Forecast evaluation with cross-sectional data: The Blue Chip Surveys

If economic forecasts are to be used for decision making, then being able to evaluate their accuracy is essential. Assessing accuracy using single variables from a forecast is acceptable as a first pass, but this approach has inherent problems. This article addresses some of these problems by evaluating and comparing the general accuracy of a set of multivariate forecasts over time. ; Using the methodology developed in Eisenbeis, Waggoner, and Zha (2002), the authors compare the economic forecasts in the Blue Chip Economic Indicators Survey. The survey, published monthly since 1977, contains forecasts of many macroeconomic variables over a relatively long time span. The forecasters are a mix of economists from major investment banks, corporations, consulting firms, and academic institutions, many of whom have participated in the survey for several years. The survey thus provides a useful set of forecasts to explore the methodologies and to investigate several aspects of forecast performance over time. ; The methodology assigns each forecast a composite score based on the standard theory of probability and statistics. This single number is easy to interpret and can be used to compare forecasts even if the number of variables being forecast, or their definitions, changes over time. ; The analysis shows that the Blue Chip Consensus Forecast, which is the average of the individual forecasts, performs better than any individual forecaster although several forecasters performed almost as well as the consensus.