Methodology & Data

The primary source is the Central Bank of Iraq's own publication of its daily foreign-currency sale: session bulletins, the market rate tables published alongside them, and the annual and monthly aggregates the Bank itself publishes. Where a bulletin survives with an intact text layer, the figures are read from the text. Where only a scanned image survives, the figures are read by optical character recognition and the record carries an OCR flag.

The register of institutions is compiled from the Central Bank's licensing lists, the Iraq Stock Exchange, and public ownership records. Enforcement status is taken from US Treasury and FinCEN publications and from the Central Bank's own announcements.

Third-party estimates — the smuggling, ghost-employee, capital-flight and budget-diversion figures in Chapter 05 — are attributed to their source, carry a Tier C badge, and are never combined with figures this file computes.

Tier A: published by the Central Bank of Iraq and recovered from its own record. All volume, rate and per-bank figures on this site are Tier A.

Tier B: audited financial statements, official registers, and primary regulatory or legal documents from other institutions.

Tier C: third-party estimates and attributed statements. Presented with the name of the person or organisation making them, and never used in this file's arithmetic.

The daily series spans 8,240 calendar days from 1 January 2004 to 23 July 2026. A volume figure has been recovered for 3,787 of them. Days for which the Central Bank published nothing, or for which the bulletin has not been recovered, are held as absent — never as zero, and never interpolated.

Consequently every aggregate on this site is a floor. The documented total of $514,997,749,994 is what has been recovered, not what was sold. Where the Central Bank published an annual total of its own, Chapter 02 shows the recovered figures against it as a coverage ratio.

The parallel rate is recorded as observed where a contemporary market quotation was recovered, as cbi-market-table where the Central Bank itself published a market rate, and as interpolated where the value is derived linearly between two observations. The composition of the series is exact: 3,793 observed daily rates, 18 from the Central Bank's published Baghdad market-rate table — 3,811 documented in total — and 4,429 derived by interpolation, giving the 8,240 days of the series. Interpolated values are drawn dashed and lighter, and every statistic computed over a period discloses the interpolated share of its days.

Premium is (parallel − official) ÷ official, expressed as a percentage. The baseline official rate in force since February 2023 is 1,310 dinars to the dollar: this is the price licensed banks pay the Central Bank at the window.

The displayed premium is never negative. The Central Bank's own 2004–2012 market table occasionally records the market rate marginally below the official rate during the dinar-appreciation era — 511 days of the series. On those days the premium displays 0.0% and the margin counts zero; the raw recorded values are preserved in the tooltips and in the downloadable series. This is a conservative choice: it never credits the mechanism with a negative cost, and it never lets a negative day offset a positive one.

The 2018–2019 interval is the weakest part of this series and is stated as such wherever it is used.

Sixty-three session bulletins published between 8 November 2016 and 28 July 2020 carry the full per-bank breakdown and survive with an intact text layer. Every row was parsed from that text; no OCR was used. Each row was checked arithmetically: cash to bank plus transfers abroad must equal the printed row total. 1,732 of 1,743 rows pass; the eleven that do not are retained and flagged.

Each session was then checked against the session total the Central Bank printed at the foot of its own bulletin. 41 of 63 reconcile exactly. The remaining 22 are retained, flagged, and not adjusted to fit.

Bank names in the bulletins carry the orthography of the source documents and are reconciled to the register by token matching, with an explicit override table for ambiguous cases. Rows that cannot be reconciled with confidence are shown under their Arabic name and carry no enforcement flag.

Two window totals are printed on this site, and they are not interchangeable. The documented floor is $514,997,749,994: the sum of the sales figures recovered day by day from the Central Bank's own bulletins, across 3,787 of the 8,240 days in the series. Added to it are the Central Bank's own published annual totals for the years in which the daily recovery is thin, contributing a further $101,276,545,245 that is not double-counted against any recovered day. The two together give the verified total of $616,274,295,239 — $616.3bn. Both components are Tier A: each rests on a Central Bank publication.

The captured margin follows the same structure. Computed on the recovered days alone, it is $23,796,027,443 — the documented figure, Tier A, and the only one presented as fact. Because that margin is computed on $515.0bn of volume while $616.3bn is verified as sold, the documented margin understates the whole by the ratio of the two volumes. Raising it in that proportion (a factor of 1.2097) gives $28,787,173,757 — approximately $28.8bn. That figure is Tier B: an estimate, printed alongside the documented one and never in place of it.

The scaling is not uniform across years. For 2013 and 2014 the Bank's own published annual totals are roughly four times the recovered dailies, so those two years are scaled year by year rather than by the era factor: the margin becomes approximately $2.94bn for 2013 and $2.75bn for 2014, against $0.54bn and $0.69bn documented. The reconciliation exhibit and the tenure tables print both values for those rows.

The true window total is higher than $616.3bn. The Central Bank held sessions on days whose bulletins have not been recovered and for which it published no annual total this file has found; 2018, 2020 and 2025 are the years where that gap is widest.

A primary market-rate source for 2018–2019 has not yet been recovered.

The change in CBI gross foreign assets after 2018 has not yet been extracted; the reconciliation therefore ends at 2018.

Window recovery for 2018, 2020 and 2025 remains thin: the verified total rises when those bulletins are recovered.

The exact post-2003 split of the $52.4bn Kuwait reparations award is stated here as $34.9bn and awaits refinement against the UNCC's own instalment record.

Audited statements after financial year 2015 have not yet been obtained; the balance-sheet table shows those years as extraction in progress.

Per-bank ledger data outside the 63 fully published sessions has not been recovered; it may not have been published.

Ownership research is in progress for 85 of the 90 institutions in the register.