CALC-060

Currency and basis-year conversion: the deflators and exchange rates the whole restatement rests on

Feeds CLM-0376. Answers QST-004. Basis: 2024 basis year, 2024 annual-average FX, throughout.

Script: the LCOE restatement. SECTION 1, lines 82-190. Constants at lines 98-145; cumulative_multiplier() line 106; to_gbp_2024() line 148.

Why it was needed. Every other record in this section multiplies by these numbers. If they are wrong, all twenty-three claims move together. Declaring them once, with their status attached, is what makes the rest checkable rather than merely readable.

Inputs. This table is the point of the record: every row is ASSUMED, and the script says so itself.

Why "recalled" is the honest word and not a euphemism. The claim's own assumes field states the reason: WebSearch and live retrieval were unavailable for that pass. This is not a case of a figure being asserted without a source type; it is a case of the right source type being named (ONS, BLS, Eurostat, Bank of England, ECB) and not reached. The distinction matters for what closing it costs: five lookups, not a re-derivation.

Working, transcribed.

Cumulative multiplier = product of (1 + annual rate) over the years AFTER
the basis year, up to and including 2024.   [script line 106-115]

UK  2018 -> 2024   1.2658
UK  2022 -> 2024   1.1009
US  2022 -> 2024   1.0712
EA  2022 -> 2024   1.0793

Worked example, the one the claim itself carries (script Section 4(a),
emitted at line 675):

  full method    EUR106.00 x 1.0793 (HICP to 2024) = EUR2024 114.41
                 EUR2024 114.41 x 0.8466 GBP/EUR   = GBP2024  96.9

  naive method   EUR106.00 x 0.8520 (2022 GBP/EUR) = GBP      90.3
                 [FX only, no inflation step - what CLM-0360 did]

  difference     96.9 / 90.3 = 1.0723,  i.e. +7.2 per cent

Result: the currency and basis-year adjustment is worth +7.2 per cent on the worked example, and single digits to low double digits generally.

State what that 7.2 per cent is a ratio OF, because this is the number most at risk of being misread. It is the ratio of the fully restated figure to the naively converted figure for one study's one percentile - Frazer-Nash ESA's EUR106/MWh FOAK p10. It is not a 7.2 per cent uncertainty on the restatement, and it is not the contribution of currency to the spread between studies. It is the size of the correction the naive method omits.

Conclusion. The smallest of the six drivers ranked in CALC-068's companion claim CLM-0385, and immaterial against a residual that is a multiple rather than a percentage. That is the whole argument for why recalled deflators do not sink the exercise, and it is sound only because the residual is 19-37x. It would not survive a residual of 1.5x.

Where this fails. Every deflator and every rate is recalled. The claim's own limits field says they MUST be checked against ONS, BLS, Eurostat, the Bank of England and the ECB before this feeds any external-facing or decision-grade output, and that instruction should travel with this record. review_due on CLM-0376 is 2026-12, the earliest of any claim in this section, which is the register correctly treating this as the perishable input it is.

A DEAD, MISLABELLED CONSTANT IN THE SCRIPT, RECORDED BECAUSE THE SCRIPT IS BEING PROPOSED AS PROVENANCE. Line 137 defines FX_2024 = dict(GBP_PER_USD=1.0/1.278, USD_PER_GBP=1.278, EUR_PER_USD=1.082/1.278*1.278). The third entry's expression evaluates to 1.082, which is USD per EUR, not EUR per USD - the name is the reciprocal of the value. No figure moves, because `FX_2024` is never used: the identifier appears exactly twice in the file, at line 24 inside the module docstring and at line 137 where it is defined, and nowhere else. The script's own next line says "(kept explicit rather than clever: define directly)" and the explicit constants at lines 141-143 are what the code actually uses. Recorded rather than fixed, and recorded rather than omitted: a mislabelled constant in a file offered as auditable working is a defect even when it is inert, because the next reader cannot tell it is inert without grepping.

Cite this