Economic forecast quality: information timeliness and data vintage effects

This paper investigates the impact of the timeliness of information releases and data vintage variation on economic forecast quality. Specifically, using a set of 63 key US economic series, we provide a concise measure of the forecast accuracy associated with use of economic activity indices with different publication lags. A forecasting model based on an economic activity index that is subject to a short publication lag (viz. the Aruoba-Diebold-Scotti index) is more efficient than competing models. Moreover, if this publication lag advantage is removed (by artificially imposing a publication lag restriction comparable to that of a competing indicator) this efficiency largely disappears. The final part of the analysis employs a novel (simulation-based) method of assessing the impact of data vintage variation on forecast accuracy, and finds that the results are somewhat sensitive to such variation.