Every restatement of a published figure, in date order, with its previous value, its new value and the reason it changed. Nothing is corrected in place.
This is the behaviour that separates a data publisher from a blog. If a number here has ever been wrong, this page says so — and if you have quoted one of ours, this is where you check whether it still stands.
7 restatements, and 11 entries in total including first publications.
PIDI-E electricity column re-pulled from EIA; 2 of 125 break-even values move by 0.001
- Was
- reconstructed to two decimal places from the published ratio and the gas column
- Now
- measured, direct from EIA Monthly Energy Review T09.08
The index's release gate was a first-run EIA refresh replacing the reconstructed electricity column. It has run. The reconstruction proved accurate: every one of the 125 published break-even values reproduces from the real EIA inputs, 2 of them moving by exactly 0.001 — one unit in the last published place — and none by more. The electricity column itself was out by up to 0.18 c/kWh, because it had been derived from the published ratio rather than measured; the ratio, and therefore the break-even, was right throughout.
PIDI-L 2026 restated from 158.5 to 160.3
- Was
- 158.5
- Now
- 160.3
The partial seed was replaced by the full Census M3 series. Census has revised the underlying unfilled-orders and shipments figures since the seed was transcribed, and the 2019 base value moved from 3.5657 to 3.5696 with it. The change is +1.77 index points, inside the ±0.5 tolerance §V2 sets for this reconciliation.
PIDI-L 2025 restated from 160.8 to 160.4
- Was
- 160.8
- Now
- 160.4
The partial seed was replaced by the full Census M3 series. Census has revised the underlying unfilled-orders and shipments figures since the seed was transcribed, and the 2019 base value moved from 3.5657 to 3.5696 with it. The change is -0.43 index points, inside the ±0.5 tolerance §V2 sets for this reconciliation.
PIDI-L 2024 restated from 163.6 to 163.4
- Was
- 163.6
- Now
- 163.4
The partial seed was replaced by the full Census M3 series. Census has revised the underlying unfilled-orders and shipments figures since the seed was transcribed, and the 2019 base value moved from 3.5657 to 3.5696 with it. The change is -0.25 index points, inside the ±0.5 tolerance §V2 sets for this reconciliation.
PIDI-L 2021 restated from 126.2 to 126.0
- Was
- 126.2
- Now
- 126
The partial seed was replaced by the full Census M3 series. Census has revised the underlying unfilled-orders and shipments figures since the seed was transcribed, and the 2019 base value moved from 3.5657 to 3.5696 with it. The change is -0.25 index points, inside the ±0.5 tolerance §V2 sets for this reconciliation.
Register counts corrected: 88 fits, 77 meeting the quality bar
- Was
- 90 fits, "38 meeting the bar"
- Now
- 88 fits, 77 meeting the bar; 39 qualifying fits on the equipment-cost basis
The register README described 90 fits and 38 meeting the quality bar. The data holds 88, of which 77 meet it; the 38 was never the pass count — it was the qualifying equipment-cost subset, now 39. The headline statistics were computed on that subset and are correct for it, but the document described them wrongly, and anyone reading "38 of 90 passed" would have drawn a far gloomier picture of the data than is true. No exponent, coefficient or quality flag changed.
Factor set rotating-cooling published at version 2026.1
6 equipment records assembled and validated against the factor set schema.
rotating-cooling, factor set version 2026.1
Factor set grid published at version 2026.1
1 equipment records assembled and validated against the factor set schema.
grid, factor set version 2026.1
Factor set generation-storage published at version 2026.1
2 equipment records assembled and validated against the factor set schema.
generation-storage, factor set version 2026.1
Factor set heat published at version 2026.1
5 equipment records assembled and validated against the factor set schema.
heat, factor set version 2026.1
Validation order corrected: the train count is now determined before the range check
- Was
- Range check applied to total capacity, so any capacity above max_single_unit refused before multiple-train pricing could be reached
- Now
- Train count determined first; the range check and every later rule apply to the per-unit capacity
Under the previous order the N-train window was (max_single_unit, valid_max], which is empty on the only record carrying a single-unit ceiling — so multiple-train pricing was unreachable in every factor set. The two limits answer different questions: valid_min and valid_max are a statistical limit, the range the curve was fitted over; max_single_unit is a physical one, the biggest machine available. In N-train the power law is evaluated per unit, which is inside the fitted range by construction. No published parameter changed.
Method changes, as distinct from restated values, are listed in the version history on the method note.