Let me elaborate on an issue that concerns almost every real-time statistic (in particular economic indicators)and which leads to on-going confusions by - and frustrated expectations of - users, namely revisions.
Revisions are corrections/updates of the history of a time series based on new evidences supported by (most)recent data. The main sources of revisions are data up-dates (by data-providers) and filter up-dates (smoothing operations by index providers). At first sight, revisions seem to be an unquestionably meaningful operation that aims at a better understanding of the past given present knowledge. Unquestionable, really?
