Combinations of forecasts: A model-building perspective

A brief summary of recent issues in the theoretical methodology and experimental application of combinations of forecasts is presented. The authors raise questions as to the meaning and interpretation of models that emphasize algebraic combinations of point forecasts minimizing mean squared error without insightful treatment of the different decisions and policies that usually accompany the forecasting context. They conclude by making a strong plea for the role of models that carefully synthesize information on inputs rather than combine forecast outputs of models based on different information sets.