In June 2025, the World Bank changed the extreme poverty line from $2.15 to $3.00 a day. The day after the change, with nothing having actually happened in the real world, there were 125 million more people in the count: the estimate for 2022 jumped from 713 million to 838 million people.
It's the best recent example of something that gets in the way of any discussion about technology and poverty: the yardstick changes more than the world does.
The numbers, with the yardstick stated
Under the new line, $3.00 a day in 2021 international prices, the World Bank estimates that 10.3% of the world's population lived in extreme poverty in 2024. The projection for 2025 is 9.9%, versus 10.5% in 2022. The downward trend still holds; what changed was the baseline.
Part of the jump is inflation, since the price base moved from 2017 to 2021. Part is a real increase in the line itself: the World Bank now counts as poor people who previously sat just above the cutoff.
None of this is manipulation. It's the normal functioning of an indicator that needs recalibrating. But it means the statement "poverty increased" and the statement "poverty decreased" can both be correct in the same year, depending on which line is used — and whoever cites only one of them is telling half the story.
Where technology really makes a difference
Setting enthusiasm aside, the mechanisms with a measured effect are few and specific:
- Mobile payments. Wherever cheap digital accounts arrived, the ability to save, receive, and transfer money without a bank branch came with them. It's the most direct and best-documented effect.
- Information priced near zero. Lessons, crop diagnostics, market prices, and health guidance no longer require travel.
- Falling cost of service delivery. Services that once required an expensive professional now fit inside an app. That holds both for those who consume them and for those who can now sell them.
What technology doesn't do on its own is distribution. Cheap internet access doesn't fix land, credit, schooling, or governance, and the list of countries where the mobile phone arrived before sanitation is long.
The lesson that applies to your company
Before accepting any impact figure — on poverty, on productivity, on an AI model's performance — ask which yardstick was used and whether it's the same one from the previous measurement. That's exactly the problem that showed up at the launch of GPT-6, where the same test scored 62.7% and 99.9% depending on the measurement harness.
Inside a company the trap is the same: changing how you measure before and after a project makes any initiative look like a success. When we help someone with AI implementation, the first thing we agree on is the yardstick, and it's measured before the tool goes in.
Frequently asked questions
How many people around the world live in extreme poverty?
Using the $3.00-a-day line the World Bank adopted in June 2025, about 838 million people were living in extreme poverty in 2022 — 10.5% of the world's population. The estimate for 2024 stands at 10.3%, and the projection for 2025 is 9.9%.
Why did the number of poor people in the world suddenly jump?
Because the yardstick changed. In June 2025, the World Bank raised the extreme poverty line from $2.15 to $3.00 a day, switching from 2017 international prices to 2021 prices. That change alone added roughly 125 million people to the 2022 count — even though nothing in their actual lives had changed.
Does technology reduce poverty?
It cuts specific costs, and that effect is measurable. A digital wallet on a phone lets people save and send money without a bank branch nearby, health and market information no longer requires traveling to get it, and services that once needed an expensive professional now fit inside an app. What technology can't fix on its own is distribution: internet access doesn't hand anyone land, credit, schooling, or good governance.