A Two-Stage Forecasting Model: Exponential Smoothing and Multiple Regression

This paper presents a forecasting technique which attempts to combine the advantages of both time series analysis and multiple regression. In this two-stage technique, an exponentially smoothed moving average model is used to forecast values of the dependent variable and/or selected independent variables as desired. These forecasts, along with data for other (lagged) independent variables, are then used as inputs to a multiple regression program. The observations are selected sequentially by the regression model so that each equation is based only upon data which would have been available at the time of the forecast, and the coefficiets of the equation are updated as new information becomes available. The final section of the paper describes a successful application of the two-stage model to a demand deposit forecasting problem.