Why do higher education students drop out? Evidence from Spain

Abstract

This paper seeks to advance our understanding of the drop‐out behavior of students in higher education. Our results are based on longitudinal data for 7000 students who embarked on short and long programs from one university in Spain and who were observed over an eight‐year period ending in 2004. The statistical analysis is carried out in a competing‐risks framework. We find that academic preparedness is one of the major influences on student completion. Additionally, older students and students who delay entry into higher education are more likely to drop out before graduating. Our analysis provides guidance about the role that financial support plays in reducing drop‐out rates; we also find that family characteristics are significant factors in explaining student drop out in long programs.

Keywords:

• higher education
• student drop‐out
• discrete‐time hazard

Acknowledgements

The research was funded in part through a research support grant from the Ministry of Education (No. SEC2003‐08855‐C03‐01). The opinions expressed in this paper are those of the authors alone and should not be attributed to the institutions with which they are associated.

Notes

1. For more details, see, for example, Postlethwaite (1995) or International Bureau of Education (2006).

2. Students who sit for the Prueba de Aptitud para la Universidad have completed upper‐secondary education and a one‐year period known as the Curso de Orientación Universitaria (university orientation course).

3. No information is available on drop out and repetition. Precise information on the distribution of students by grade is also lacking, and the distribution of graduates by year of entry into the program is generally poorly documented.

4. We consider here that the vast majority of programs in the faculty sector lasts for five years.

5. The University of Málaga enrolls more than 38 000 students; it is one of the eight largest institutions in the country. Although there is no precise information to compare the performance of students in Málaga, we consider that, overall, this institution is fairly representative of the Spanish higher education system. To illustrate, the graduates/students ratios in the faculty and university school sector are around 13% at Málaga, compared with 13% and 15%, respectively, at the system level (see Table 1). However, higher tecnical school graduates in Málaga tend to finish more rapidly their studies than their counterparts nationwide. We estimate that around 40% of graduates in this institution complete their degree within the intended time frame (Lassibille & Navarro Gómez, 2007); according to the rough information supplied by the Ministry of Education, this proportion is around 30% at the national level.

6. We roughly estimate that less than 2% of entering students stop out during our period of observation.

7. As pointed out by an anonymous referee, students who drop out during a given year may have different reasons for leaving depending on the exact date when their decision occurs. However, given the nature of our data, we are not able to identify such differentiated effects.

8. Among these students, we estimate that around one‐third switches to another program at the University of Málaga; we are not able to evaluate the proportion of students who transfer to another institution.

9. Normalized score at pre‐university examination = (X − M) / S, where X is the student's score on the pre‐university examination, M is the average score of students taking the same subject, and S is the standard deviation.

10. The amount of financial aid students received is not available in the data‐set.

11. The final data‐set contains 28 999 person‐period records on 6991 persons.

12. When using the non‐adaptative Gauss–Hermite quadrature, the coefficients of the variables change with a relative difference up to 167% between the models with 12 and 16 integration points. With the adaptative Gauss–Hermite quadrature, these relative differences are less than 5%.

13. ρ = σ_u ² / (σ _u ² + σ_ε ²), where ε are logistic, independent and identically distributed with mean zero and variance σ_ε ² = π² / 3.

14. We tested for possible interaction effects between gender and study programs; none of the coefficients of the interactions was individually significant at standard levels. These results may be obtained from the authors on request.