This paper argues that a typical use of regression models to target student recruitment efforts is theoretically unsound and may therefore be operationally inefficient. It presents results from a study using a predictive model to identify the prospective students on whom recruitment efforts have the greatest impact. The model uses four kinds of variables: demographic, academic, geographic, and behavioral. Application of the model at the State University of New York (Stony Brook) found that predictive variables significantly and positively related to student enrollment at the 1 percent level were: high-yield high school average, high-yield Scholastic Assessment Test scores, high-yield zip code, and open house attendance. Significant predictive variables related negatively to student enrollment included White or Hispanic ethnicity, U.S. citizenship, regular admission status, early application, and on-campus housing request. The model was used to demonstrate the efficacy of an experimental program of increased contact with admitted students and their parents. Findings indicated that a modest increase in recruitment activity increased the enrollment of students with relatively low enrollment probabilities but did not improve the recruitment of students identified by the regression model as more likely to enroll. Results suggest that the typical use of predictive modeling to identify "hot prospects" may be inefficient and ineffective. (DB)