Guide

How to read the register

What each register is, what every field and code means, and how to follow a figure from the book to its document.

The three registers · Source fields · Claim fields · Specialist questions · How far a source was read · Claim status · Evidence tiers · Version history · Withheld quotations · Archive and working records · Codes · Live and snapshot · Data

The three registers

The source register is a record of documents: one row for each document the book opened or tried to open. The claim register is a record of assertions: one row for each quantitative, legal, institutional or current claim that more than one chapter relies on, or that is contested, superseded or otherwise unable to stand on its source row alone. The specialist register records the fifteen specialist reviews the book identified and did not commission. A claim names the sources it rests on, and those names are links. To start from a chapter, use By chapter.

Fields of a source row

Identifier: The row's permanent identifier. It never changes and is never reused.
Document: What the document is: its publisher, title and date.
Where to find it: The public address, where there is one, and whether a copy is held in the author's archive.
Kind of document: The kind of document, in the author's words.
Evidence tier: Its place on the six-level scale of kinds of document. See Evidence tiers.
Published: When the publisher issued it, as precisely as the document says.
First opened by the author: When the author first opened it.
Retrieval record: How it was retrieved: the date, the result and, for a failed retrieval, what was tried.
What the book relies on it for: What the book uses it for, chapter by chapter, and what it is permitted to support.
How far it was read: How far the author got with it. See How far a source was read.
Notes and amendments: The author's notes on the document, and every later amendment, oldest first. See Version history.

Fields of a claim row

Identifier: The row's permanent identifier.
The claim: The claim as the book states it, or a summary of it.
Where the book uses it: The chapters and passages that use it.
Status: One of eight values. See Claim status.
Source: The source rows it rests on. Each identifier is a link.
Kind of source: The kind of source.
Published: When the source was published.
Read by the author: When the author read it.
What is measured, and its limits: What is measured, over what period, on what basis and for what population. The most important field for checking a figure.
Conflicting evidence: Any evidence that points the other way, and how it stands.
Specialist question: A question only a specialist could settle, if there is one.
Other passages that depend on it: Other passages or rows that depend on this one.
Rule for using it: The rule for using the claim: what may and may not be said on it.
Row version: The row's own version number.

Fields of a specialist question

Identifier: The question's permanent identifier.
Field: The field of expertise needed.
Editorial issue numbers: The numbers of the editorial issues it answers.
Questions: What the specialist would be asked.
Who could answer them: Who could answer them.
Reviewer: Who reviewed it, or why nobody did.
Date: When its disposition was decided.
Outcome: What the book does instead, given that no review was obtained.
Status: How the question stands.
Chapters affected: The chapters it bears on.

How far a source was read

The source register keeps two scales apart. One records what kind of document a source is (its evidence tier); this one records how far the author got with it. A figure resting on an account of a document, rather than on the document, does not enter the book's prose.

READ: opened and read
READ_IN_PART: reached in part; the row names which part was read
PARTIAL: only part of the document, or only an account of it, could be reached; the row states the limit
VERIFIED: checked against the document itself
HELD_NOT_READ: a copy is held and has not yet been read; nothing in the book rests on it
TO_READ: retrieved; the reading is still owed
DISCOVERED: found by a search and not yet opened
RETRIEVED: retrieved to record that a route was run
TO_OBTAIN: not yet obtained; the retrieval is still owed
NOT_OBTAINED: retrieval was tried and failed; the routes tried and not tried are recorded
RETRIEVAL_FAILED: the publisher's site served nothing usable; the attempts are recorded
SEEN: its text was seen only in an extract or a search service, not read on the page; no figure rests on it
NOT_ESTABLISHED: checks were run and did not settle the question; the row records them
OPEN: a record that stays open, such as a list of documents still to find
REJECTED: a rejection record: the wrong document, kept so that it is not used again
SUPERSEDED, or SUPERSEDED BY a named row: replaced by a later row, which the row names

Claim status

The claim register uses eight values, and the discipline is in not collapsing them.

