Forecasting with an Input-Output Model

Abstract

One of the more difficult problems of the modeling of corporate activities is in the area of forecasting equipment retirement. This paper develops a statistical forecasting method based on an input-output model suitable for use in this particular problem as well as in similar areas. The model assumes that present output is a linear, time invariant function of current and past input values. The model parameters may be developed by employing a constrained gradient technique. The use of this method is illustrated by developing retirement frequency and survivor curves for several input-output time series of electric utility plant data. These results are obtained without the assumption of particular shapes for the retirement frequency functions. The results are compared to those obtained with more conventional techniques. This comparison leads to the conclusion that the method may be advantageously employed for retirement studies where the equipment class exhibits a very long life or where the equipment mortality characteristics do not tend to behave in a fashion that is easily anticipated.