What Would It Cost to End Extreme Poverty?

Policy Context

The share of the world’s population living in extreme poverty—defined as living on less than $2.15 in 2017 PPP dollars per day—has declined dramatically, from an estimated 41% in 1981 to 8% in 2024. Still, hundreds of millions of people continue to live below very low consumption thresholds. This raises an immediate policy question: could extreme poverty be eliminated more rapidly through redistribution?

Study Design

Researchers wanted to understand the least expensive approach to target cash transfers to families experiencing extreme poverty all over the world. Unfortunately, limited data is available; many governments in low- and middle-income countries don’t have detailed or reliable poverty data which makes targeting interventions to the most vulnerable challenging. To bridge these gaps, researchers used advanced statistical methods to estimate measures of poverty based on assets they can easily check, for example, what housing materials, how many rooms it has, or the appliances a family owns.

Using real survey data from 23 countries, the researchers trained computer models to predict how poor a family likely is based on those easy-to-check details. Then they used those predictions to decide who should get cash and how much.

The research team tested two approaches: flexible cash transfers tailored to each family, versus a universal basic income (a fixed amount). They compared both to other ideas, like giving everyone cash regardless of need, or guessing poverty status with simpler/older methods. They also checked how close each approach gets to an ideal world where we know exactly how poor every single person is, and estimated what it would cost to roll this out worldwide.

Results and Policy Lessons

• Ending approximately all extreme poverty would cost around 0.3% of global GDP. These new estimates  capture how much it would cost to directly transfer enough income to each household to bring almost all of them above the $2.15/day poverty line.
• Costs in context. This figure is around 50% more than was (until recently) spent on foreign aid, but around 1/7 as much as the world spends each year on alcoholic beverages.
• Old data, new methods. By applying state-of-the-art statistical learning methods to the kinds of data governments already collect, the study is able to move beyond hypotheticals and provide benchmarks that can guide real-world decisions. The approach also demonstrates how artificial intelligence can be harnessed to address pressing social issues
• Smarter targeting saves billions. A universal basic income (UBI) set at the extreme poverty line would cost about five times the cost of the data-driven approach.  In most countries, data-driven targeting can thus reduce poverty by the same amount as UBI at a fraction of the cost.