Estimating errors in student enrollment forecasting

the long version

The purpose of this paper is to demonstrate how longitudinal data can be used to determine variances, and hence confidence bounds, on student enrollment forecasts in addition to finding the forecasts themselves. The cases of known admission numbers and unknown admission numbers, but with an assumed Poisson distribution, are both considered. The model takes into account different admissions at fall and spring semesters, and also allows for differences in the continuation fractions for these different semesters. Normal approximations are used to calculate the probability that a total enrollment lies in a given interval. Numerical examples illustrate the results.