Assessing the accuracy of businesslevel forecasts

This article presents original work on some aspects of forecasting at the individual business or firmlevel. In particular, two ways are suggested for assessing the accuracy of these forecasts based on the calculation of average percentage errors and the construction of a 95 per cent confidence interval. It is found that the quality of forecasts tend to become increasingly unreliable after two years and that the decay in forecast quality is inversely related to the frequency of the time series that is the less often a time series is updated/published the faster the deterioration in forecast quality.