What a New Study Gets Wrong About Child Poverty and the Social Safety Net

The New York Times this week reported on a new child poverty study with two big findings. First, child poverty has fallen by 59 percent between 1993 and 2019. Second, the most important reason for the decline was the expanding social safety net. The first finding is true but not exactly breaking news. The second finding is simply false—and bolsters a counterproductive progressive framing around anti-poverty policy.

The new report was written by a seven-person team at Child Trends that used data largely assembled by Columbia University’s Center on Poverty and Social Policy. The researchers found that child poverty fell from 28 percent in 1993 to 11 percent in 2019. This is a huge accomplishment but a trend that has been observed for a decade now. 

My American Enterprise Institute colleague, Richard Burkhauser, and other researchers published a paper in 2019 reporting that overall poverty fell dramatically over 60 years. Three years earlier, I wrote a report that found child poverty fell from 18 percent to 8 percent between 1993 and 2014. And in a 2014 report, the Congressional Research Service found that poverty among single mothers dropped from 38 percent in 1993 to 24 percent in 2013. Another colleague, Bruce Meyer, wrote a paper with James Sullivan in 2012 that found poverty among single-parent families fell from 27 percent in 1993 to 9 percent in 2010. 

This significant accomplishment challenged conservative and liberal assumptions alike. The U.S. had not “lost” the war on poverty after all, and perhaps the safety net had done some good. But it also threw into doubt the notion that the welfare reforms of the 1990s had failed abjectly. In recent years, conservative scholars have seemed more comfortable accepting the reality of a decline in child poverty, coming as it did after welfare reform. 

Liberals may now feel safe embracing the decline in poverty since this latest report claims that it was primarily due to safety net spending, but it’s not that simple. 

The study first purports to show that economic factors can explain only a third of the drop in child poverty between 1993 and 2019. Using statistical modeling, they estimate the change in child poverty from a change of a given magnitude in unemployment (among the general population), labor force participation (among single mothers), the minimum wage, median hourly wages (among all workers), and overall economic growth per capita. Using those estimates and the actual change in each of these factors from 1993 to 2019, they estimate how much poverty reduction was due to each.

Yet the modeling they conduct relies on a slew of hidden assumptions. One methodological choice in particular is deeply problematic—their use of “year fixed effects.” As the authors note, this modeling decision means that economic changes that occur uniformly for all states (say, broad-based growth) are neglected for purposes of estimating the relationship between those changes and child poverty. It means, in their words, “Our models may thus underestimate the role of these variables in the decline in child poverty.”

Because of this problem, they find that a rising tide—increasing gross domestic product—lifts no boats—does not affect child poverty. Nor do rising median wages. In contrast, increases in the minimum wage, which do vary primarily by state, do affect child poverty. Remember, a change that occurs uniformly across states, such as an increase in the federal minimum wage, gets statistically controlled away.

Notably, when the Child Trend researchers shift to looking at the influence of the safety net, they reject this sort of modeling: “Because changes in federal policies often affect all states at the same time, this method does not allow us to evaluate associations between federal policies and child poverty rates.” The same logic would argue against “year fixed effects” in the economic analyses; increased economic growth that provides a boost to families in every state gets whittled down to just the boost that varies across states in their modeling. Economic growth that reduces child poverty equally across states does not register as reducing poverty at all.

Instead, the Child Trend team shifts to a different kind of modeling. They start with child poverty rates based on family income before any government benefits are taken into account. Then they add in, successively, different benefits and look at what happens to poverty rates. Adding in the earned income tax credit, for instance, lowers the pre-tax and -transfer child poverty rate by more in 2019 than in 1993 and thereby accounts for some of the decline in child poverty. Taken together, safety net programs reduced the child poverty rate by 2.8 points in 1993 and by 8.8 points in 2019, a difference of 6 percentage points.

Though I don’t think the Child Trends folks are aware of it, their results actually undermine the view that this impact makes safety net expansions the most important part of their story. Consider that the study finds that child poverty fell by 16.5 percentage points between these years. If economic factors explain a third of that drop, then that credits 5.5 points of the decline to those factors. Even with the study’s modeling bias against economic factors, that is nearly as large as the effect of the safety net (6 percentage points). 

The problems go deeper. If Child Trends wants to say that “healthy economic conditions alone were not sufficient to protect children from poverty,” then they should apply the same language to the safety net. What their analyses show is that after adding safety net benefits to pre-tax and -transfer income, the child poverty rate goes down. That sidesteps the fact that without earnings and other non-transfer income, safety net benefits alone are usually not enough to raise families with children above the poverty line. And for every family that is lifted above the poverty line only once you count their (now greater) safety net benefits, other families may be lifted out of poverty once you count their (now greater) pre-tax and -transfer income.

Child Trends’ own analyses show the greater importance of rising pre-tax and -transfer income (earnings, business income, asset income, private transfers such as child support, and retirement income). By their lights, the safety net reduced child poverty by 6 percentage points versus what the decline would have been without the expansion of the safety net. However, the actual child poverty rate fell 16.5 percentage points. The difference—10.5 points—is the effect of additional private income (net of taxes paid), and it is 75 percent larger than the effect of the safety net. (Click here for more on the importance of private income.)

And there’s another problem: Moving from $1 below the poverty line to $1 above the poverty line is an arbitrary indicator of success. Greater safety net benefits and greater earnings might improve a family’s well-being even if they remain below the poverty line or even if they start out just above it.

