MULTIVARIATE TALENT FLOW ANALYSIS: A PILOT STUDY 1

This heuristically‐oriented study employed a wide range of data from the College Board Admissions Testing Program file on college‐bound students to explore a multivariate approach to talent‐flow analysis, or analysis of differences between admissions‐stage groups. Multivariate talent flow analysis, a largely undeveloped area of admissions‐related research, may be thought of as having two related general objectives: a) to analyze the correlates of dichotomous flow criteria—i.e., to determine which of several student variables, or talent descriptors, are associated with filing an application for admission (the flow criterion is application versus nonapplication), being offered admission (accept vs. reject), and enrolling (enrolling vs. not enrolling), respectively; and to assess the relative importance of the student variables as factors affecting, or involved directly or indirectly in, advancement/nonadvancement at each admissions stage; and b) to predict the respective flow criteria—i.e., to determine rates or probabilities of application, acceptance, and enrollment, respectively, according to score‐level on composites of relevant student variables optimally weighted to discriminate between criterion groups at each admissions stage. Results of the exploratory study, using data for college‐bound students sending CB score reports to one State University, strongly suggest that the type of information yielded by using the multiple regression or discriminant model to analyze differences between admissions‐stage groups with respect to a variety of academic, personal, and demographic data (talent descriptors) should be of value to colleges and universities interested in assessing and evaluating their admissions and recruitment policies and procedures. Moreover, the optimally weighted discriminant composite of variables found to differentiate admissions‐stage groups provides information that can be used to facilitate prediction of yield—e.g., to determine which candidates are most likely to complete an application or to enroll if offered admission. From the point of view of accountability in admissions, systematic analysis of the correlates of the accept/reject decision would appear to be relevant. By regressing the accept/reject criterion on appropriate student variables it would be possible to ascertain the relative role in admissions of, say, academic variables (such as tested ability and previous academic record) and nonacademic and/or socially relevant variables (such as, for example, socioeconomic status, sex, minority vs. nonminority status). Knowledge of the correlates of individual decisions to apply or to enroll if accepted should be useful to institutions interested in identifying high‐yield candidate categories in connection with their recruitment programs. Analysis of student data from a variety of colleges and universities representing different admissions modes and circumstances is needed to establish the extent and nature of variation in the patterns and the correlates of application, admission, and enrollment according to institutional differences in admissions policies and emphases.