Shareable intelligence

Madagascar

4 objects, each built so its context cannot be cropped away. Every one states the scenario range rather than a single value, carries the confidence level and the date the evidence was last checked, and links back to the full page it came from.

Copy a card's link to share it. Screenshot one and the qualifications come with it, because they are inside the frame rather than under it.

Country outlook · Eastern Africa

Madagascar

Three scenarios, two horizons. The range is the finding — a single number would imply knowledge nobody has.

Measure	Latest observed	2031	2036
Output per person	$599 2025	$525 – $673	$471 – $742
Electricity access	43% 2024	50% – 58%	54% – 66%
Internet use	19% 2024	25% – 32%	29% – 40%

Growth bands: -2.17% · 0.21% · 1.96% a year per person, from Madagascar's own 15-year record (2011–2025) — mean ± 0.85 standard deviations, clamped to its own 5th–95th percentile. Not an outside estimate of what this country might do.

Who this outlook leaves out

A cohort that reaches working age faster than jobs, training places or clinics are built
Rain-fed farming households, first and hardest, in a bad season
A state that must regulate technologies it has too few trained people to assess

2031 versus 2036 · Madagascar

Two horizons, asked the same question

By 2031

Household electricity access moves from 42.5% of the population (2024) to between 50% and 58% by 2031, depending on the scenario. Simple continuation gives 53% — a rise of 10.6 percentage points.
In people rather than percentages: about 17.6 million would still be without it in 2031 under continuation — 1.2 million fewer than in 2024, because the population grows as the share improves.
Internet use moves from 18.7% of the population (2024) to between 25% and 32% by 2031, depending on the scenario. Simple continuation gives 28% — a rise of 9 percentage points.
In people rather than percentages: about 27.1 million would still be without it in 2031 under continuation — 483,989 more than in 2024, because the population grows as the share improves.
Output per person spans $525 to $673 by 2031. The $148 between them is under a quarter of today's $599.
That spread is the distance between scenarios, not a margin of error — and it is a national average, which does not identify whose income moved.
The population reaches about 37.5 million — roughly 4.8 million more people than in 2025, about 22.1 million of them of working age.
This figure is the same in all three scenarios. Most of the people who will be alive in 2031 have already been born.

By 2036

Household electricity access moves from 42.5% of the population (2024) to between 54% and 66% by 2036, depending on the scenario. Simple continuation gives 60% — a rise of 17 percentage points.
In people rather than percentages: about 16.8 million would still be without it in 2036 under continuation — 2 million fewer than in 2024, because the population grows as the share improves.
Internet use moves from 18.7% of the population (2024) to between 29% and 40% by 2036, depending on the scenario. Simple continuation gives 34% — a rise of 14.8 percentage points.
In people rather than percentages: about 27.6 million would still be without it in 2036 under continuation — 1 million more than in 2024, because the population grows as the share improves.
Output per person spans $471 to $742 by 2036. The $271 between them is a quarter to a half of today's $599.
That spread is the distance between scenarios, not a margin of error — and it is a national average, which does not identify whose income moved.
The population reaches about 41.6 million — roughly 8.8 million more people than in 2025, about 25 million of them of working age.
This figure is the same in all three scenarios. Most of the people who will be alive in 2036 have already been born.

What has to happen first

For the Acceleration band to describe the decade, growth per person has to hold near 2% a year. This country has reached that rate before — its best year in the window was 2.1% — but has not sustained it across the 15-year record the band is drawn from.

What the growth model cannot see · Madagascar

Climate exposure & disaster risk

Binding constraint — This, not the growth model, is the operative question.

117,000 new displacements from disasters were recorded in 2023. Freshwater withdrawals stood at 11% of renewable resources (2022). 69% of employment is in agriculture (2025). Climate is not an environmental topic here — it is the household income of most of the workforce, and a bad season is a nationwide wage cut.

Why the band cannot contain it

Climate impact is non-linear and arrives as discrete events. A decade of trailing growth contains the droughts that happened, not the ones now becoming more likely.

What would invalidate the modelled band

A multi-season drought or a major flood event affecting the main producing regions. Discrete climate events are outside any band built from recent averages.

Assessed in words and deliberately not converted into a probability. This platform does not publish likelihoods it cannot derive from evidence.

AI evidence state · Madagascar

5 of 20 dimensions rest on a document or a measurement

3 dimensions have not been examined at all — which is a statement about this platform's research, not about Madagascar. There is no readiness score here: ranking countries by a weighted sum of twenty evidence states would manufacture a number nobody measured.

4 Measured A published measurement exists for this country and has been read.

1 Documented A policy, law, contract or programme document exists. No measurement of its effect has been found.

5 Inferred Assessed from adjacent indicators. Nothing AI-specific was found for this country; the reading is circumstantial.

7 Not evident Searched, and nothing was found. Informative — but the absence of a published record is not proof that nothing exists.

3 Unexamined This platform has not looked. This is a statement about the research, not about the country.

Full brief for Madagascar · Open in the atlas · Method · How to read the labels