I Corrected My Own Book

A standalone article drawing on my forthcoming book The Numbers That Do Not Hold. Part 1 of 13 in the series What the Record Shows. Every figure carries its publisher, unit and date; every calculation that is mine says so.

---

On 21 October 2025 the World Bank disclosed its assessment of what the war had done to Syria and what it would cost to rebuild. That same week my earlier book, Rewriting the Rules, reached the shops. Its first chapter told the reader that "rebuilding the country is expected to cost between $250 and $400 billion".

The Bank's figure was US$216 billion. Where my book needed a comparison it took the upper figure, and at that end it was out by roughly a factor of two.

I could have corrected the figure quietly in a later printing and said nothing. It seemed a poor way to argue for transparency. So I went back to check the rest, expecting to spend a fortnight updating numbers. This article is about what I found.

What the Bank measured

A number needs its sentence before it is used, so here is the sentence. The World Bank's The Syrian Conflict: Physical Damage and Reconstruction Assessment (2011–2024), dated August 2025 and disclosed on 21 October 2025, puts the cost of reconstruction at US$216 billion in December 2024 US dollars. It calls that "a conservative best estimate" and places it inside a range of US$140 billion to US$345 billion. It is the cost of rebuilding to current standards, net of what had already been rebuilt by 31 December 2024. It is not the cost of the damage, which the same assessment puts at US$108 billion, also in December 2024 dollars: a different measure of a different thing, what the war destroyed, valued as a loss to the country's pre-war stock of buildings and infrastructure.

Set my range beside that. The upper figure, US$400 billion, sits US$55 billion above the top of the Bank's range and is about 1.85 times its best estimate, which is my division. The lower figure, US$250 billion, falls inside the Bank's range, about 1.2 times the best estimate on the same division. My comparisons, with one exception I come to below, used the top.

Chart

One bill, two ranges

US$ billion. The World Bank's range for the cost of reconstruction and its conservative best estimate, the range Rewriting the Rules printed, and the Bank's figure for physical damage, which is a different measure

The range and its missing source

My book gave no source for the range. The same range, in the same figures, is in a commentary the Atlantic Council published on 3 January 2025, Diana Rayes's Dispatch from Damascus. It says that "the estimated cost of rebuilding the country ranges between $250 billion and $400 billion", figures that, "given the extensive damage to the country's infrastructure, appear increasingly accurate". It gives no source for them either: the sentence's one link, on the word rebuilding, leads nowhere. My book printed the range in October 2025 as what rebuilding was "expected to cost", citing nothing, so nothing in it told a reader where its bill came from, or that the commentary carrying the same range gives no source of its own.

Figures near US$250 billion had been in circulation since at least November 2017, when the United Nations Special Envoy told the Security Council that rebuilding would require at least that much. A number like that does not need a fresh source to feel sourced. It needs only to have been repeated, and that is the condition in which I printed it. The defect is mine: my book reported the range without saying where it came from.

The same floor turns up elsewhere, joined to a different ceiling. A European Parliamentary Research Service briefing of March 2025 says the cost of rebuilding "could range from US$250 billion to US$923 billion", and an International Labour Organization factsheet of October 2025 gives "USD250 to USD 900+ billion". The ceiling is not a cost of rebuilding at all, and how it became one is the next article in this series.

Five days

The range was not the only number carrying my argument. The first chapter of Rewriting the Rules also set the bill against "over $30 trillion in assets" of environmental, social and governance investment, and concluded: "if the entire ESG market were compressed into a single calendar year, Syria would only need to capture about five days' worth of that flow. Not months. Not weeks. Less than a week."

The arithmetic holds on its own inputs: US$400 billion against US$30 trillion, spread over a year, is 4.87 days on my division. The line worked. It was repeated back to me at conferences. It was also meaningless.

Assets under management are a stock: what is held at one moment, most of it already invested and owned by someone with a mandate. A reconstruction bill is paid out of a flow, out of budgets set each year by institutions that have to find the money without selling what they already hold. Dividing one by the other produces a figure with the shape of an answer and no content.

The pool would not hold still either. The Global Sustainable Investment Alliance's 2024 review, published on 11 November 2025, reports US$16.7 trillion, which looks like a collapse of nearly half. It is not one. The Alliance changed what it counts, a change it describes as a 50 per cent reduction in the scope of assets considered, and on the consistent basis it says the quantity rose by nearly US$5.5 trillion, or 49 per cent, over two years. The headline fell by nearly half while the thing it named grew by half. So the passage was wrong three ways at once: it compared a stock with a flow, it used a number whose definition was about to change, and it treated the pool as a fixed feature of the landscape. All three errors pointed towards optimism, which is the direction an author is least likely to audit.

