Sequential unordered logit applied to college selection with imperfect information

Abstract

The authors examine the implications of logit-based choice models for the input-transducer and output-transducer functions of an organization (in their application, a university). Logit models, defined as causal statistical models in -which the dependent variable is the odds that a particular event occurs, can be used to predict outcomes when deciders must choose among discrete alternatives (as in selecting a college or university). Unlike previous logit studies, the role of information flows between the organization and organisms outside the organization (potential students) is considered. It is shown that: (1) Static conditional logit models may yield inaccurate selection probabilities if, in reality, selection is made by recursive elimination of alternatives with new information added after each elimination; (2) a sequential unordered logit model can be developed which permits adjustment of the choice set during the information collection process; (3) this model can be used to evaluate the effectiveness of marketing policies. The model is applied to college selection and recruitment of college-bound high school seniors in southern Illinois.