Drop-out decisions in a cohort of Italian universities

In this chapter, we study the determinants of student drop-out decisions
using data on a cohort of over 230,000 students enrolled in the Italian university
system. The empirical analysis reveals that the probability of dropping out of
university negatively correlates with high school grades and student age, controlling
for the course of study and university fixed effects. The benchmark estimation
suggests a negative correlation between high school final grade and drop-out
probability.We also find that enrolling late at the university increases the likelihood
of dropping out. In line with the literature, our results suggest that women have a
lower propensity to drop out. Our dataset allows differentiating between students
who leave their homes to enroll at university (off-site students) and on-site students.
We find that off-site students drop out significantly less than those who study in their
hometowns.We provide significant evidence that off-site students are a self-selected sample of the total population. Accordingly, we use an instrumental variable (IV)
approach to identify the causal relationship. The IV estimation shows that studying
off-site negatively affects drop-out decisions and more so for students growing up
in the south of Italy who typically study off-site in the Center-North of Italy. Taking
advantage of a more detailed dataset concerning students enrolled at the Università
di Sassari, we show that the choice of the degree is also important to predict the
magnitude of drop-out. Specifically, we resort to a bivariate probit specification
to account for self-selection into the course of study, finding that the estimates of
the determinants of drop-out and the predicted probabilities are heavily affected.
Accounting for self-selection, we show that an unconditional comparison among
degrees is misleading, as some degrees attract more heterogeneous students than
others, as far as skills and motivation are concerned. For instance, regarding the
effect of gender, we show that while the estimation without selection suggests that
women drop out less, once we account for selection, the contribution of women to
drop-out becomes either positive or negative, depending on which course of study
they choose. In line with these results, policymakers should tailor drop-out reducing
policy interventions to the specificities of each course of study.