The OWID Brief: Around half of… | Our World in Data Summary

1. Key Themes

Theme: Mental health burden is far larger than point-in-time statistics suggest

Lifetime prevalence is roughly 50%

Standard health statistics capture only a snapshot, which understates cumulative risk. A multi-country study shows how risk accrues over a lifetime.

"They estimate that roughly half of the population will experience one of 13 common mental health disorders by the time they reach 75."

Gender gaps differ by disorder type

Depression and anxiety skew female, while substance use disorders skew male.

"Depression is the most common; by the age of 75, one in three women has experienced depression at some point in their lives. One in five men has."

"For anxiety, the share is similar, as is the gender gap. For substance use disorders, it's the opposite."

The true figure may be higher still

The study's methodology likely undercounts.

"In this article, we look at these estimates in more detail and discuss why the true share might be even higher."

Theme: The future of global emissions hinges on China

Coal use outside China has been flat for ~40 years

The rest of the world has plateaued while China drove nearly all growth.

"Outside China, the world's use of coal has barely changed since the late 1980s. It was just over 20,000 terawatt-hours back then, and almost four decades later, it is still the same."

"But the change has been very different in China. The country has increased the use of coal enormously, and today, China is burning more coal than the rest of the world combined."

Climate trajectory depends on one country's decisions

"The future of global coal use, and the emissions that come with it, depends heavily on what happens in China."

Broader emissions signals are mixed

Global fossil CO2 is projected to decline even as coal dominance persists in places like Poland (headlines only).

"Global CO2 emissions from fossil fuels are set to fall in 2026"

"Solar and wind power have grown quickly in Poland, but coal still dominates"

Theme: AI is accelerating its own development, but the evidence is opaque

AI-assisted R&D speed-up at a frontier lab

METR assessed how much AI is speeding up Anthropic's own research.

"On a set of well-defined, easy-to-grade tasks, METR estimates Anthropic is now improving its models about twice as fast as between March 2024 and October 2025. It gives a ballpark estimate that 80% of this could be down to AI."

No visible bottlenecks, with possible further acceleration

"METR's researchers saw no major bottlenecks for the next 12 months, and warn the speed-up could still accelerate."

Transparency gap for third-party verification

"Unfortunately, Anthropic decided to redact much of the evidence METR found most informative."

"...find it very unfortunate that AI labs don't have stronger transparency obligations, and that we have to trust METR on such important questions."

Theme: Extreme poverty reduction is stalling

Historic progress was dramatic, driven by China and India

"In 1950, six in ten people lived in extreme poverty. By 2025, it was one in ten, even though the world's population more than tripled. Most progress since 1990 came from China and India."

Progress has slowed and could reverse

"But progress has slowed over the past decade, and as I've written in a recent OWID article, I'm worried that progress against extreme poverty might come to a halt."

"...new projections show the global share in extreme poverty falling to 1.2% by 2050 if every country cuts poverty as fast as South Korea once did, but if current trends hold, they expect little progress until 2030 and a rise by 2050."

Theme: Migration is economically significant but politically fraught

Economic role of migration

"In destination countries, migration can fill employment and skills gaps and contribute to economic growth."

"For example, remittances (money sent back to family or friends by immigrants) can be an important source of income in their country of origin."

Public concern and scale

"But migration is also among the top concerns for the public in many countries. In politics, it is often divisive and controversial."

"Hundreds of millions of people live in a country that is different from the one where they were born. In some countries, the majority of the population was born abroad."

2. Contrarian Perspectives

The AI speed-up may be a one-time boost, not an intelligence explosion While Ajeya Cotra argues for recursive acceleration, Toby Ord offers a more deflationary reading, and the newsletter author sides with him tentatively.

"Ajeya Cotra, also of METR, argues on her Substack that this could mean AI speeding up its own development faster and faster, and soon. Oxford researcher Toby Ord says the speed-up may be a one-time boost from training a bigger model, which the report can't rule out. I'm a bit more in Ord's camp for now..."

Headline poverty progress masks a stall, and "business as usual" implies reversal Against the celebratory long-run narrative (6 in 10 to 1 in 10), the forward-looking projections are pessimistic: gains were concentrated in two countries and have decelerated.

