Joshua Blumenstock has spent much of his career alleviating the pain of poverty in countries such as Togo, Afghanistan, and Bangladesh, project by project. But with advances in AI-driven machine learning, Berkeley economists have set their sights on a more comprehensive goal: calculating how much it would cost to eradicate extreme poverty around the world.
The answer, according to a new research paper, is $318 billion a year, or 0.3% of global gross domestic product, enough to lift hundreds of millions of people out of extreme poverty. That’s a huge amount of money, but Blumenstock and his co-authors found that it’s still tiny compared to what humans spend each year on alcohol and cosmetics.
“Our hope is to come up with more realistic estimates of the costs of eradicating extreme poverty through direct cash transfers so that lack of realism does not become an excuse for inaction,” said Blumenstock, co-director of the Center for Effective Global Action (CEGA) at the University of California, Berkeley. “We now have the most accurate numbers possible. And it’s actually not that big, so that’s reassuring.”
“The numbers tell us that it’s not crazy to set your sights on big, ambitious goals,” said co-author Paul Niehaus, a professor of economics at the University of California, San Diego. “One reason for this is that extreme poverty has already fallen significantly in recent decades…but also because advances in data science have made it possible to design out-of-the-box policies that maximize support for the poorest people.”
This research is innovative not only for its estimated cost but also for using AI-driven machine learning statistical analysis to solve such a serious human challenge. This powerful technology helps policymakers identify the world’s poorest people based on their living conditions, spending habits, assets and other detailed surveys of poverty. An application of machine learning devised by researchers allows them to assess with unprecedented precision the amount of aid needed to bring people up to $2.15 per day, a benchmark commonly used by international organizations to define extreme poverty.
The research report, “What does it take to end extreme poverty?” was published online by the National Bureau of Economic Research in December.
Blumenstock is a professor at the University of California, Berkeley, School of Information and Goldman School of Public Policy. In addition to Niehaus, other co-authors include CEGA data scientist Leo Selker and Stanford University’s Roshni Sahoo and Stefan Wager.
Overall, the researchers looked at poverty in 23 low-income countries for which detailed economic data is available at the household and individual level. It concluded that extreme poverty rates in these countries, which account for about half of the world’s poorest people, could be reduced to less than 1% of the population with an annual investment of $170 billion.
They then applied it to the world’s other poorest countries and concluded that such poverty could be nearly eradicated with annual investment of just 0.3% of the world’s annual economic output, or $318 billion. By comparison, they write, this is slightly more than has been spent on foreign aid in recent years. However, this is only a fraction of the annual global spending on alcohol, which is 2.2% of gross global economic product, and the annual spending on cosmetics, which is 0.6% of gross global output.
“This could make ending poverty completely more affordable than ever before,” they write on the CEGA website.
A long-standing goal without a clear path
Ending extreme poverty has long been a goal of development agencies, religious groups, and civil society organizations. Ending poverty by 2030 is listed as the number one Sustainable Development Goal by the United Nations, and this goal has also been recognized in the foreign policies of some Western countries. Major financial institutions are tracking the eradication of global poverty. In 2022, the World Bank has raised the extreme poverty threshold to $2.15 per person per day.
We want to be optimistic, even in those moments when the odds seem to be against us.
Joshua Blumenstock
Effective domestic policies in low-income countries, combined with economic growth and aid programs, have dramatically reduced extreme poverty. Extreme poverty has fallen from 41% of the world’s population in 1981 to 8% in 2024.
However, improvement has been slow. In recent years, voters in the United States and other countries have objected to foreign aid spending. So how can we foster continued progress?
Many experts call for a universal basic income with policies designed to keep individuals and families from falling below a certain floor. But that’s an inaccurate way to provide assistance. Blumenstock and co-authors write that this approach can be very expensive because flat benefits give some households more money than they need to get above the poverty line.
Now, advances in advanced machine learning tools give policymakers new options to tailor the aid sent to households to more accurately meet their needs.
Countries disadvantaged by persistent poverty are now collecting fine-grained data using longitudinal surveys to assess household income, land ownership, family size, education, food consumption, housing construction characteristics, and other details.
Make assistance more accurate and efficient
New research shows that rapid advances in AI-driven machine learning are enabling policymakers and aid agencies to perform detailed statistical analysis on existing survey data to more accurately assess aid needs and goals.
Blumenstock gave the example of Togo, a small West African country that has worked closely with CEGA. The 2019 Harmonized Survey on Household Living Standards collected a wealth of information on the living conditions of approximately 6,000 households. Based on these results, policymakers in Togo will be able to directly calculate which of the surveyed households are suffering from extreme poverty and how much aid is needed to lift them above that threshold.
But a key challenge is understanding the poverty and needs of the millions of households that did not participate in the survey. Machine learning allows policymakers to identify a number of easily observable criteria, such as house size, number of rooms, roofing material, and even how the house looks on satellite images, that, taken together, can provide a good indication of a household’s underlying needs.
These metrics can be used to determine how much aid to allocate to many people across the country who did not participate in the study.
“Statistical machine learning can help policymakers understand how to make more nuanced decisions about how much each household should receive,” Blumenstock said. Phone-based banking software is then used to send relatively accurate payments to each household, helping them get above the extreme poverty line.
Ron & Keta via Flickr
The authors write that there is evidence that injecting cash into extremely poor communities can boost economic growth. And compared to a universal basic income approach, targeted aid would save hundreds of billions of dollars.
Political challenges and reasons for optimism
However, the new working document includes some caveats, but focuses primarily on costs.
Most low-income countries cannot afford to make these payments without external support. At the same time, the authors acknowledge that the United States and other governments are significantly reducing foreign aid starting in 2023.
Adding to the burden, the targeted aid payments they detail are meant to supplement, not replace, existing aid programs. And if international organizations agree to the World Bank’s decision last year to raise the threshold for extreme poverty to $3.00 per person per day, the expected cost of ending extreme poverty will rise significantly.
But the important value of this study is that it simply shows that the costs of eradicating extreme poverty can be relatively manageable if policymakers use new tools to make aid payments more precise and targeted.
The authors calculate that for an American with an average annual income of $45,000, the cost of eliminating extreme poverty in this way would be about $135 per year. They also consider a scenario in which billionaires invest to wipe out poverty in their countries, alleviating human suffering and establishing a noble personal legacy.
“Even in those moments when it looks like the odds are stacked against us, we want to be optimistic,” Blumenstock said.
