What Happens When Official Data Is Revised
Statistics get republished with different numbers, on a schedule, and it is not a mistake being corrected. What revisions are for, which kinds exist, and what they do to a chart built on the old figures.
A revision is a design feature, not a correction
Statistical agencies publish on a fixed timetable rather than waiting for complete source data, because a number that arrives two years late is useful to nobody. The consequence is that the first version of a statistic is built from partial coverage, and better versions follow as survey responses, tax records and administrative registers catch up. The revised figure is what the agency now believes; the first one is what it could establish at the time. Neither is an error in the ordinary sense.
The routine kind
Most revisions are scheduled and narrow in scope: the same reference period is reissued once or twice as more of its own source data lands, and recalculated seasonal factors are applied. Agencies state in advance when each series is revised and how far back, and many publish summaries of how large their past revisions have been. That summary is the honest answer to how much weight a first print can bear, and it is specific to each series rather than a general rule.
The structural kind
Periodically an agency re-anchors an entire series: a new base year, new weights, a new industry classification, a full register replacing a sample. These benchmark revisions can move values going back years, and they are announced and documented in advance because they break comparability with everything calculated before them. A level shift on a long chart is often one of these rather than an event in the economy.
What it does to your notes and charts
A chart drawn today from a current database shows the revised history, not what was known at each point in it. Anything you concluded at the time was concluded on different numbers. This matters most for any rule or study calibrated on history: tested against revised data, it had access to figures nobody held on the day. Some agencies publish real-time or vintage datasets — the series as it stood on each past date — precisely so this distinction can be respected.
Reading a release that carries revisions
When a release restates earlier periods, read the revision table before the headline. The new figure's relationship to the previous month is defined by whichever version of that month is now in the series, so a headline comparison can change direction purely because of what was revised. The agency's own commentary usually says what drove the revision, and that is a better explanation than any reconstruction after the fact.
What revisions do not mean
A large revision is not evidence that a statistic is unreliable or that anyone was misled: the size of revisions is a documented, expected property of series built on incomplete initial coverage. Nor does a revision say anything about what comes next. And a revision arriving after a position was taken changes the record, not the outcome — the fill happened at the price it happened at, whatever the data now says about the month it happened in.
Where the answers actually are
Every question in this guide is settled by three documents the publisher already provides: the revision policy, the methodology note and the release calendar. They say when revisions happen, how big they have been, what changed at the last benchmark and where the vintage data lives. No secondary source needs to be trusted for any of it.
Capital at risk. Trading forex and CFDs carries a high level of risk and may not be suitable for all investors — most retail CFD accounts lose money. Never trade with money you cannot afford to lose. Read the full risk disclosure