Free access / Consumer evidence
Economic data vintages in free trading research
Preserve the value available at the decision date instead of substituting a later revised economic observation.
An observation date is not an availability date
An economic series can describe an earlier month while becoming available later. A research record therefore needs both the period described and the release or vintage relevant to the decision. The current value on a public series page is not automatically the value a past researcher could have read. ALFRED's documentation distinguishes original releases and later revisions. That distinction is useful without pretending that a macroeconomic series is a free buy or sell instruction. A data resource can answer a research question while leaving the trading rule entirely unspecified.
Retain transformations and units
Keep the series identifier, units, seasonal treatment, frequency and any transformation beside a saved value. A percentage change, index level and annualised growth rate answer different questions. If a rule uses several observations, retain the precise inputs rather than only the final chart. A revision can affect an input and its derived transformation. State whether the example uses actual downloaded observations or invented numbers. This publication supplies a fictional vintage case; it does not claim an independently reproduced economic strategy or an API-powered production signal feed.
Build the decision-time view
For a historical decision, choose the information available by the declared decision time and freeze that selection. If only a release date is known and the strategy needs intraday timing, the finer clock remains unresolved. Do not infer exact release seconds from a daily vintage label. Record missing releases and unavailable series instead of forward-filling them without a stated rule. The important research output is a reproducible information set, not a visually smooth historical curve constructed using observations published after the decision.
Separate access from redistribution
An openly readable data page does not by itself settle every reuse or redistribution permission for every underlying series. Preserve source metadata and the relevant terms before packaging observations into a new product. Do not place an unsupported blanket licence in a machine dataset simply because the hosting institution is public. A free research workflow can link to the original source and retain a private decision record without claiming ownership of the observations. Its final conclusion should identify the chosen vintage, transformation and unanswered timing question.
Completed fictional consumer case
Fictional case: a monthly indicator is first shown as 2.0 and later revised to 2.4. Leila's invented rule would trigger only above 2.2. The decision-time view therefore records no trigger, while the revised chart would misleadingly show one. These values and the rule are invented solely to explain the issue. No real FRED series, investment result or release clock is asserted. Leila keeps the two vintages separate and concludes that a current chart alone cannot reproduce her historical decision information.
| Record | Evidence object | Bounded finding |
|---|---|---|
| Initial vintage | Invented 2.0 | Below invented 2.2 boundary |
| Later vintage | Invented 2.4 | Revision changes apparent condition |
| Historical decision | Initial available value | No retrospective trigger |

Primary references and their limits
Public documents reviewed 13 October 2026. A link supports its stated mechanism or access description, not a completed account trial or this desk's invented case.