We modelled school location choices of post-secondary students in the Greater Toronto and Hamilton Area using data from two student travel surveys. We compared the performance of random utility models (multinomial logistic regressions) and machine learning methods (random forest models), discussing the merits of each. Our key findings include the effectiveness of gravity models compared to more advanced accessibility models, important relationship between student’s living arrangement and school location choice, and the different patterns in behaviours of university and college students. We also find that while the random forest models obtain similar accuracy to the classic econometric tools, they are less suitable for predictive applications or extensions to unseen data.