About the data
This is the first edition: thirty-seven chapters in eight parts, the identity scoreboard, the key statistics reference page, the glossary and the bibliography, and the nine interactive tools in the Explore menu, all of them built on the fifty-four country cards. Every number the book quotes from a country is a field on one of those cards or a key in the verified population anchors, and the print edition is set from the same files this site renders, so the two cannot disagree.
The edition is cut against fixed vintages, listed in full in the table below: World Population Prospects 2024 (medium variant), the IMF World Economic Outlook of April 2026, World Bank International Debt Statistics 2025, ILOSTAT modelled estimates of November 2025, the Boston University Chinese Loans to Africa release covering 2000 to 2024, Africapolis 2025, the World Bank FY26 fragility list, the ND-GAIN 2026 index, Afrobarometer Round 9, and the World Bank’s live World Development Indicators and Statistical Performance Indicators at the pull dates the last column records. Two sources this edition deliberately does not use, and says so where the absence shows: the only complete register of African Eurobond issues, which is licensed against commercial use, and the world motor-vehicle production tables, which are now behind a commercial-use licence of their own. What was not established in time is written out in words rather than estimated. The site is refreshed between editions; the book is not.
Every tool on this site, and every number the chapters quote from a country, reads the same fifty-four records: one card per UN-member African state. This page says what those records carry, which release each field comes from and when that release was last pulled, how the confidence flag is worked out, and what a blank means. You can download the whole set from the country cards page as cards.csv or cards.json.
The identity the cards are built around
From \(Y = (Y/E) \times (E/L) \times L\), where \(Y\) is output, \(E\) is employment and \(L\) is the working-age population:
GDP growth ≈ growth of output per worker + growth of the employment-to-working-age ratio + growth of the working-age population
All three are annual growth rates over the same window, 2014 to 2024 on the cards as they stand, which is what makes them add. The card carries each term, their sum (the implied rate), and the growth actually observed, so the gap between the last two is visible rather than hidden. Term one, working-age population growth, is on all 54 cards. Terms two and three are on 53: they come from the ILO’s modelled estimates, and term two is modelled employment aged 15 and over divided by the UN’s 15 to 64 population, a mismatch of age bands that the book states rather than smooths. The within-sector and between-sector split of term three is on 21 cards, the countries the Groningen database covers.
What the cards carry, and from where
| Block | What is on the card | Source |
|---|---|---|
| Demography | Population in 2024, 2050 and 2100, median age, total fertility, dependency ratio, working-age population growth | UN World Population Prospects 2024, medium variant, ages 15 to 64 unless labelled |
| Output | GDP in current dollars, GDP per capita, GDP at purchasing-power parity, observed growth, a five-year forecast | IMF World Economic Outlook, April 2026, 2025 levels; World Bank World Development Indicators where the IMF publishes no figure |
| Employment | The employment-to-working-age term, output per worker, wage-employment and non-agricultural shares, employment by sex, youth not in employment or education, informality | ILO modelled estimates, November 2025 |
| Structural change | The within-sector and between-sector split of output-per-worker growth | Groningen Growth and Development Centre, Economic Transformation Database |
| External debt | Total external debt stock and its creditor classes: multilateral, Paris Club, China, other bilateral, eurobond, other private | World Bank International Debt Statistics 2025 |
| Chinese lending | Cumulative loan commitments, 2000 to 2024 | Boston University Global Development Policy Center, Chinese Loans to Africa database |
| Data quality | Statistical Performance Indicators score | World Bank Statistical Performance Indicators |
| Data quality | Last census, last GDP rebase, last labour-force survey, IMF dissemination standard | Entered by hand, one citation per country per field, carried in each record’s _sources map |
Two things follow from the last row. The data-quality fields are not machine-fetched from a single register, because no single register holds them; each one is cited to the statistical office, the IMF release or the survey report it came from, and the citation travels with the number in the download. And the Chinese-lending figure is loan commitments rather than debt outstanding, which is a different quantity from the International Debt Statistics line beside it, so the cards set the two side by side rather than netting one against the other.
Vintages, and when each was last pulled
Each field in a downloaded record carries a source key in its _sources map. This is what those keys mean, which release each names, and the day that release was last pulled into the cards.
