What is the difference between an economic observation date and a data vintage?

Name the observation, release and retrieval clocks

The St. Louis Fed describes ALFRED as a collection of vintage versions of economic data. Its real-time API documentation distinguishes information available today about the past from information known as of a past date. Its vintage-date endpoint lists dates when values changed or new values were released; it excludes releases that did not change a series.

Those concepts do not automatically supply a complete publication calendar or an exact intraday timestamp. The ALFRED help page describes fallback sources for release dates and notes that additions generally follow release rather than proving instantaneous availability. A research record should distinguish source publication, database availability and the time the researcher actually retrieved a version.

Build an original release-and-revision ledger

Use a wholly invented index, not a real economic series. Its observation period ends 31 March 2026. In the teaching calendar, the first value, 102.0, becomes public on 20 April at 09:00 UTC. A revision to 103.5 becomes public on 20 May at 09:00 UTC. Stipulate no earlier access or other updates.

A fictional rule flags a case when the index value is at least 103. At a 25 April cutoff, the eligible value is 102.0, so the rule does not flag. At a 25 May cutoff, the eligible value is 103.5, so it flags. Selecting the later value merely because both rows describe March would change the earlier result using information that was not yet available.

Make the selection rule explicit

For each decision, first identify the intended observation period. Then retain only versions with a verified availability time no later than the decision cutoff, and select the latest eligible version within that set. Keep both the observation identifier and version identifier in the resulting record. The April decision joins to the first release; the May decision joins to the revision.

If no version is eligible, the result is unavailable at that cutoff, not a zero value. A later revision is not evidence that the original user or model should have known the replacement number. Reconstructing an earlier decision and estimating the latest historical measurement are different tasks.

Test an intraday boundary that dates alone cannot resolve

In the same fictional calendar, a decision at 08:30 UTC on 20 April precedes the 09:00 publication and has no eligible value. A decision at 09:30 may use 102.0 under the stipulated immediate-availability assumption. Both decisions share the same calendar date.

The FRED API’s real-time date boundaries are inclusive dates. That fact does not establish whether a particular value was available at 08:30. If only date-level vintage information is available, document the intraday gap or use a justified conservative convention rather than inventing a release timestamp. Actual access delays and later database corrections may also matter.

Method, sources, and limits

The calendar and threshold are original hypothetical inputs checked on 27 October 2026. No API key was created, no economic series was downloaded and no trading signal was tested. AI assisted the draft; independent AI verification is required before publication and is not qualified human financial review. The worksheet is a data-join design, not a claim that a particular application implements vintage-aware execution.