Analysis and Prediction of Telephone Demand in Local Geographical Areas

An approach to forecasting the demand for local area telephone service is presented in this paper. The specific problem discussed is the forecasting of main stations in three Michigan metropolitan areas. Several different statistical models are used. The first class of models introduced used adaptive exponential smoothing and is based solely on the past history of the time series involved. Although appropriate data at the local area level are very difficult to obtain, two exogenous time series related to household formation are used to construct more elaborate models for one of the areas. The various models are evaluated by both the average absolute and the root-mean-square forecast error. In terms of these criteria, the first class of models referred to above performs reasonably well while the second set does considerably better. This argues strongly that future improvements in forecasting accuracy will be made by the more extensive involvement of exogenous variables.