A behind-the-scenes look at data to measure poverty
Digest #57
A note to our readers that the editors and authors of the #DataValues Digest will be taking a break from publication through the end of August to plan for future editions. If you’ve got an idea or you’re interested in submitting a post to the Digest, please email jmclaren@data4sdgs.org.
We are surrounded by consternation regarding the situation of the most disadvantaged people in face of the pandemic, cost of living increases, the global recession, debt burdens and inflation, conflicts and climate crises. But, as I read the speeches and glimpse the anxiety sparked by them, I wonder, quietly and a bit wistfully, when we will be able to report confidently on the current reality.
My colleagues and I measure multidimensional poverty. The key insight of multidimensional poverty is that poor people’s lives are complicated, precisely because several things are going wrong at the same time. As Nobel Laureate in Economics Amartya Sen observed tenderly, “human lives are battered and diminished in many and various ways” at the same time.
How do we measure this? With economist and scholar James Foster, we innovate on the ‘counting approach’—which has underlain monetary poverty measures since the early 1970s—to answer the question “who is poor?”. We look at different problems that afflict the same person or family and count how many things are going wrong at the same time. Some issues are more important than others, so they weigh a bit more. If a person is burdened by a critical mass of deprivations, they are classified as multidimensionally poor.
Think with me, for a moment, about the data needs for this project. Multidimensional poverty metrics use data about basic issues of poverty for a single person or a single household. Maybe someone in the household is undernourished, or a child is out of school, or no one at all has completed six years of schooling, or they lack drinking water, adequate sanitation, electricity, quality housing, or clean cooking fuel. These are among the issues we track with our colleagues at the United Nations Development Programme (UNDP) when we measure the acute global multidimensional poverty index (MPI) across over a hundred developing countries and find that 1.1. billion people are poor, defined as being deprived in one-third or more of the weighted indicators.
The findings that emerge from our 2023 study—which in the case of the global MPI covers 110 countries and 6.1 billion people—are striking.
For example, we observed that 25 countries cut multidimensional poverty by half in four to 12 years, including India and China. We noted that, in 15 years, 415 million people left multidimensional poverty in India and that the poorest groups—children, poor states, disadvantaged castes, and tribes—experienced the fastest reduction. But, across the world, 824 to 991 million poor people were deprived in sanitation, housing, and/or cooking fuel—many in all three at the same time—and that 600 million people were undernourished or lived with someone who was. And it’s hard to fathom that over 80 percent of multidimensionally poor people in Sub-Saharan Africa, which is home to around half of these poor people, don’t have access to electricity at home—not even a solar light bulb.
So, naturally, we want to know: How has the pandemic affected poverty across these 6.1 billion people? In the early days of the pandemic, we, like many, pivoted to make projections and simulate impacts on poverty to set our best guess. Later, we analyzed the wonderful phone surveys that the World Bank and UN agencies carried out to use those data to the best effect possible. But their coverage of the poorest is patchy.
Other poverty questions surge, too: Is the world set to halve the global MPI, which is a target during this Third Decade for the Elimination of Poverty? Will UNDP have reduced poverty for 100 million poor people by 2026? How many people plunged into poverty due to recent polycrises? To answer such questions, we need data for the poor and vulnerable populations. Then we can tell if poverty has gone up or down, where poverty is worst, and which deprivations constitute poverty for a group or in a place, so that actors can respond quickly and energetically to intercept and push back human harms.
But there’s a problem. The global MPI is built from data from household surveys that we access for free—mainly from Multiple Indicator Cluster Surveys of UNICEF and Demographic and Health Surveys of USAID, plus national surveys. We’re so grateful for these free, high quality, detailed data that can be disaggregated into 1,281 subnational groups and by other variables, from rural/urban areas to gender of the household head. Every year we update the global MPI using any new surveys.
So for how many of the 110 countries do we have data that were gathered in 2021 (which was still during the pandemic in many places) or 2022? It’s 7 countries. For 103 countries, the data are older. Why? Is it just too expensive?
When the late and beloved economist Sir Tony Atkinson chaired a World Bank Commission on Monitoring Global Poverty, he recommended a non-monetary adjusted headcount ratio (MPI) be used alongside monetary measures to monitor poverty. Speaking to the data required, he rightly observed that the stakes are higher because “it is necessary to have a data source at the level of the individual or household covering all relevant dimensions.” But he immediately counted the number of required questions, “The multidimensional poverty indicator for Colombia is based on some 38 survey questions, that for Pakistan on 54 survey questions, and for Costa Rica on 77 survey questions.” He observed that monetary consumption poverty measures require more questions per household. In one example, “the 1993–94 survey for Cambodia had a detailed consumption recall list of some 450 items.” Therefore,” Atkinson wrote, “it should not be assumed that a nonmonetary approach is more data-demanding.”
That is an important point because most people presume that multidimensional poverty measures require more, not less, data than monetary poverty measures. Inspired by Atkinson’s study and crestfallen by the paucity of post-pandemic data, this year we, with UNDP colleagues, counted the questions that underlie the global MPI we apply to surveys representing 6.1 billion people. It’s 43 questions. In fact, we only use about 5 percent of the questions on the wonderful surveys that we freely access.
But the hard fact remains that post-pandemic data on these 43 questions are not available for nearly all of the 110 countries we cover.
For Niger, the poorest country in the MPI, the most recent survey was from 2012. This year we dropped South Sudan and Burkina Faso, our next poorest countries, from online data tables as their data—from 2010—is out of date.
Data providers globally are storming ahead with innovations or wrestling to manage data deluge. This era will be remembered for breakthroughs—on AI but also satellite, administrative data, and cell phone data. We have worked on poverty with all these sources. But when it comes to the acute poor across 110 developing countries, the data fall short. So where are the data scientists, the visionary engineers, the creative funders, who are investing in a labor-intensive employment-creating opportunity of gathering high quality household survey data on the poorest, and/or making data breakthroughs on our poverty data bottleneck?
One of our team pointed us to Forbes’ site on real-time billionaires. It updates their holdings every hour. We yearn for the data needed to build such a site—updated just once a year, not even once a day —with and for the poorest on the planet.
 | A guest post by
|