Yuchen Zhu

University of Florida · Economics · 2026–27 job market

Fields

Labor, Econometrics, Applied Micro

Job market paper

Predicting College Major Switching with Machine Learning

Major switching can benefit students, but it is costly when it happens late. Advising offices therefore want to know in advance who is likely to switch, so they can help students stay in their major or find a better fit. Using administrative data from the University of Florida, I build a machine learning model that gives each student an approximately unbiased and easily interpretable switching probability. Information available before college predicts switching only weakly, so advising offices cannot reliably identify likely switchers when students first arrive. Prediction improves as students’ records accumulate. Performance in a major’s core courses is key: it predicts even switches that occur several terms later. At the same advising capacity in each term, the model identifies 18 percent more switchers than the university’s current advising system.