new open-access study published in Social Indicators Research now offers one of the most detailed statistical portraits to date of how these careers actually unfold, drawing on the administrative records of more than 57,000 bachelor’s students at the University of Padova.

The research, conducted by Claudia Franceschini, Anna Giraldo and Silvia Meggiolaro of the university’s Department of Statistical Sciences, tracks every student enrolled in a first-cycle degree programme between the 2016/2017 and 2020/2021 academic years, following their trajectories until September 2024. That means each student was observed for at least four years, well beyond the official three-year duration of an Italian bachelor’s degree, a crucial design choice in a system where students face no strict deadline for finishing and often graduate years late.

What sets the study apart methodologically is its treatment of university careers as a set of competing risks. Rather than asking simply whether a student drops out, the authors model dropout and graduation as two rival events, each with its own instantaneous hazard, using cause-specific Cox proportional hazards models. Because the administrative records contain exact event dates rather than just academic-year labels, the models operate in continuous time, capturing whether a student withdrew in October or June, information that coarser yearly analyses would erase. Students still enrolled when the observation window closed were treated as right-censored, a standard survival-analysis technique that uses partial information without biasing the estimates.

A further refinement concerns course changes. Students who switch degree programmes within the same university are often counted as dropouts in cruder studies, but the authors treat a change of course as a time-dependent covariate that switches on at the moment of transfer. This allows the models to estimate how switching relates to the subsequent hazard of leaving or graduating, while the authors are careful to note that the association is descriptive rather than causal. After excluding strategic first-week enrolments, foreign diploma holders and second careers, the final dataset contained 57,097 student careers.

The covariates span the ‘micro level’ of dropout research: gender, age at enrolment, citizenship, place of residence, secondary school type and final grade, regularity of the school career, employment status, part-time status, disability and a precise household economic indicator known as ISEE, which combines income, assets and family size. Because about a third of students had missing ISEE information, the researchers created a separate missing category, later discovering that these students form a strikingly heterogeneous group that behaves in unexpected ways.

The headline finding concerns secondary school background. Graduates of vocational institutes faced the highest hazard of dropping out and the lowest hazard of completing a degree, with graduation hazard ratios around 0.58 to 0.66 relative to academic high school leavers, while students from technical schools fell in between. Each additional point on the secondary school final exam reduced the dropout hazard by roughly one to three percent and raised the graduation hazard by two to four percent across most fields. The one exception was professional health sciences, where entrance examinations select a highly motivated population whose prior schooling apparently matters far less.

Field of study proved so consequential that the authors estimated completely separate models for four disciplinary areas: scientific studies, humanities, professional health sciences and social sciences. Gender effects, for example, flipped direction depending on the outcome: in scientific studies men were significantly less likely than women to drop out, yet in the other three fields men showed a markedly lower hazard of graduating. Economic status was broadly protective, with students from higher-ISEE households less likely to withdraw and more likely to finish, except in the social sciences, where the income gradient on dropout disappeared.

Perhaps the most policy-relevant result involves switching courses. Contrary to the intuition that changing programmes helps struggling students find their footing, a course change was associated with a significantly lower hazard of graduating in every destination field, and it reduced the dropout hazard only for humanities students. In other words, for most students a switch delayed or jeopardised completion without preventing departure, a pattern the authors interpret as evidence that many students arrive with insufficient guidance and choose their initial programme on faulty premises.

Employment at enrolment told a similarly field-specific story. Holding a job substantially increased the dropout hazard among humanities and social sciences students and lowered the graduation hazard in the humanities and professional health sciences, confirming the corrosive effect of the work-study conflict, while showing no significant association in scientific or social science graduation. Part-time enrolment was rare overall and displayed the worst dropout trajectories, though graduation comparisons were regulated out of reach, since part-time students may take twice the standard time.

The authors are careful about limits. Because Italian administrative data cannot distinguish students who left higher education entirely from those who transferred to another university, ‘dropout’ here means formal departure from Padova without re-enrolment there. The pandemic overlapped several cohorts, so the models included cohort fixed effects and a time-dependent variable capturing the shift to fully online teaching in 2020-21 and hybrid teaching the following year. P-values were corrected for false discovery using the Benjamini-Hochberg procedure, and robust sandwich standard errors guarded against within-subject correlation.

The practical payoff is targeted intervention. Universities cannot change a student’s school background or family income, but they can identify the profiles at greatest risk, from vocational-track entrants to working students in the humanities, and direct tutoring, psychological support and financial aid accordingly. The results also make a strong case for better pre-enrolment guidance in Italian secondary schools, so that students arrive knowing what they are choosing. Since part of Italian public funding now depends on career progression, the incentive for institutions to act on these hazard curves has never been sharper.

Subject of Research: Competing-risks survival analysis of university dropout and degree completion among Italian bachelor's students

Article Title: Modelling University Career Paths: A Competing-Risks Analysis of Dropout and Degree Completion at the University of Padova

Article References: Franceschini, C., Giraldo, A., & Meggiolaro, S. (2026). Modelling University Career Paths: A Competing-Risks Analysis of Dropout and Degree Completion at the University of Padova. Social Indicators Research, 184(3), Article 52. https://doi.org/10.1007/s11205-026-03944-3

Image Credits: AI Generated

DOI: 10.1007/s11205-026-03944-3

Keywords: university dropout, graduation, survival analysis, competing risks, Cox proportional hazards, higher education, University of Padova, course change, socioeconomic status, secondary school background, administrative data, Italy