What this covers
Bruce Meyer of the University of Chicago makes a systematic challenge to recent claims of millions of Americans living in extreme poverty on $2–4 per day. Conversing with Russ Roberts, Meyer argues that these figures rest on flawed survey methodology rather than actual living conditions. He contends that linking survey records to administrative data—tax filings and government program records—reveals that most people classified below $2 or $4 daily income actually have substantially higher incomes than surveys alone report. When in-kind benefits like SNAP and housing assistance are counted, measured extreme-poverty numbers drop by more than half.
The conversation ranges across measurement problems in the Current Population Survey, where response rates are falling and income under-reporting is endemic. Meyer documents that the material conditions of America's poorest have improved dramatically—90% of the bottom fifth have air conditioning, and their housing size now matches what the middle fifth occupied 30 years ago. He challenges Kathryn Edin and Luke Shaefer's estimates and argues his best count is around a couple hundred thousand in extreme poverty, likely an overestimate. The discussion also examines Consumption vs Income as measures, why Sensitivity to Assumptions in research on intergenerational mobility produces starkly different conclusions than widely reported, and Meyer's case for Targeting vs Universality—arguing that a tailored safety net with caseworkers beats Universal Basic Income because people's needs are heterogeneous and cash alone cannot address addiction or mental illness.
Meyer argues that headline claims of millions of Americans living in 'extreme poverty' are artifacts of survey error and incomplete income measurement; once administrative data, in-kind benefits, and assets are accounted for, near-zero Americans live on $2-4 a day, and the existing targeted safety net reaches more people than official numbers suggest.
- Linking survey records to tax and program administrative data shows most of those classified below $2/day actually have far higher income
- In-kind benefits (SNAP, housing) and assets must be counted, and doing so cuts extreme-poverty counts by more than half
- A targeted, human-administered safety net is preferable to Universal Basic Income because needs are heterogeneous
Survey income data severely under-reports actual resources available to low-income households.
- The Edin/Shaefer and Deaton extreme-poverty studies start from surveys that generally do not include the homeless, and Meyer's rebuttal uses those same surveys; the homeless are studied separately using the decennial census and American Community Survey.
“the Edin and Shaefer work and the Deaton work starts from surveys that don't generally include the homeless.”
- People are less willing to discuss income in surveys than their sex lives, and combined with falling response rates and rushed interviews, this leads to large under-reporting—for example, about half of pension recipients identified in 1099-R tax records did not report pension income in the survey.
“Income turns out to be more sensitive in surveys than people's sex lives. They're more willing to talk about sex than their income.”
- By linking anonymized individual census survey records to tax and government program data, Meyer's team shows that only a small fraction of those classified as below $2 or $4 a day cannot be ruled out as having much higher income; the vast majority actually have far higher incomes than the survey alone reports.
“We take the survey data that these authors, in fact, were using, and we link it to tax and government program data at the individual level... we see that a tiny fraction of these individuals can't be ruled out as having much higher income--only a small fraction.”
- The widely cited 50% figure for the share of people who can expect to be better off than their parents is misleading because it doesn't account for family size or for the under-reporting of income in the CPS; correcting these would raise the share, yet the figure remains highlighted on the Opportunity Insights website.
“when you look at their data in their Appendix, they say, 'Well, actually if you change this assumption and that assumption, it's not 50% are going to do better than their parents: It's 70%.' Well, 50% and 70% are really different.”
- The current safety net already reaches a much larger fraction of people and reduces poverty more than official numbers indicate, because large fractions of recipients of SNAP, Unemployment Insurance, cash welfare, pensions, and Social Security/SSI do not report it in surveys—so taking survey reports at face value falsely suggests many people are falling through the cracks.
“the current safety net already reaches a much larger fraction of people than official numbers indicate. And the poverty reduction of existing programs is much greater than officially advertised.”
Official poverty measures fail to account for government programs and tax benefits that reduce deprivation.
- The most common definition of extreme poverty is living on less than $2 per person per day, though some use $4 a day, and standard definitions usually do not account for taxes and in-kind benefits.
“the most common definition of extreme poverty is those who are living on less than $2 per person, per day. Others have used other definitions--less than $4 a day. And, others have accounted for taxes and in-kind benefits. The most standard definition really doesn't.”
- Accounting for in-kind benefits such as SNAP and housing assistance cuts the measured extreme-poverty count by more than half, and accounting for Medicaid and Medicare would reduce it even further; whichever correction is applied first matters most.
“it goes down by more than half once you account for in-kind benefits. We only account for SNAP and housing benefits; we don't account for health insurance.”
- Because food and housing are basic and constitute a majority of what even middle-class people spend on, poverty measures should account for what government has done to reduce poverty through in-kind benefits (SNAP, housing, Medicaid) and tax credits (EITC, Child Tax Credit), which standard and official measures fail to do.
“I'm a strong believer that we want to account for what the government has done to reduce poverty over time.”
