Back to basics: Data revisions

With few exceptions, most economic data undergo revisions. Although frequently neglected, data revisions may have implications, not only for economic analysis, but also for policy decisions, as revisions may alter current assessment and forecasts of economic developments. In this paper, we reassess data revisions analysis and its impact on forecasting, presenting an encompassing and unified perspective on this subject. For this purpose, we built a real-time database for Portuguese exports and imports of goods. We present a broad set of the measures typically used to gauge revisions and add to this discussion by clarifying the relations between revisions to deferent types of series (for example, revisions to month-on-month and year-on-year rates of change). Furthermore, regarding the (un)predictability of revisions, we suggest an alternative testing approach. The key feature of this approach is that it takes into account both in-sample and out-of-sample performances. We also discuss the impact of revisions on forecasting, focusing on short-term forecasting of first releases. Even though not accounting for data revision implications can lead to suboptimal results, our findings reinforce the need for a case by case analysis.