Estimating errors in student enrollment forecasting

This paper shows how longitudinal data for student enrollments and attendances can be used to obtain variances and confidence limits in the forecasts of future enrollments. The effect of historical data on conditional expectations and variances of enrollments is explicitly included in our formulas, as is the “odd-even” effect of admissions during fall and spring semesters. Thus, it is possible with little effort to obtain confidence intervals for forecasts (in absolute numbers or in percentage terms) from the same longitudinal data that provide the forecasts themselves. We include calculations for the special cases of large cohort sizes and Poisson admissions.