Individual level data for the entire cohort of undergraduate mathematics students of a relatively small US public university was used to estimate the risk that a student will switch major to another one before degree completion. The data set covers the period from 1999 to 2006. Survival tables and logistic models were estimated and used to discuss student's switch from their initially planned mathematics major to another. An important goal of low enrollment university departments is to increase enrollment and to attain a high retention rate. Major switching indicates a failure of these departments to retain potentially able students. Students' retention is also an important issue for university programs evaluation and funding. The goal of this work is to investigate the timing of switching major, to a different one, by mathematics students. We estimated the probability of occurrence of this event for different school terms and when this event is most likely to occur. The empirical results suggest that the probability of a mathematics student switching major varies from a high 23% early in student's major enrollment to a low of about 6% in later semesters. Mathematics education majors, however, showed inferior risk of switching major. Gender differences were also examined, showing no significant gender differences.