Early Identification of College Dropouts Using Machine-Learning: Conceptual Considerations and an Empirical Example (2019)
Joint work with Ingo Ispording, published in IZA Research Report No. 89.
Can and should we use modern big data methods to predict which students will drop out of their studies? Such an early detection of at-risk groups of students would allow practitioners and policy-makers to timely intervene through targeted counter-measures as for instance mentoring, counseling, or institutional support. This report discusses the feasibility, benefits and potential hidden costs of such an approach which would allow to mitigate substantial costs of educational career frictions.