- Employment of single mothers with little education rose from about 45% in the early 1990s to 75% by the end of the 1990s, driven by welfare reform that supported work through expanded Medicaid availability, a greatly expanded EITC (worth roughly $5,000-$7,000 to a low earner depending on children), the Child Tax Credit, and additional childcare funds.
“The employment of those single mothers with little education was about 45% in the early 1990s, and it went up to 75% by the end of the 1990s, with the changes in welfare”
Material living standards of America's poor have improved dramatically over recent decades.
- Other indicators of well-being—quality of housing, appliance ownership, and ability to pay bills—confirm that many people classified below $2 or $4 a day actually have characteristics resembling middle-class families.
“that's confirmed by looking at other indicators of well-being like the quality of housing those individuals are living in, the appliance ownership, the difficulty paying bills.”
- There is a tendency in the economics profession to be uncomfortable reporting good news—those who report good news are dismissed as Pollyannas—even though material well-being of those at the bottom has dramatically improved, as shown by consumption, under-reporting-corrected income, and housing quality.
“there is a tendency in the profession to be uncomfortable reporting good news. That, you seem more serious if you're reporting bad news.”
- Many people classified as poor actually hold substantial assets they could draw upon, and when their housing and appliances are examined they resemble middle-class or slightly-above-average families rather than poor ones, so excluding such asset-holders is justified.
“the individuals that have high assets that we say are not poor, when you look at their housing and the appliances they own, they look like middle-class families.”
- Material conditions of America's poor have improved dramatically: 90% of the bottom 20% have air conditioning—a level the middle 20% reached only about seven years earlier—and you must go back roughly 30 years to find the middle 20% living in apartments as large as those the bottom 20% occupy today.
“if you look at this bottom 20%, 90% have air conditioning, either central or a room unit. And, the middle class, or the middle 20%, had 90% about seven years ago.”
- Because the system already does a better job than the official narrative indicates at meeting the material needs of low-income people, the focus should shift to ensuring people have adequate education and are connected to work or skills-building, since some people are poorly prepared for the modern workforce and lack the intangible skills useful in the workplace.
“we need to do a better job of connecting people to education and training and work than we currently are.”
Correcting for under-reporting and benefits reduces extreme-poverty estimates to a few hundred thousand.
- Setting aside the homeless, people pay rent, and you cannot rent a place for $60 a month for an individual or $200 a month for a family of three, so the claim that people are getting by on less than $2 per person per day does not make sense.
“if you're a single individual, you can't rent a place for $60 a month. You can't find a place for a family of three for $200 a month. So, it just doesn't make any sense.”
- Meyer's best estimate of Americans in extreme poverty is around a couple hundred thousand, and even that is likely an overestimate because his administrative data is incomplete (omitting unemployment insurance, TANF, Workers' Compensation, and capital gains), so the true figure may approach zero—versus the claimed 10 million.
“we come up with a couple hundred thousand individuals; but we're using incomplete administrative data.”
- Claims that millions of Americans live in extreme poverty are deeply misleading because looking at the very tail of the income distribution mostly captures outliers and reporting errors rather than truly poor people.
“The problem is, if you look at the very tail of the distribution, there's a good chance you're looking at outliers, often errors. And that's exactly what's going on here.”
Homeless people form a persistently poor group with substantial program access and partial workforce engagement.
- The best estimates put the US homeless population at a little over half a million at a point in time, based on the HUD Point-In-Time count, with the Decennial Census giving a slightly lower number in the same ballpark.
“the best estimates of the homeless population in the United States at a point in time is a little over half a million. That's from the HUD Point-In-Time count.”
- The roughly half a million homeless at a point in time are not having a bad day, month, or year but a 'bad lifetime'—data on their income and program receipt over time show they have had low income for many years and will continue to, indicating homelessness is generally a persistent rather than transitory condition.
“they aren't having a bad day or a bad month or a bad year. We can see that they're having a bad lifetime in the data.”
- Universal Basic Income would be a mistake because it would discourage work, be inadequate for those who are truly needy, and be too costly for those who don't need support; a better system targets the genuinely deprived and uses humans (social workers, caseworkers) to tailor cash, education, training, childcare, and treatment to individual needs, since people's needs are heterogeneous and cash alone cannot solve problems like addiction, mental illness, or abuse.
“I think Universal Basic Income would be a mistake: that, we want to encourage people to work... a Universal Basic Income would be inadequate for those who are really needy, and would discourage work and be too costly for those who don't need the support”
- Homeless people are more connected to government programs and the labor market than commonly assumed: about 80% of sheltered and 70% of unsheltered homeless receive SNAP (about 75% overall), many receive Medicaid, and about half have some formal work visible on tax records in the year they are homeless—enabled by electronic benefit cards and the ability to receive correspondence at post offices, churches, or friends' homes.
“about 80% of the homeless who are in shelters are receiving food stamps, and about 70% of those who are unsheltered--who are on the street--are receiving SNAP. And... about half of those who are homeless have some formal work that you can see on tax records”