Testing the efficiency and rationality of City forecasts

This paper examines the accuracy of city forecasts of UK data releases as collected by Standard and Poor's MMS. It finds that the median forecast calculated by MMS are, for the majority of series, biased and inefficient. The numerical size of these deviations is found to be small, unlike in some other studies. Evidence is presented that revisions may take place during the period between the collection of the forecast and the publication of the data. If significant forecast revisions take place this would compromise the usefulness of forecast surveys in the estimation of `news' effects on financial asset prices. Regression analysis does not suggest that revisions can be systematically measured by changes in current dated variables or by news from other variables. News effects therefore measure the direct effect of the forecast error rather than also including the effects of revisions to forecasts of, as yet, unpublished data.