A few weeks ago, Matthew Desmond, a Pulitzer-prize winning sociologist, released an adaptation of his upcoming book, “Poverty, by America,” in the pages of the New York Times Magazine. He claims there has been “no real improvement” in poverty reduction over the last several decades despite the expansion of anti-poverty programs like the EITC, CTC, and SNAP. He believes this is because the root cause of poverty—what he calls exploitation—has gone ignored. He goes on to list problems in the labor, housing, and financial markets as well as solutions to them. While his descriptions of these and his prescribed solutions to them are well and good for the most part, he, for whatever reason, fundamentally misrepresents the progress on poverty over the last thirty years, and misunderstands why people fall in poverty in the first place.
Lets start with his characterization of poverty trends:
On the problem of poverty, though, there has been no real improvement — just a long stasis. As estimated by the federal government’s poverty line, 12.6 percent of the U.S. population was poor in 1970; two decades later, it was 13.5 percent; in 2010, it was 15.1 percent; and in 2019, it was 10.5 percent. To graph the share of Americans living in poverty over the past half-century amounts to drawing a line that resembles gently rolling hills. The line curves slightly up, then slightly down, then back up again over the years, staying steady through Democratic and Republican administrations, rising in recessions and falling in boom years.
What accounts for this lack of progress? It cannot be chalked up to how the poor are counted: Different measures spit out the same embarrassing result. When the government began reporting the Supplemental Poverty Measure in 2011, designed to overcome many of the flaws of the Official Poverty Measure, including not accounting for regional differences in costs of living and government benefits, the United States officially gained three million more poor people. Possible reductions in poverty from counting aid like food stamps and tax benefits were more than offset by recognizing how low-income people were burdened by rising housing and health care costs.
In the first paragraph, he points to trends in the government’s Official Poverty Measure (OPM). Here is what that measure looks like over time:
From looking at this time series, his claim that the poverty rate has barely improved since 1970 would seem correct, right? Wrong. To derive the official poverty rate each year, the U.S. Census Bureau adds up all of a family’s gross before-tax income—including earnings and wages, capital income, and Social Security income—and compares it to a threshold based on that family’s size and composition, adjusted for inflation each year. If total family income falls below that threshold, then all members of that family are said to be in poverty. One problem with this measure is that it excludes in-kind transfers and is calculated before taxes. That means that some of the nation’s biggest anti-poverty programs, like the Earned Income and Child Tax Credits, SNAP, and Section 8 housing vouchers, are completely disregarded from the official poverty rate’s definition of income.
Desmond tries to get around this by pointing to the Supplemental Poverty Measure (SPM), a new measure of poverty the Census Bureau began publishing in 2011, which accounts for taxes and in-kind transfers like those mentioned above. He notes that even if you look at the SPM, it shows there are more poor people than the OPM. But if you pay close attention, you’ll notice Desmond points to the SPM only at a particular point in time, not its change over a broader period, which is central to his claim that poverty has stagnated.
As you can see, the trend in the SPM is very different than that of the OPM. According to this measure, the poverty rate fell from 17.2% in 1970 to 12% in 2019, a 30% decrease.1 You might still be thinking to yourself, “That’s true, but the poverty rate stagnated for most of the time in between, with almost all of the improvement occurring at the tail ends of the graph.” That’s correct, but the reason for this has to do with the SPM’s own methodological quirks. While the SPM’s definition of income is better than the OPM’s, its poverty thresholds are more problematic for tracking how many people can afford a particular set of goods. Unlike the OPM, which compares family income against a constant basket of goods, the SPM compares disposable cash resources against thresholds that change in real terms. These thresholds are equal to a 5-year moving average of the 33rd percentile of expenditures on food, clothing, shelter, and utilities, with some other adjustments to account for other basic needs and family size. If society as a whole gets richer over time, leading to greater consumption of those goods across the entire population, the SPM thresholds will move upwards as well. This, in my view, makes the SPM more a measure of inequality in the consumption of particular goods than one of poverty, since it better captures how many people fall within a particular distribution of consumption, not how many people can afford a set level of consumption.
