Les progrès dans les prévisions : météorologie et économique

Abstract

-Meteorological and economic modelling and forecasting have a number of common features, which suggest that it may be interesting to compare their recent progress in forecasting. We concentrate on two aspects of forecasting: first, measures of the value added of forecasts, and the evolution of these measures over approximately the last twenty to thirty years; second, the estimation and representation of uncertainty of forecasts. We follow several particular quantities as representative of different types of forecasts: temperature, GDP growth and financial market volatility (continuous variables); probability of precipitation and probability of recession; 0/1 predictions of precipitation or recession (binary forecasts); tornadoes and market crashes (rare events). We describe the available information on the evolution of forecast quality, and describe also the evolving methods for understanding and representing uncertainty in the forecasts.