Inequality of opportunity in selection procedures limits diversity in higher education: An intersectional study of Dutch selective higher education programs

internationals who were already living in The Netherlands (giving them a Citizen Service Number), but for whom parental data are missing as parents live abroad. Six out of eight of the multivariable models of these clusters showed no significant association between income/ assets category and the odds of admission. Due to the high amount of missing data, these results must be interpreted with caution. In this research, we assumed that data was missing completely at random. However, there are methods for handling missing data, e.g. multiple data imputation [66]. In future work, we could investigate the usefulness of multiple data imputation. Our focus in this paper was to create association models for the two academic years under study in the Dutch HE context. These models are not prediction models, which could be con- sidered a limitation. Furthermore, the models are not validated in other contexts or other years. To generalize our results to other contexts, we would need to create prediction models which are validated based on data from an external context and from different time periods, which we do not have. Because of the explorative nature of our study, in which our goal was to find out which variables are more or less related to gaining admission to selective programs, we did not adjust the p-values for multiple testing. Finally, in Table 3, Group ALL and NONE consist in some study clusters of small numbers of applicants and/or admitted students. We addressed this limitation by creating figures simi- lar to Fig 2 for all clusters (in S1 File). Future research

The diversity of the student populations as shown in S3 Table only applies to the students at selective programs. As shown in S1 Table, there are many HEIs that do not use selection for the same study programs. Therefore, S3 Table lacks the composition of the student populations of comparable non-selective programs (e.g. in Nursing, where only one out of 17 HBOs used selection). It is possible that self-selection has an effect both on whether and where students apply to such programs, and that certain groups would rather avoid a selection procedure compared to other groups. Thus, further research could investigate potential differences in stu- dent diversity between selective and non-selective programs. This could also aid in mapping the diversity and representativeness of the future workforce in these fields. Although our quantitative analyses found widespread inequality in selection to Dutch HE, our data cannot explain what causes this inequality (e.g. do certain selection instruments dis- advantage certain groups? Are certain instruments used more often in some clusters than in others?) and why there are such large differences between clusters. Future work could answer these questions with qualitative or mixed-methods research. Lastly, as we were not able to acquire data on parental jobs for this study other than jobs in the healthcare professions or in education, future research could investigate if other types of parental jobs also have an influence on the odds of admission. Supporting information S1 Table. Clusters of study programs. (DOCX)

S2 Table. Results from the univariable logistic regression. *p < .05; **p<0.01; ***p<0.001. Ref. = reference category; PSVA = Primary, Secondary or Vocational or Agrarian; TMSDI: Turkish, Moroccan, Surinamese, Dutch Caribbean, or Indonesian; Urbanity Degree of postal code of residential address in 2015. Results in bold are statistically significant (p < .05). (DOCX) PLOS ONE Inequality of opportunity in Dutch selection procedures limits diversity in higher education PLOS ONE | https://doi.org/10.1371/journal.pone.0292805 October 13, 2023 20 / 24