The rule the book takes from it is simple to state. Before a financing comparison is made, both sides are identified as a stock or a flow, and if they are not the same kind of quantity the comparison is not made. That rules out most of the comparisons that make Syria's reconstruction sound easy, including several of mine.

Two kinds of error

What I found when I went back divided in two.

Some of the errors were mine: of reasoning, of scope or of omission, made with the evidence available and defensible only as honest mistakes. The five days is one. The range printed without its source is another.

The others were of a different kind. They were figures that had been sourced when I wrote them and had since come loose from their sources. Pages went down. Agencies restated numbers without saying they had. A citation that once led somewhere now led to another citation, and then to nothing. None of those figures was invented and none was taken carelessly. Each was in circulation, repeated by people with no reason to mislead, in documents that looked exactly like sources. Chapter 2 of the book follows three of them, case by case, and sets out what they have in common.

That second kind is not a confession. It is the finding. Syria's documentary record decays while people are using it, and a number can be properly sourced one week and unsupportable by the time anyone checks it. That matters to anyone deciding whether to lend, give or invest, because the institutions that could finance reconstruction at scale are the ones obliged to check. A development bank preparing a board paper asks where a number came from, in writing, and needs an answer that survives a legal review. A question nobody can answer is, to that institution, an answer.

Forty per cent

One claim sits awkwardly across the two. It anchored arguments in three chapters of my earlier book. "Here is a number that tells you everything about Syria's pre-2024 governance: 40%", it began. "That is the estimated portion of the national budget that disappeared through corruption each year. Not inefficiency. Not waste. Not bureaucratic bloat. Outright theft." It is also the exception I mentioned: the one passage that measured anything against the floor of the reconstruction range, saying that a decade of that theft "would have covered nearly 20% of the total reconstruction cost before the revolution even began".

I went to find where the forty per cent came from. On 11 August 2026 I searched, in English and in Arabic, the anti-corruption literature on Syria, the published budget reporting from 2022 onwards and the Arabic economic press, under corruption, embezzlement, public funds and budget leakage and their Arabic equivalents. I found no audit, no assessment and no study behind it, and that statement is only as wide as that search.

The nearest documented figures of that size are not theft at all. In the World Bank's Syria Macro-Fiscal Assessment of June 2025, budgeted subsidies were 41 per cent of total budgeted expenditure in 2021 and 42 per cent in 2022, against 21 per cent in 2018 and 2019, 9 per cent in 2020, 30 per cent in 2023 and 18 per cent in 2024. Two years of seven, and subsidies measure a distortion in what the state spends, not money that vanished. The two should not be allowed to borrow each other's authority.

The forty per cent did not come loose from a source, because I have not been able to establish that it ever had one, and that is worse than either kind of error: I printed a figure in three chapters without ever having held the document behind it.

What the corrections became

So the fortnight became a second book. The Numbers That Do Not Hold is not a new edition of Rewriting the Rules. It starts from the correction and asks a different question: what Syria would have to be able to show, in title, accounts, definitions, institutions, payment channels, project economics and responsibility, before its need for reconstruction can be financed. Evidence is one constraint among several, and the book does not pretend otherwise: security, law, currency and whether a project earns anything can each stop a transaction on its own.

The corrections are in it, all of them gathered at the back and most stated where they arise. They run from the bill and its missing source to a blockchain land registry I proposed and now withdraw, and to export credit agencies I said could insure Syrian risk, whose own pages offered no cover for Syria on either side of my earlier book's publication. The book gives them no total, and Chapter 2 explains why it prints no count of the failing claims it examines either.

Two of my own earlier articles on this site used the same bill, one the whole range and one its top, US$400 billion. Each now carries a dated correction note at the top, with the World Bank's figure, and its text is otherwise as published.

The evidence register behind the book is public, with each source, its date and what remains unverified, and it is online before the book. The rows behind this article are there, among them the Bank's figure and the forty per cent with the search behind it.

"Syria has no shortage of numbers," Chapter 1 says. "It has one number produced by a full assessment of the whole conflict period, and that is the one almost nobody quotes with its sentences attached." What that number can bear, and what would have to be added to it before anyone could lend against it, is where the book begins.

---

What the Record Shows, part 1 of 13. This article is drawn from my new book, The Numbers That Do Not Hold: Evidence, Risk and Financing Syria's Reconstruction, which comes out in November 2026, in English in paperback and on Kindle and in Arabic on Kindle. Its first chapter asks what the World Bank's US$216 billion measures, what it does not, and what would have to be broken down before it could become a financing plan. The book's public evidence register is already online. If you want to know when the book is out, the book page has a waiting list: one email, on the day it goes on sale.

The forthcoming book is The Numbers That Do Not Hold; passages drawn from it are from its current approved chapters and may change before publication.