First release and revision: comparing versions of economic data
Keep both values from one release, distinguish observation periods from publication and retrieval, and reconstruct the calculation known at the time.
A value describing a past month can change in a later statistical release. That does not make a later number available to an earlier reader. A vintage is a retained version of a dataset: identify which version supplied both numbers before calculating a change.
Three dates answer different questions
The observation period says what time the value describes. The release date says when the source made a version public. Your retrieval time says when you actually obtained it. A database’s ingestion or vintage date may differ from the original release and from your download.
If a real source gives only a date, keep that precision and mark the exact time unknown. Do not invent an hour. A new download date alone does not identify which release the downloaded file contains.
One fictional index, two coherent pairs
Index X is entirely invented. It is monthly, uses the same index units, base and treatment in both versions, and compares January with February 2030. Every date and clock time below is invented; none is a BEA release schedule or a real GDP observation.
V0 publishes 100 and 105 on March 10 at 09:00 UTC. V1 publishes revised levels of 102 and 104 on April 10 at 09:00 UTC. Both endpoint values change. Within V0, 105/100 − 1 = 1/20 = 5%; within V1, 104/102 − 1 = 1/51 ≈ 1.960784%. These are changes in an index, not investment returns.
Our own teaching example; all figures are hypothetical.
| Version | Observation periods | Published | Previous level | Current level | Change |
|---|---|---|---|---|---|
| V0 | 2030-01 → 2030-02 | 2030-03-10 09:00 | 100 | 105 | 5% |
| V1 | 2030-01 → 2030-02 | 2030-04-10 09:00 | 102 | 104 | ≈1.960784% |
Public release and a reader’s saved record
Reader A retains V0 at 09:02 and writes an entry at 09:03 on March 10. On April 10, reader B obtains V1 at 09:01; A retrieves it only at 09:05. Public availability does not prove that a particular reader received or used a release.
The eight events below separate public information from the locally retained evidence. When reviewing at 09:06 on April 10, A keeps the original March calculation and appends a distinct later-version result. The newer pair cannot be substituted into the earlier record.
| UTC | Actor and event | Public / locally retained | Permitted conclusion |
|---|---|---|---|
| 2030-03-10 08:59 | Readers A and B, before release | Neither teaching pair is public | Neither shown change can yet be attributed to public knowledge. |
| 2030-03-10 09:00 | Publisher releases V0 | Public V0: 100 → 105 | The coherent V0 change is 5%. |
| 2030-03-10 09:02 | A saves V0; writes at 09:03 | A retains V0: 100 → 105 | The dated original entry uses 5%; V1 is not yet public. |
| 2030-04-10 08:59 | A checks the pre-revision state | Public V0; A retains V0 | The later pair is still unavailable in this teaching timeline. |
| 2030-04-10 09:00 | Publisher releases V1 | Public V1: 102 → 104; A still has V0 | A has not automatically received the new release. |
| 2030-04-10 09:01 | B retrieves V1; A has not updated | B: V1; A’s saved copy: V0 | B can describe about 1.960784%; A must identify its older copy. |
| 2030-04-10 09:05 | A saves V1 separately | A retains V0 and V1 | Keep both pairs, source versions and retrieval times. |
| 2030-04-10 09:06 | A reviews the March entry | Original V0 and later V1 remain distinct | Keep the old 5%; append about 1.960784% as the revised-version calculation. |
Correct arithmetic can still mix versions
104/100 − 1 = 1/25 = 4% is arithmetically correct, but 104 is from V1 and 100 is from V0. It describes neither coherent release in this example. Identify the series, both periods and the same release before dividing.
New growth minus old growth is 1/51 − 1/20 = −31/1020 as a fraction of one. Multiplying by 100 gives −155/51 ≈ −3.039216 percentage points. A percentage point is the unit for a difference between two percentages; do not call this −3.039216 percent or −31/1020 percentage points. No annualization is performed.
Retain a dated source record
Save the series and source, units/frequency/base/treatment, both observation periods, release/version identifier, publication date or time with timezone and known precision, retrieval time, source URL and retained file, both values, formula and result. Keep the original entry time and append a separately dated review; a checksum can help identify the retained file.
For real data, distinguish the original publication from database loading and your local copy. ALFRED explains that release-date provenance can vary and that calculations from rounded published levels may differ from a headline rate. This example does not calculate a trading result, forecast or probability of profit.
Use the journal without inventing a product feature
Make the dated data record outside Elliott. Then use the existing scenario-journal lesson to keep observations separate from interpretations and to append reviews. Its first-calculation link concerns chart analysis, not calculating a macroeconomic series.
Elliott provides no macroeconomic feed, vintage repository or macroeconomic calculator, and this lesson does not promise importing macro observations as OHLC candles (open, high, low and close). An economic observation is not a trading signal. This is supplementary practice in preserving evidence.
Which pair reproduces A’s March 10 entry, which gives the revised change, and why is 4% neither release’s result?
Use 105/100 − 1 = 5% for the original entry and 104/102 − 1 = 1/51 ≈ 1.960784% for V1. The 4% calculation combines the V1 current value with the V0 baseline. Preserve the old entry and add the new version in a later review.
Sources
- BEA — why GDP estimates are revised (historical explanation, July 13, 2015)
- FRED — real-time periods and information known in the past
- ALFRED — vintage data, release dates and growth formulas
The 2015 BEA article explains revision mechanisms; its old schedule, numerical shares and revision cycle are not asserted as current. FRED/ALFRED explain version and date distinctions. No actual GDP series or restricted API was used.
Sources checked: 2026-10-21T02:06:55.960Z
This material and its translations were prepared with AI assistance. Educational calculations were checked programmatically. Sources explain concepts; examples and files were created for this article.
Next step
First retain the source and version in your own dated record outside Elliott. Continue with the scenario journal to preserve the original facts; its existing next step leads to the first chart calculation.
Keep a scenario journal →