| Source key | Source | Vintage | Last pulled |
|---|---|---|---|
| WPP2024 | UN World Population Prospects | WPP 2024, medium variant | Not recorded |
| AFRICAPOLIS2025 | Africapolis | 2025 release with 2050 projections | 25 October 2026 |
| HOORNWEG2017 | Hoornweg and Pope | 2017, Environment & Urbanization, 2100 city projections only | Not recorded |
| WEO-2026-04 | IMF World Economic Outlook | April 2026 | 13 September 2026 |
| WDI | World Bank World Development Indicators | The live series, as they stood when last pulled; the urban share carries its own year after the @ |
25 October 2026 |
| IDS-2025 | World Bank International Debt Statistics | IDS 2025, December 2025 release | 25 October 2026 |
| SPI | World Bank Statistical Performance Indicators | The live series, as they stood when last pulled | 25 October 2026 |
| AU-AFCFTA-REGISTER | African Union AfCFTA signature and ratification register | The register’s own date follows the @; ratified means the instrument has been deposited |
25 October 2026 |
| PAPSS-COVERAGE | PAPSS Payments Network Coverage sheet | The sheet’s month and year follow the @; membership is its participant column, not inbound reachability |
25 October 2026 |
| WDI-FLOWS | World Bank World Development Indicators, cross-border flows | Remittances, foreign direct investment and net aid; the latest year each series carries for that country follows the @ |
25 October 2026 |
| USGS-MCS2026 | USGS Mineral Commodity Summaries | MCS 2026, with 2025 estimates | 25 October 2026 |
| CLA-2024 | Boston University, Chinese Loans to Africa | 2000 to 2024 release, November 2025 | 25 October 2026 |
| ILO-2025-11 | ILO modelled estimates | November 2025 | 25 October 2026 |
| GGDC-ETD | Groningen Economic Transformation Database | The release current when last pulled | 25 October 2026 |
| VIIRS-VNL21 | NOAA and EOG VIIRS nighttime lights | VNL v2.1 annual composite, with the year appended after @ |
Not recorded |
| SP-2026-02 | S&P Global Ratings, African sovereign debt outlook | February 2026 | Not recorded |
| WB-FCS-FY26 | World Bank Group list of fragile and conflict-affected situations | The FY26 list, published 8 July 2025; a country in neither of its two columns reads “not on the FCS list” rather than nothing, because the list covers the whole membership | 25 October 2026 |
| ND-GAIN-2026 | ND-GAIN Country Index, University of Notre Dame | 2026 release; the vulnerability score alone, never readiness, with the score’s year after the @. The tercile ranks a country among the African countries the index covers, not against the world |
25 October 2026 |
| AFROBAROMETER-R9 | Afrobarometer, Lived Poverty Index | Round 9, fieldwork October 2021 to June 2023 across 39 countries; each card carries its own country’s fieldwork year after the @ |
25 October 2026 |
| SHELF | The report shelf: IMF Staff Country Reports and World Bank country reports | No edition: what the Fund and the Bank had published in the three years before it was pulled, so the pull date is the vintage | 25 October 2026 |
| MANUAL | Entered by hand | The citation follows the @, one per field per country |
Not pulled |
| UNCTAD-TRADEMATRIX | UNCTADstat, Merchandise trade matrix (analytical) | The release on the pull date; the year of each card’s export figures follows the @ |
25 October 2026 |
Two columns, two different things. The vintage is the edition a number belongs to and does not move: the 2024 revision of World Population Prospects will say the same thing in five years as it says today. The last column is the day that edition, or that live series, was last read into the cards, and it is what tells you how stale the figure in front of you is. For a fixed edition the distinction rarely matters; for the World Bank series, which are revised continuously, the date is half the number’s address. Where the last column reads not recorded, the release is a fixed edition whose reading day was not kept, and the edition rather than the day is what the figure depends on. Every country card repeats these dates for the releases that card draws on, above its own list of fields.
A key ending in @2024 names the observation year, which differs by country wherever a source publishes on its own schedule. Not every key above is on a card yet: the Hoornweg, USGS and VIIRS series are listed because the book quotes them elsewhere, and the nightlights field on the cards is still empty.
Four of these releases carry most of the headline numbers, and every card page and the compare view now stamp them at the top: the World Economic Outlook edition, the day the World Development Indicators were pulled, the International Debt Statistics release, and the revision of World Population Prospects. The stamp is read from the same two places this table is, _data/VINTAGE.md and the pull date recorded beside each source’s own output, so a card, the compare view and this page cannot say different things.
The confidence flag
Three tests, all of them about the statistical system rather than the economy:
- a census within fifteen years;
- a Statistical Performance Indicators score of 70 or better;
- a labour-force survey within five years.
Pass all three and the card reads high. Fail one and it reads medium. Fail two or three, or have a census older than the year 2000, and it reads low. If one of the three inputs is itself missing, the flag is left uncomputed rather than counted as a failure, because not knowing when a country last ran a survey is ignorance about a statistical system and not evidence against it. As the cards stand that is 11 high, 20 medium, 20 low and 3 uncomputed, and 34 of the 54 have a labour-force survey within five years. The three uncomputed are not gaps in the research: the Central African Republic and South Sudan have never run a labour-force survey, and Eritrea has never held a census and has no statistical-performance score.
The flag is a summary, and summaries destroy information. Nigeria fails the census test on a 2006 enumeration, passes the survey test, and sits a point below the score threshold; the flag says low and does not say which of the three did the work. That is what the three fields underneath it are for, and they are all on the card and in the download.
What a blank means
A blank is a blank. Nothing on this site substitutes a zero, an average, or a neighbouring country’s figure for a number that is missing. Three phrasings are used, and they mean different things:
- not yet fetched: a series that no source has yet supplied for that country, or that does not exist for it;
- not recorded: a field that is entered by hand and has not been established for that country, such as a census year for a state that has not run one;
- not yet computable: a number derived from another that is missing, which is why a country can have an observed growth rate and no implied one.
