Electricity · countries · –

Who actually decarbonised, and how?

Almost every country claims a transition. Only some changed what comes out of the socket. This reads the low-carbon share of electricity for every country with data at both ends of the period, ranks them by how far they moved, and then asks the harder question: which lever did the moving.

Countries ranked: of in the panel
Excluded, and named: lack
Observations: country × year
Years in panel: to
Query engines: 0bytes of WebAssembly

Where everyone stands

Of the countries in the panel, have data at both endpoints. The without — including Ukraine — are listed at the foot of this page rather than silently dropped. is thinner than its neighbours (195 countries against 212 in 2022): a gradient, not a clean cliff.

The spread is the story: countries gained percentage points or more over the period, and of the ended lower than they started. An average over the set conceals both.

Movers and fallers

% of electricity · change in percentage points

Figures are set in a monospaced face with tabular figures, so the columns align on the decimal. An empty cell would mean unknown, never zero.

Teal rises, oxide falls, and a country whose share did not move at all is drawn muted rather than pushed to one side or the other. These are the only two hues on the page and both mean something. The colour is also redundant: the direction of each line already says it, and so does the Change column above — which is what makes it admissible at all, and why the seven pathways below carry no hue whatsoever.

How they got there

Each country gets exactly one label from a deterministic rule: did its low-carbon share fall (went backwards), barely move (stalled), or rise — and if it rose, which of four drivers contributed most. There is no clustering and no random seed. Countries missing a primary series at either endpoint are labelled insufficient data and shown as their own group rather than dropped.

gas_swap is orthogonal: a country can add renewables and switch coal for gas, and forcing that into a mutually exclusive category would misdescribe it.

The eight groups

countries, n =

The bar is weight, never identity: the label carries which group this is, and the count carries how big it is. With seven categories a hue channel would exceed what is safely distinguishable and would fail outright in greyscale.

Pathways and swaps

countries per group

Method note — stated, not buried

A note on the "other low-carbon" driver. Where a country's other_renewables series is missing, this driver is inferred as the part of the low-carbon change the four named drivers do not account for, rather than being left blank. That is an assumption, not a measurement: it is what lets China, India, Russia and 17 others appear here at all. It changes which countries are shown far more than which label they get — of the seven countries labelled other-low-carbon-led, only Thailand relies on the inference.

How fast, and from when

Section 01 uses single-year endpoints, which are noise-sensitive. Drag the window below and the ranking, the trend fits and the leaderboard all recompute — in plain JavaScript, over rows already in memory. Nothing is queried; there is no engine in this page.

The -country set is held constant across every window, so different windows rank the same countries and "who improved fastest" stays comparable. The table and the chart below are a bounded view: the top of those , ranked by change over the window. The other are recomputed on every brush and simply not drawn.

The window always ends at 2024 and starts no later than 2023. A one-year window is not a window: every country's change would be exactly zero by definition and there would be no line to fit, so the slider stops one year short rather than offering a state that reads as "nobody improved".

Fastest improvers

– · percentage points, and pp per year

An empty cell above means unknown, never zero — the reading is missing, not zero percent. No cell is empty on today's data; every one of the countries has a reading for all 25 years.

Who is missing, and why

of panel countries have data at both 2000 and 2024. These do not, and are excluded from sections 1–3. This is the list §1 promises: named, not silently dropped (§4.5). Every count here and in the prose above is read from the same loader artefact, so the page cannot quote two different numbers for the same fact.

Named exclusions

of panel countries

Kosovo is in the source file but carries no ISO code, so the panel filter excludes it before this list is built — it is not one of the . A future refresh would need to assign it a code, not add the country. (The design spec's §10 records Kosovo as "absent from the source"; it is not, and tests/test_panel.py pins the correction.)