The interaction between dropout, graduation rates and quality ratings in universities

Abstract

This paper investigates in a non-parametric framework whether academic programmes maximize their student graduation rates and programme quality ratings given the first-year student dropout rates. In addition, it explores what institutional and programme characteristics explain this interaction. The results show a large variation in how academic programmes are able to deal with the selective nature of first-year dropout. Nevertheless, we can accurately explain the variation among programmes by programme and institutional characteristics. It seems that universities can maximize the relation between first-year dropout, graduation rates and quality ratings in several ways: (1) by improving student programme satisfaction, (2) by better preparing certain groups of students for higher education, (3) by supporting male students, (4) by supporting ethnic minority students, (5) by attracting older staff, and (6) by strengthening the selective nature of the first year (ie, increasing the academic dismissal policy threshold).

Keywords:

• student dropout
• student graduation
• conditional efficiency
• nonparametric estimation
• universities

Acknowledgements

We would like to thank participants of the Sixth North American Productivity Workshop, the Dutch Ministry of Education and ORD 2014, Wim Groot, Henriëtte Maassen van den Brink, Kees Boele, Subal Kumbhakar, Tommaso Agasisti, the associate editor and two referees for valuable insights and comments. We gratefully acknowledge the support with the data received by Studiekeuze123, especially by Bram Enning and Constance Dutmer.

Notes

1 Note that graduation rates and quality ratings are often not measured in a consistent way. We incorporated the used definitions/measurements in Tables A1 and A2 in Appendix A.

2 However, few studies have investigated the direct relationship between satisfaction and graduation rates.

3 Note that academic programme, given their dropout, can choose not to obtain the highest graduation rate and quality rating as possible (ie, the Free Disposability Assumption). Further in the section, we discuss this more deeply.

4 Note that in the DEA literature, this ‘trade-off line’ is known as the ‘efficient frontier’.

5 The only exceptions are the numerus fixus programmes in medicine, dentistry and other areas that have a limited number of student places. Numerus fixus means that programmes can choose which students they will accept. Note that only 0.30% of the available places can be filled in by this practice (Jongbloed, 2003).

6 Thus, the student graduation rate is measured conditional on first-year dropout.

7 Note that these sample sizes differ because we only included programmes which have information on all the control variables. Programmes with missing values are thus removed. There are more academic programmes with missing values in the academic year 2010–2011 compared with 2011–2012. This is not surprising since the data gathering of Studiekeuze123 is improving every year.

8 The number of bootstrap replications only matters for the statistical inference. The conditional order-m model has been estimated in R by using the integral formulation, as this procedure is more time efficient and precise.

9 Note that when we combine the two years in one data set and we do the above analyses again we find similar results. Moreover, we find a significant negative correlation between the graduation and dropout rate and an insignificant negative relation between graduation rate and the quality ratings.

10 In line with Jongbloed et al (1994, 2003), we divided the academic programmes into an arts, a sciences and a medical cluster and reran the analyses. The results of the arts cluster are robust with the current findings. The results of the sciences cluster showed differences concerning the influence of the control variables. Owing to insignificant power we did not find results for the medical cluster.