Restricted forecasts using exponential smoothing techniques
A methodology is presented for obtaining optimal forecasts with exponential smoothing (ES) techniques when additional information, other than the historical record of a time series, is available. Such information is usually given as linear restrictions on the future values of the series and may come from: (i) expert judgments, (ii) alternative forecasting methods or (iii) scenarios to be portrayed. Appropriate usage of the additional information improves the forecasts' accuracy and precision. Here we provide closed expressions for the restricted forecasts obtained with the most frequently employed ES methods and emphasize the potential usefulness of the proposed methodology in practice.