Abstract
The study examines whether the reduction in the timeframe for the completion of university studies in Greece will affect students of different socio-economic background disproportionally. To this intent, it assesses the influence of the status of students, defined as working and non-working, on the duration of studies but the relevance of other variables, notably students' family income, is also examined. Combining administrative and survey data, the study estimates the probability of students' graduating after 4 and 6 years, respectively. The results reveal the existence of strong income discrimination between working and nonworking students. Among the non-working students, those coming from poor families complete studies earlier than wealthier students. For the working students, there is no income effect.
Acknowledgements
We wish to thank an anonymous referee of this journal for his/her constructive comments and suggestions which helped us to substantially improve the paper. We are grateful to the administration of the University of Macedonia for giving us permission to access individual students' records on a limited basis. We are indebted to Theodore Panagiotidis and Theologos Dergiades for their help and advice on the quantitative part of the paper. We also wish to thank the following members of the administrative secretariat for their assistance in recording the data analysed in this article: Josephina Dimitrova, Stelios Haritakis, Eleftheria Kiale, Niki Kougoula, Zoi Manou, Fotis Sarigiannis, Maria Tsapakidou, Vicky Valani and Thomai Zia.
Notes
1. For the majority of degree programmes, the expected duration of which is 4 years, the law implied that students failing to graduate after 8 years of studiesor failing to attain a pass mark on any one course after 8 attempts would be dismissedfrom universities.
2. The concordance statistic (Cij) is estimated from the following formula: , where, T is the sample size, Sit is a binary variable receiving the value of 1 when a student is working and 0 otherwise, Sit is also a binary variable receiving the value of 1 when a student is prolonging its studies above 4 years and 0 otherwise. The significance of the concordance statistic is conferred from the a coefficient whose value is obtained by the following moment condition:
. In the GMM estimation, we use the Bartlett kernel option with a fixed bandwidth of 3; for further discussion on the concordance statistic, see, Harding and Pagan (Citation2002). For an exposition of the way that the concordance statistic is implemented, see Dergiades and Tsoulfidis (Citation2007).
3. The presence of endogenous right-hand side variables in a probit model presents familiar problems of estimation. IV estimation was employed in the cases where endogeneity was suspected. Likelihood ratio tests were estimated under the null hypothesis of exogeneity. The FIML estimators were not qualitatively different form the ones reported in Table .