Where every number comes from

Confidence grades

Every figure carries a grade. A: official source, no more than two years old. B: official but older, or an IMF forecast. C: compiled by a third party from official and modelled data, or calculated by dotzedw from several sources. D: a dotzedw estimate.

Rankings, medians and five-year change

Ranks and medians are worked out within the region you have selected, using each country's latest figure; where years differ, the year is shown next to the value. Shares and rates are compared in percentage points, levels such as GDP as a percentage change. Whether a change is good or bad depends on the indicator, and is said in words as well as colour.

Countries

Country figures come from the World Bank, the IMF, Our World in Data, Natural Earth and power plant databases. Each data card names its source, licence and the date it was retrieved. Where a country has no recent figure it is left blank rather than estimated.

Calculated figures

People without electricity = population × (100 − access rate) ÷ 100. Adults without an account = population aged 15 and over × (100 − account ownership) ÷ 100. GDP per person falls back to GDP ÷ population only where the World Bank series is missing.

Investor Lens scores

Each input is scored 0 to 100 by rank within the region (best = 100), inputs are averaged into pillars and pillars are weighted into an overall score. Ranks keep one extreme value from distorting everyone else. Missing inputs are left out and flagged, never filled in.

Maps and colours

Choropleths use an equal-area projection so large countries do not look more important than they are. Colour steps are quantiles for skewed data and even steps otherwise; values either side of zero use a diverging scale. All palettes are tested for colour-blind separation.

Industries and sectors

Industry values are value added at current prices in US dollars from the UN's national accounts, grouped by ISIC revision 3: agriculture (A–B); mining and utilities (C and E); manufacturing (D); construction (F); trade, restaurants and hotels (G–H); transport, storage and communication (I); and finance, real estate, public and other services (J–P). Growth uses constant 2020 prices. Values move with exchange rates, so compare shares across countries and real growth over time.

Each pillar's trade is the sum of the product codes (HS 2012) listed for its subsectors; a product can count in two subsectors. Deals are placed in a subsector from the words in the deal record; those that match none are shown as not yet placed.

Chains and their outlets come from All The Places, which reads brands' own store finders every week. We keep the latest month and one run a quarter since mid-2023. Store finders are added all the time, so growth and new chains compare like with like: only store finders that worked in both runs count, and a chain gaining or losing a third or more of its outlets (and at least 25) in one step is left out, as that comes from a store finder changing what it covers. The trend is today's count carried back by that growth. Independent shops are not counted, and growth is shown only where at least 50 outlets are compared.

Output per worker and output by province

Output per worker is value added in an industry (UN national accounts, current US dollars) divided by the people working in it (ILO modelled estimates, which the ILO builds from labour force surveys), in the seven industries the UN accounts use. The five-year change uses constant 2020 prices. The ILO does not cover Seychelles.

Output by province or region is quoted from each national statistics office in its own currency and prices, with shares of the national total. Stats SA, ZIMSTAT, INE Angola and Tanzania's National Bureau of Statistics publish it; ZIMSTAT gives only each province's largest industries, INE Angola three broad sectors and Tanzania totals only. Where nothing is published, the page says so.

Trade

CEPII's BACI database reconciles what exporters and importers report to UN Comtrade, product by product (HS 2012), in US dollars and tonnes. Stages of processing follow the UN's Broad Economic Categories. Each opportunity explains its own working.

Elections

Seats, rules and results since 2020 are compiled by dotzedw from electoral commissions, parliaments, court rulings and published results, each with its source. Earlier elections come from the V-Dem and V-Party datasets (Varieties of Democracy Institute, v16, CC BY-SA 4.0), which also supply the clean elections index and the share of women in parliament; figures derived from them are shared under the same licence. Where both cover an election, the key figures are compared on every data run. Vote shares are of valid votes; turnout is of registered voters.

City maps

Each city is measured over the same kind of area: its built-up urban centre in the GHSL Urban Centre Database with a ring around it, kept on its own side of any national border. The area is divided into a hexagonal grid (H3, cells of about 0.74 km²; about 5 km² on the free map).

Building footprint, counts and heights are Google's yearly satellite estimates for the middle of each year from 2016 to 2023, read at 4 m and added up for each cell. They show patterns, not plot-by-plot surveys, and very fast growth may be somewhat overstated. The map of buildings by year shows each 8 m square by the year a building first appears there. People are the GHSL population grid for 2025, at about 90 m.

Shops and services come from OpenStreetMap and Overture Maps, merged where the same place appears in both within 60 m (500 m for shopping centres), with chains named from brand tags and names; coverage is strongest for chains. Drive times use OpenStreetMap roads at typical off-peak speeds and 15 km/h for the last stretch from road to cell; walking uses roads and paths at 4.5 km/h. Shopping centre catchments reach 30 minutes' drive. City data is rebuilt every month.

Rents and prices

Few cities in the region have a regular house price or rent index, so the city pages quote published figures exactly as their sources give them, with the period and a link to the report, and say so where nothing is published. Where a report gives two figures for the same thing, both are shown. Each is checked when its source publishes again.