Market evidence has multiple dates. The observation period tells you when the measured activity occurred. The publication date tells you when the source released the information. The retrieval date tells you when SAID obtained or reviewed it. A revision date may tell you when the source changed a previously published value.
These clocks can be days or months apart. A release published in September may describe transactions completed in July. A census value retrieved today may describe conditions years earlier. A daily-updated file may contain records whose underlying transactions were registered after a delay.
Publication lag
Lag is not a flaw by itself. Administrative and statistical systems need time to collect, validate and publish information. The problem begins when publication recency is mistaken for observation recency.
Incomplete recent periods
Transaction datasets can be especially sensitive to late registrations. The newest month may look weak simply because not all records have arrived. Comparing an immature period with a complete prior period can create a false decline.
Revisions and vintages
Official series are sometimes revised as additional information arrives or methodologies change. A robust system should not treat the newest value as though it had always been known. Preserving release context makes it possible to distinguish what was observable at a given time from what was later corrected.
Practical reading
- Use the observation period to understand when the measured condition existed.
- Use the publication date to understand information lag.
- Use the retrieval date to understand SAID's evidence freshness.
- Use revision metadata to understand whether history has changed.
Recency is therefore not one property. A value can be newly published but describe an older period, or recently retrieved but belong to a slowly updated series.