This paper examines alternative techniques for projecting freshman enrollment in specific academic departments. Departmental enrollment projections provided by four different projection models are compared to actual departmental enrollments at a selected institution. Two of the models use only historical data, while the other two models are sensitized to current developments as indicated by the expressed major choices of prospective freshmen. The use of discriminant analysis to establish differential enrollment probabilities is also explored.
Although different models do a better job for different curricular departments, the smallest mean squared error across all departments was obtained with the simplest projection technique. The use of the preliminary major choice of prospective freshmen did not improve departmental projections, and the student characteristics explored in this study did not differentiate enrolled from non-enrolled students adequately enough to improve enrollment projection accuracy.
Based on the results obtained at this one institution, therefore, it would appear that simple and straightforward projection models can be as useful as complex and sophisticated models.