Improved methods of combining forecasts

Authors

  • Clive W. J. Granger,

    1. Department of Economics, University of California, San Diego, La Jolla, California 92093, U.S.A.
    Search for more papers by this author
    • Nottingham in 1959 where later he held the Chair of Applied Statistics and Econometrics. He was a Harkness Fellow at Priceton in 1959–1960 and has also been a visitor at Stanford, Vienna, Canberra and elsewhere. He has published seven books and over 90 articles on forecasting, time series analysis, spectral analysis, econometrics, speculative markets and pricing theory. He has been Professor of Economics, University of California, San Diego since 1976 and is currently Chair of Economics. He is also Fellow of the Econometric Society and associate editor of several journals, including Applied Economics, Energy, Journal of Econometrics, Journal of the American Statistical Association, Business and Economic Statistics and Journal of Financial Economics.

  • Ramu Ramanathan

    1. Department of Economics, University of California, San Diego, La Jolla, California 92093, U.S.A.
    Search for more papers by this author
    • Mathematics and Statistics from the Indian Statistical Institute and a Ph.D. in Economics from the University of Minnesota. His interests are applied econometrics, economic growth and international trade. He has written an advanced book on growth theory and published numerous articles which have appeared in leading journals in the U.S.A., U.K., Australia and India.


Abstract

It is well known that a linear combination of forecasts can outperform individual forecasts. The common practice, however, is to obtain a weighted average of forecasts, with the weights adding up to unity. This paper considers three alternative approaches to obtaining linear combinations. It is shown that the best method is to add a constant term and not to constrain the weights to add to unity. These methods are tested with data on forecasts of quarterly hog prices, both within and out of sample. It is demonstrated that the optimum method proposed here is superior to the common practice of letting the weights add up to one.

Ancillary