I conducted my own analyses leveraging the Current Population Survey (CPS). Because Child Trends and the Times are so focused on the role of safety net benefits, I looked at the family incomes of single mothers not in school and in the bottom third of the education distribution. (To identify the bottom third of the education distribution, I looked at all women ages 18 to 54 and not in school.) 

The most comprehensive measure of family income I have readily available through 2019 includes all pre-tax and -transfer income, public cash transfers (welfare, Supplemental Security Income benefits, unemployment and worker’s compensation benefits, public retirement and disability benefits, and energy subsidies), refundable tax credits (the earned income tax credit and additional child tax credit), and several noncash transfers (food stamps and subsidized school lunch benefits). It does not include housing subsidies or Medicaid or Medicare benefits, which, as of 2019, were no longer in the CPS data that I use. (Child Trends uses Supplemental Poverty Measure data that includes housing subsidies and benefits from the Special Supplemental Nutrition Program for Women, Infants, and Children (WIC); otherwise, our income types are very similar.) For simplicity, I ignore taxes, except for the refundable credits that are, effectively, cash transfers. 

Unlike Child Trends, I compare business cycle peaks, which ensures that any increase in pre-tax and -transfer income or decline in transfer income is not due to comparing the recession year of 1993 to the peak year of 2019. I also use a better inflation measure than they do (the “implicit price deflator for personal consumption expenditures,” for the measurement nerds out there.) Between 1989 and 2019, the mean of my measure of comprehensive (pre-tax, post-transfer) family income rose by 49 percent among less educated single mothers. 

To assess how important different types of income were to this increase, I looked at the rise in aggregate family income within this group, beginning with pre-tax and -transfer income, then adding refundable tax credits to workers, then cash transfers, then SNAP and school lunches. At each step, I looked at how the increase in income compared with the increase in comprehensive income, assessing the incremental amount of the increase in comprehensive income “explained” by adding in income types.

The mean pre-tax and -transfer income of less-educated single mothers rose 71 percent between 1989 and 2019. In the aggregate, the increase amounted to 101 percent of the rise in the group’s post-tax and -transfer income.

The earned income tax credit and additional child tax credit, which only go to families with pre-tax and -transfer income, accounted for another 3 percent of the increase in family income among less-educated single mothers. Non-cash transfers accounted for only 1 percent of the increase. 

Cash transfers amounted to a 4 percent decline in family income over the period. (When added together, these percentages sum to 100 percent.) Welfare reform is responsible for the decline in the contribution of cash transfers to income. Yet by incentivizing work, it is also responsible for some of the increase in private income, along with some of the increase in income from the EITC and ACTC, which go only to workers.

Between the peak years of 1989 and 2007, I can look at a family income measure that includes housing subsidies, Medicaid, and Medicare. (See my 2016 report for details of how I value these benefits.) Between those 18 years, the mean of this comprehensive measure of family income rose 18 percent among single mothers not in school and in the bottom third of education. 

Fully 94 percent of this increase can be credited to a rise in pre-tax and -transfer income. Another 7 percent was due to Medicaid and Medicare. The earned income tax credit and additional child tax credit were responsible for 5 percent of the increase, SNAP and school lunches accounted for another 2 percent of the rise. Meanwhile, housing subsidies accounted for none of the increase, and cash transfers amounted to a decline in income of 9 percent. 

Now, the CPS misses a lot of transfer income, so it is likely that the analyses above understate the increase in income among single mothers over time and the share of the increase that is attributable to various safety net benefits. But the increase in private income is, by far, the most important driver of increases in the well-being of single parents. To neglect this is to downplay the importance of economic growth and human capital investment in poverty reduction, and the crucial role of work incentives within anti-poverty programs. 

The importance of single parenthood is also given short shrift by Child Trends. Sharply declining teen pregnancy rates explained 52 percent of the decline in deep child poverty. Living with two parents was also strongly associated with avoiding child poverty, but because the change in living arrangements was small, it did not explain much of the decline in poverty.

The study’s incomplete analyses focused on income from safety net programs reinforce progressive policies that come down to “just giving people money.” There is no consideration of the possibility that welfare reform helped poor children by encouraging parental work and marriage and could serve as a model for other programs. At one point in the report it looks like the authors might acknowledge the importance of policies to revive marriage when they begin with, “Drawing on our finding that the proportion of children in two-parent families is strongly associated with child poverty …” But they then recommend an assortment of liberal priorities having little to do with family stability: paid leave, flexible work scheduling, and reforming incarceration and child welfare to be less punitive. They also manage to include greater access to abortion as a recommendation, pegged to the teen pregnancy results.

“While increases in the share of children living in two-parent families were strongly associated with [declines in] child poverty, we recommend caution to readers in interpreting this finding. We specifically recommend caution in developing policy interventions that directly encourage parents to maintain or create two-parent households. Incentives to marry or otherwise maintain two-parent households could have the effect of directing resources away from children in single- or no-parent households,” they continue. “Such incentives could also trap families who are experiencing domestic violence. A narrow focus on two-parent households may also miss opportunities. Research illustrates the role of extended family—for example, Black grandparents living with their grandchildren—in supporting children and parents.”

The Child Trend researchers write that “the United States must continue its collective efforts to further extend the past quarter century’s decline in child poverty rates. The lessons of this decline provide powerful insights into how we can continue to reduce child poverty. The past has much to say, and we should listen.” 

The past does have much to say about poverty reduction, but it seems like a lot of people do not want to listen. And then a lot of consumers of research and the news—a lot of policymakers—go misinformed. Poor children deserve better.

Scott Winship is a resident scholar and the director of poverty studies at the American Enterprise Institute.