In the JSON download a blank is null; in the CSV it is an empty cell. In the identity bars a missing term is drawn hatched, carries no width, and holds the sum back rather than passing as a flat zero. About a fifth of the cells across the fifty-four cards are empty today, and the pattern of the emptiness is itself one of the book’s subjects.
The shelf on each card
Under the data-quality section, each card carries a short shelf: the latest IMF Article IV staff report for that country, the latest Selected Issues paper where that paper is about labour or the public finances, and the latest World Bank country economic update or country economic memorandum where the Bank has published one. Each row links to the report and carries one line of what it argued.
The vintage is the day it was pulled, recorded in _data/shelf/_shelf.json and listed above under the source key SHELF. A shelf has no edition: it is a snapshot of what the Fund and the Bank had published on the day it was read, and it goes stale by publication rather than by revision. The window is three years, so “the latest” always means the latest of something recent enough to describe the economy in front of you.
Four rules, and each of them is the reason this is a shelf and not a news feed.
- The line under “what it argued” is the publisher’s own summary, quoted. It is a verbatim extract of the abstract the IMF or the Bank published with the report, cut at a sentence boundary, and the row says whose words they are. Nothing here is a paraphrase, and nothing is a summary written by a machine and dressed as the report’s. No paywalled text is read: the abstract is published beside the record.
- A summary that says nothing leaves the line blank. The IMF files many Selected Issues papers under the abstract “Selected Issues”; where that is all there is, the card says the summary says nothing rather than inventing a sentence.
- A country with no Article IV in the last three years says so, with a date. Five do: Senegal, Tunisia, the Central African Republic, Sudan and Eritrea. The row names the year of that country’s last Article IV consultation and links to it. Eritrea’s was in 2003. That is a fact about the relationship between a country and the Fund, and it is worth as much as a recent report; it is not a gap in this book’s research.
- A country with nothing to show shows nothing. No empty box, no placeholder row.
The IMF half is read from the Crossref record of the IMF Staff Country Reports series, which carries the Fund’s own titles, report numbers, abstracts and DOIs; the Bank half from the World Bank’s Documents & Reports service. Both are the publishers’ own machine-readable records of the same documents, and every row links back to the publisher.
The map’s industry layers
The nine industry layers on the Growth Pole Map do not come from the cards. Each is the site table its sector chapter’s own figures were drawn from, _figures/chNN/chNN_sites.csv, one file per chapter from 12 to 19 and from Chapter 35, and every site carries the source string and the claim-ledger identifier of the row it came from, both of them in its popup.
| Layer | Chapter | Sites drawn |
|---|---|---|
| Oil and gas | 12 | 40 of 40 rows |
| Critical minerals | 13 | 37 of 37 |
| Power | 14 | 34 of 34 |
| Auto plants | 15 | 27 of 27 |
| Financial centres | 16 | 32 of 32 |
| Cables and data centres | 17 | 40 of 40 |
| Tourism hubs | 18 | 38 of 38 |
| Ports and border posts | 19 | 35 of 35 |
| Data centres and AI power | 35 | 9 of 9 |
A row reaches the map only if its coordinates fall inside the map, its country is one of the fifty-four, and its claim-ledger identifier resolves. A row that fails is not quietly dropped: it is listed, with the reason, in the layer file the map reads, and the layer’s own note says how many rows are in that list. Nothing is in that list today: Chapter 18’s twenty-one rows citing ledger identifiers that had not been written yet now load with the rest.
Two things these layers are not. They are not a cadastre: a layer holds the mines, plants, ports and landings its chapter argues about, so the counts above describe a book’s attention and not the population of anything. And they are not on the country cards, so nothing here feeds the identity decomposition or the confidence flag. A blank cell in a sites file behaves the way a blank behaves everywhere else on this site: a capacity, a throughput or a year the chapter did not establish reaches the popup as words, never as a zero.
How a citation was checked
The bibliography’s year is the year of the document. Where one author has two works in the same year, a letter after the year tells them apart, and the text uses that letter. A bracketed [incomplete: ...] on an entry means a field is still unpinned. No page number in the book is a guess. Where a sentence rests on a publisher’s summary, a press account, or a promoter’s own figure, the sentence says so, and the bibliography note says so again. The claim ledger under the drafts records the stronger distinction, between a figure read off the source and a figure that is only qualified.
Card numbers are a different kind of check. Every figure the chapters quote from a country is a field on one of the fifty-four cards, regenerated from the release named in the vintages table, and the print edition is set from the same files this site renders.
Downloads and source
cards.csv: 54 rows, one per country, with the dotted field names the chapters quote.cards.json: the same 54 records, plus the_sourcesmap that says which release each field came from.independence_years.csv: independence year, date, former ruler and basis for each of the 54, used by Chapter 3’s data exercise.- The repository behind the site, with the chapter drafts and the code that builds every number, is not yet public. Everything needed to check a card against its source is in
cards.json, whose_sourcesmap names the release behind each field.
All 54 country cards · Compare cards · Growth Pole Map · Sovereign Debt Tracker · Convergence Calculator