Satisfied but Gone: Why What Freshmen Do Predicts Retention Better Than What They Say

Universities administer freshman surveys hoping that early attitudes will flag students at risk of leaving. We linked three entry cohorts ( N  = 8,016) at a Taiwanese university of technology to enrollment, attendance, and grade records, letting what students said compete with what they did. Attitudes lost. Satisfaction, expectation fit, and willingness to recommend barely distinguished leavers from stayers, whereas first-semester grades and unexcused absences separated them sharply, and the strongest survey-based signal was silence, as nonresponse carried a fivefold attrition gap (10.6% vs. 2.0%). Three routinely stored fields yielded a cross-validated AUC of 0.80, a random forest added nothing, and the signals held among students still enrolled after year 1, predicting who would quietly disappear a semester later. Clustering campus-life ratings showed that coolness toward coursework tripled departure risk, whereas dissatisfaction with facilities posed no risk. We outline a low-cost early-alert workflow built on the survey that came back unanswered.