Dynamics of Exponential Smoothing with Trend and Seasonal Terms

Abstract

The characteristics of seasonally adjusted, exponentially smoothed forecasts are studied through the frequency response and impulse response functions. The appropriateness of using exponential smoothing for updating the seasonal base series is called into question, due to its tendency to store random noise for long periods. Exponential smoothing with trend is also examined. It is concluded that Brown's formulation has the property of being critically damped, so that its response to changes in the time series under study is the most rapid possible without overshoot.