Enhancing Study Success in Dutch Vocational Education

About 50 percent of all Dutch students enroll in a vocational education program. All efforts of vocational institutes are aimed at guiding students towards a diploma; yet, on average, about 28 percent of the students drop out. The lower the level of education, the higher the likelihood of unemployment (9,3% of all 15- to 25-year-old people are unemployed. When no start qualification is obtained (vocational level 2) the unemployment rate is 12,5% for this group.). Thus, students who drop out of vocational education have a higher risk of unemployment and poverty, which is a major social. Vocational institutes are aware of this social mission, and the life-changing consequences for students with these problems and are motivated to reduce the number of dropouts. Despite the abundance of dropout-related research, most of the articles are related to secondary education, high school, or college. Research on student dropout in vocational education is surprisingly scarce. There is a need for more information about the dropout decision in vocational education and the performance of dropout prediction modeling needs an improvement. Therefore, this dissertation aims to answer the following research questions: - What drivers determine student dropout in vocational education? - How can we improve early identification of dropouts? This dissertation addresses these research questions in five studies. The first three studies relate student expectations, cognitive skills and personality traits, and the differentiation skills of teachers to reduce student dropout. The final two studies use machine learning techniques to identify students at risk of dropping out and offer a systematic way to act on this. Main findings and conclusions The findings show that (1) dropped-out students did not have different expectations, as measured by our expectations questionnaire, about the vocational program than successful students, (2) personality traits and cognitive ability as measured by the formative entry test appear, unlike in other educational tracks, not to determine study success, (3) teachers do not tailor their lessons to the class in front of them, (4) our method that uses the dropout predictions of machine learning algorithms to construct sensitivity/precision trade-off plots can be used for targeted student dropout prevention, and (5) Teachers can increase the accuracy of predictions at the start of the program using machine learning algorithms. In addition to the main findings, dropped-out students had lower grades than successful students. Interestingly, this was already the case after the first quarter. This is in line with previous studies focusing on college students, where first grades in the program tend to be much better predictors than any other variable that students bring to college. This suggests that the first weeks of college are particularly decisive when it comes to subsequent dropout and suggests programs intended to stop dropout may have to start early after matriculation. The targeting of potential dropouts at an early stage of the program is important to assess if students can be helped in preventing undesired dropout. Machine learning algorithms enable better risk predictions, but –in themselves- do not provide hands-on advice to educational institutions about which students should participate in dropout prevention programs and when they should be invited. The precision-based approach introduced in this dissertation provides a way of doing just this. Policymakers can use this approach to make accurate dropout predictions and start intervening. To dropout is not the mirror image of persistence. Knowing why students drop out does not tell us (directly) why students persist. In the world of action, we must keep on sharing our knowledge with everyday practice and try to help institutions address pressing practical issues of persistence using scientific knowledge.