SUPPORTED: a primary document was opened and says this
QUALIFIED: the document says this within a limit, which must travel with the claim wherever it appears
CONTESTED: two comparable sources disagree, and neither has been superseded
SUPERSEDED: the same body has replaced it; the row is kept as part of the record
UNSOURCED: nothing stands behind the claim yet, which is a bar on printing it as fact
AUTHOR_ESTIMATE: the author's own calculation; the row carries the formula, assumptions, period and currency
SPECIALIST_REVIEW: it may be raised as a question and not settled as advice
AUTHOR_DECISION: the choice is the author's, and the row holds it open until it is made

Evidence tiers

The tier records the kind of document, from 1 (strongest) to 6. A source low on the scale establishes that a claim was made, never by itself that it is true. Some rows give a tier with a qualifier, such as one tier for a document and another for its metadata.

Tier 0: outside the hierarchy: a record kept so that a wrong document is not used again
Tier 1: a primary document held in the original, or an authenticated original
Tier 2: enacted law, official regulation, a court decision, an official dataset, an audited account, an executed contract or an approved funding document
Tier 3: an original peer-reviewed study or an official technical report
Tier 4: reputable investigative or specialist reporting with a named method and sources
Tier 5: an institutional press release or state media, which is self-reporting
Tier 6: secondary commentary, an aggregator, advocacy or an unattributed report

Version history

The register keeps changing after the book is fixed in print. A row's notes are kept as a trail, oldest first, and a correction is added to the trail rather than written over what was there, so a reader can see what a row said before and when it moved.

Withheld quotations

Long verbatim quotations from third-party documents are not republished here, because the permissions the book obtained are for the book. Each is replaced where it sat by a marker giving its length in words. Follow the citation to read it at the source.

Archive and working records

The documents the book rests on are held in the author's archive and not published, because most are third-party copyright material. Where a row names a file, the page shows a plain description in italics with a dotted line; hover over it to see the file's name. The JSON and CSV downloads keep the names as written.

held in the author's archive: A copy of the document is held. The row's other fields say where it came from and when; the public address, where there is one, is in Where to find it.
an internal working record: A record of how the book was made, such as a review, a search log or the basis for a correction. It is not evidence and is not published; the row states what it found.
the author's decision log: The log of every decision about the book, numbered D-1 onwards. Not published.
a draft of the book, a checking script: A working draft of a chapter, or one of the programs that check the book and the register. Not published.

Codes you will meet

SRC-123: A source row. Every one is a link to its page.
A claim identifier, such as DEPOSIT-001 or CH5-002: A claim row. Every one is a link to its page.
SR-001: A specialist question. A link to its page.
D-327: an entry in the author's decision log, which records each decision about the book and is not published.
AQ-87: a question put to the author for a ruling; not published.
MIN-2313: a minor finding held in the internal ledger for a later pass; not published.
AR-1441: an entry in the glossary of Arabic terms used by the Arabic edition.
v83, ar_v25: A version of a chapter: Chapter 5 v83 is its eighty-third numbered version, and ar_v25 the twenty-fifth version of its Arabic.
l.229: A line of a chapter in the version named beside it.
Class 1, 2 or 3: The kind of correction: 1 a mechanical substitution, 2 a correction of evidence on an opened document, 3 a rewrite.
Diff review: An independent check of a correction, by a reviewer who did not make it, before it is approved.

Live register and snapshot

This edition rests on the register as it records the evidence to 18 October 2026, the book's closing date; a document read after that day is dated where the book uses it and in the register. The snapshot is that state, read from the repository's history; the live register is what the rows say today. Where a row has moved since, the live page says so and the snapshot page says what it said before.

Data

Each register can be downloaded whole as JSON or CSV from the front page, with the same quotations withheld and every file name kept as written.