Accurate enrollment forecasts are a prerequisite for reliable budget projections. This is because tuition payments make up a significant portion of a university's revenue, and anticipated revenue is the immediate constraint on current operating expenditures. Accurate forecasts are even more critical to revenue projections when a university's revenue allocations are tied by a state funding formula to credit-hour enrollment. Unfortunately, budget forecasters often extrapolate the latest enrollment change indefinitely into the future. Such judgment can mistake short-term seasonal and cyclical variations for a change in the long-term growth trend. The risk of such a mistake is especially high when the growth trend is moderate, as in the 1970's, because short-term variations are more visible than they are during periods of rapid growth, as in the 1960's. Under present conditions, if a temporary increase in enrollment one year is extrapolated as a trend change to the following year, underbudgeted years will alternate with overbudgeted years. This means year-end travel and procurement curtailments. The study suggests that this type of error can be avoided by using traditional techniques of time series analysis to identify fairly stable short-term enrollment patterns. Forecasters can then view enrollment fluctuations as a composite of short-term variations and the long-term trend variation. The techniques are demonstrated as applied at a large urban public university. (Author)