"...if current trends hold, they expect little progress until 2030 and a rise by 2050."

Mental health "prevalence" figures likely understate the true burden Conventional point-in-time estimates already seem large, yet the authors argue lifetime risk could be higher still.

"...discuss why the true share might be even higher."

3. Companies Identified

Anthropic

Description: Frontier AI lab
Why mentioned: Case study for measuring AI-accelerated AI research; criticized for redacting evidence
Quotes: "METR estimates Anthropic is now improving its models about twice as fast as between March 2024 and October 2025." / "Unfortunately, Anthropic decided to redact much of the evidence METR found most informative."

METR

Description: Organization that tests AI systems for risks
Why mentioned: Conducted the third-party assessment of AI's effect on Anthropic's research speed
Quotes: "From mid-August to late October, the organization METR, which tests AI systems for risks, had access to Anthropic's data and staff to assess how much AI itself is speeding up Anthropic's own research."

World Bank

Description: International financial institution
Why mentioned: Source of extreme poverty estimates and projections to 2050
Quotes: "World Bank researchers Christoph Lakner and Samuel Kofi Tetteh Baah sum up the Bank's estimates of extreme poverty since 1950, with projections to 2050."

Energy Institute

Description: Publisher of the Statistical Review of World Energy
Why mentioned: Data source for the global coal consumption chart
Quotes: "Source: Energy Institute, Statistical Review of World Energy (2026)."

4. People Identified

John McGrath

Description: Lead researcher on a multi-country mental health study
Why mentioned: His study underpins the lifetime-prevalence estimate
Quotes: "A study by John McGrath and colleagues tries to answer this question by drawing on data from 29 countries and over 150,000 people across the world."

Ajeya Cotra

Description: Researcher at METR who writes on Substack
Why mentioned: Argues AI could accelerate its own development faster and faster
Quotes: "Ajeya Cotra, also of METR, argues on her Substack that this could mean AI speeding up its own development faster and faster, and soon."

Toby Ord

Description: Oxford researcher
Why mentioned: Offers the counter-view that the speed-up may be a one-time boost
Quotes: "Oxford researcher Toby Ord says the speed-up may be a one-time boost from training a bigger model, which the report can't rule out."

Christoph Lakner and Samuel Kofi Tetteh Baah

Description: World Bank researchers
Why mentioned: Authors of the extreme poverty estimates and projections
Quotes: "In this blog post, World Bank researchers Christoph Lakner and Samuel Kofi Tetteh Baah sum up the Bank's estimates of extreme poverty since 1950, with projections to 2050."

Hannah Ritchie

Description: OWID researcher
Why mentioned: Authored the mental health article and co-led the migration topic update
Quotes: "(This topic update was led by Hannah Ritchie, Esteban Ortiz-Ospina, and Tuna Acisu.)"

5. Operating Insights

Use third-party audits and standardized, easy-to-grade tasks to measure AI productivity gains METR's approach of benchmarking on well-defined tasks over defined time windows offers a template for quantifying AI's real impact on R&D velocity.

"On a set of well-defined, easy-to-grade tasks, METR estimates Anthropic is now improving its models about twice as fast as between March 2024 and October 2025."

Plan for China-dependent scenarios in energy and climate-related ventures With rest-of-world coal flat, demand, policy, and emissions outcomes are largely a function of one market's trajectory.

"The future of global coal use, and the emissions that come with it, depends heavily on what happens in China."

Don't rely on snapshot metrics when sizing health or wellbeing markets Lifetime incidence (about half the population) is a very different addressable-need figure than point prevalence.

"People can develop mental health disorders at any point in their lives, or experience recurring episodes over many years."

6. Overlooked Insights

Crude vs. age-adjusted rates can tell opposite stories Aging populations can make headline disease burdens rise even as underlying risk falls, which matters for healthcare demand forecasting.

"Population aging is pushing up Chile's crude cancer death rate, even as age-adjusted rates are falling"

Demographic aging is already substantial globally

"One in ten people in the world is over 65 — how does this compare in the youngest and oldest countries?"