So how can we get around this? Fortunately, researchers at Columbia University’s Center on Poverty and Social Policy built an anchored version of the SPM. That is, they picked poverty thresholds for a single year—2012—and then adjusted it only for inflation for all years before and after. This ensures the thresholds reflect a constant, real bundle of goods and services to compare family resources against. With this, we can see just how many people were able to afford a similar basket of goods and services across multiple years
And here we get a much clearer view of poverty over time. According to this measure, the poverty rate fell from 21.6% in 1970 to 11.2% in 2019, a nearly 50% decrease. When looking at just the poverty rate among children, the decline is even greater. In 1970, 25%, or one in four, children lived in poverty. By 2019 just 11.8% did. This is consistent with alternative consumption measures, which estimates poverty based on people’s actual consumption, rather than their income or resources.
With the exception of the OPM, none of the poverty measures we looked at paint a picture of stagnation. Across almost all definitions of poverty, it has fallen. A reason for this that we should celebrate is the dramatic rise in spending on programs like Medicaid and CHIP, and tax credits like the EITC and the CTC.
Between 1979 and 2019, means-tested transfers to the lowest income quintile grew from 32% of average income to 64%, and refundable tax credits like the EITC and CTC from 1.2% to 12.4%. This led led to a pretty dramatic rise in incomes after taxes and transfers for the bottom fifth of households. Over the entire period, average post-tax-and-transfer income of the bottom fifth of households grew by 94%.
Desmond notes the growth of these programs, but fails to appreciate them and their effects on poverty. Why? Well, I think he is confused as to where to poverty actually comes from. As much as the flaws that he points to in labor, housing, and financial markets are real, they are not why people are resource poor in the first place. In a market economy like ours, people can only receive income from two sources: labor and capital. If a person neither can work nor has any claim on capital, such as a 401(k) retirement account, then they receive no income at all. Unbeknownst to most people, a majority of people in the United States neither have paid employment nor significant financial assets. 52% of all people in the United Sates do not work at all.2 This is because they are either kids, students, disabled, elderly or retired, or caring for their home or a family member.
In a previous post, I pointed to Matt Bruenig’s great piece on redistribution to demonstrate how certain conditions drag people down the income ladder.
The bottom third of people by market income are almost all children, carers, disabled, unemployed, or retired or elderly. These are exactly the people who cannot and should not work and earn an income. Even if we strengthened collective bargaining rights as Desmond suggests, it would do nothing to raise the incomes of people in the bottom third of the personal market income distribution. Some might argue this inaccurate since people within families share resources, so even if a child does not have an income of their own, they are still able to benefit from an increase their parent’s earnings. But even when we look at the composition of the population by family market income, it is still non workers who make up the bulk of the people, especially among those in the bottom 5%.
What brings up those nonworkers upwards in the income distribution is not higher wages, but the addition of social insurance benefits like Social Security and Unemployment Insurance income, which specifically go to those without a job. And this really gets to the center of the problem. To tackle poverty, governments do not even need to focus on it. All that is necessary is the transfer of income from those who receive market income—workers and owners of capital—to those who don’t—children, the disabled, the elderly, and the like. While Desmond gets this wrong, he is right that the growth in our social programs discussed earlier is not enough. Our welfare state is still embarrassingly stingy, means-tested, and reliant on work and earnings requirements. That needs to change if we are going to end poverty. But if we are going to succeed in that, we need recognize it and be honest about the facts.
While the Census Bureau only began publishing the SPM in 2011, researchers at Columbia University’s Center on Poverty and Social Policy were able to produce the measure for previous years using the Current Population Survey Annual Social and Economic Supplement public use microdata that the SPM and OPM are based on.
Authors’ calculations based on Flood, Sarah, et al. “IPUMS CPS: Version 10.0” 2022 Current Population Survey Annual Social and Economic Supplement, Minnesota Population Center, University of Minnesota, 2022. https://doi.org/10.18128/D030.V10.0.
Wrong