On August 22, the 20th anniversary of welfare reform being signed into law, my new paper, “Poverty After Welfare Reform,” was released. It sprawled into a 45,000-word opus that ate my summer. Due to its length, the report has benefitted from a “Piketty Effect” in that criticism has been slow to come simply because it takes too much time to get through it.
But Slate Moneybox columnist Jordan Weissmann made it through, and he is not buying my claim that child hardship is lower or no higher than in 1996. Unlike many liberal commentators, Weissmann concedes that it is “basically uncontroversial” that child poverty fell after 1996. Even by the official measure—which overstates poverty and understates its decline over time—the likelihood that a child (or a child of a single mother) was poor in 2014 was lower than it was in 1996.
But the official poverty measure obscures how much child poverty has fallen. It does not include the value of noncash benefits in income when determining who falls below the poverty line. Nor does it account for benefits in the form of refundable tax credits for working families. Since taxes have fallen over time, the official measure’s reliance on pre-tax income understates how much disposable income has increased. And because it treats unmarried couples living together as separate families—a living arrangement that has grown more common—the official measure overstates poverty and understates its decline.
Partially correcting these problems (but ignoring employer and government health benefits entirely) puts the 2014 child poverty rate—and the rate among the children of single mothers—at an all-time low. Rather than 45 percent of the children of single mothers being poor in 2014 (the official estimate), 26 percent were. Child poverty is almost surely rarer still today.
Weissmann is typical of reform’s critics, however, in maintaining that the poorest of the poor are worse off today than in 1996. He takes up a number of pieces of evidence that my study also considers. Most prominently, sociologists Kathryn Edin and Luke Shaefer have claimed that “extreme poverty”—living on no more than $2 a day per person—has doubled since welfare reform. Weissmann himself has viewed the Edin-Shaefer claim skeptically in the past, so it was surprising to see him defend their research.
Edin and Shaefer themselves found that 80 percent of the increase in extreme poverty they reported disappeared after including food stamps, housing benefits, and refundable tax credits in income. My analyses using a different data source also initially found a rise in extreme child poverty from 1996 to 2014, but it completely disappeared after taking food stamps into account.
Weissmann displays one of my charts (below) to argue that I still find an increase in extreme poverty among the children of single mothers. But as I state in the paper, the trend for these children is volatile because the data is murky at such deep levels of poverty and there are relatively few children of single mothers in the data at such levels. After taking into account taxes and noncash benefits other than health coverage, the 2014 extreme child poverty rate is higher than in 1996 by one percentage point, but the increase from 1996 to 2013 is only 0.3 percentage points. The 2009 rate is the same as in 1996.
I then show extreme child poverty trends after making two more adjustments that I take great pains to justify. Using a better cost-of-living adjustment to raise the poverty line over time makes essentially no difference for extreme poverty trends. Valuing health benefits as income, however, puts the 2013 extreme poverty rate for the children of single mothers at the 1996 level (even though the 2014 level is higher).
But these results, like those of Edin and Shaefer, should be dismissed out of hand, because it is well-known that both private sources of income and government benefits are grossly underreported in household surveys, the subject of an appendix in my study. As is apparent from the chart, when I partially accounted for such underreporting (of some government benefits), extreme child poverty is practically nonexistent in 2014 (or in 1996).
Weissmann implies that this result is a consequence of a final adjustment I make that shifts away from looking at per-person dollars, but it is evident in the chart that that adjustment makes little difference in comparing 1996 to 2014. He also takes seriously my finding that the increase in extreme cash-income poverty began in the 1970s, while I argue in my paper that one should not believe any of the increase.
Extreme poverty also supposedly rose among married college graduates, which certainly casts into doubt the idea that welfare reform increased extreme poverty among children but also should raise concern that what we are seeing is simply a consequence of bad survey measurement. A new paper by Robert Rector of the Heritage Foundation, for instance, found that several forms of hardship were basically nonexistent among households supposedly living on $2 a day per person.
Weissmann has also apparently forgotten the Brookings Institution paper that he previously plugged that found barely anyone spends less than $2 a day and that a large fraction of families supposedly living on $2 a day said they were living on less than one cent per day.
In addition to citing Edin and Shaefer’s increase in extreme poverty, Weissmann notes estimates from other sources that “deep poverty” (living under half the poverty line) has increased. Like many critics of reform, he provides the 1995-to-2005 estimated increase in deep child poverty estimated by the Center on Budget and Policy Priorities (2.1 percent to 3.0 percent) but does not mention that it fell to 2.6 percent in 2012.
In fact, the trend I show in deep poverty is very similar to the CBPP one. You can’t tell that from Weissmann’s piece because he displays my chart for extreme poverty and says it shows trends in deep poverty. Here’s the right chart:
Furthermore, I go into great detail as to why the small increase I find from 1996 to 2012—0.6 percentage points—is too imprecise to merit confidence. (Weissmann defends the CBPP increase by noting that it is statistically significant, but statistical significance tells us how much an estimate is affected by random errors, while the problems I emphasize in my paper involve nonrandom error such as underreporting, imputation of values, and survey nonresponse.)
The increase is also probably an overestimate (potentially masking a decline in deep poverty) because of underreporting of earnings. While I had access to data to correct for underreporting of some government benefits, I had no such data to correct earnings. As I discuss extensively in my study, we know about the widespread prevalence of earnings underreporting, in part, because of the pre-welfare-reform work of Edin herself. Edin’s research revealed that single mothers receiving welfare benefits needed outside sources of income to make ends meet, but they would not typically report such income for fear of losing their welfare benefits.
Simply put, the available data on deep and extreme poverty is not good enough to provide trend estimates reliable enough to say whether severe hardship is higher in 2014 than in 1996. The change was probably pretty small, in part because few children live in such hardship. Meanwhile the more reliable trend in child poverty (deep, extreme, or neither) is unambiguously down.
Weissmann also cites evidence that the number of “disconnected mothers” (not working, in school, or receiving cash welfare benefits) has risen since 1996. But underreporting of work is a problem for these figures too. One might think that if work underreporting did not worsen after 1996 that it cannot explain an increase in disconnection. However, unreported employment only matters for ascertaining whether a mother is disconnected if she is not receiving cash welfare. As receipt of welfare became rarer and rarer, a steady rate of employment underreporting would have made more and more single mothers appear disconnected in the data even absent a real increase. While they no longer have to worry about losing cash welfare by reporting earnings, they still have to worry about losing SNAP benefits.
Furthermore, single mothers officially poor and actually not working benefited from expansions in the safety net elsewhere, in particular the growth of health benefits. As I write in my paper, citing a U.S. Government Accounting Office study, many families don’t receive cash welfare benefits because they stand to see benefits from other government programs decline if they enroll. Other single mothers reacted to welfare reform by moving in with family members or boyfriends who did have earnings or other sources of income.
In addition to disconnection, Weissmann also notes increases in the number of food stamp recipients without cash income. That estimate is also affected by underreporting of earnings and reliance on other noncash benefits. And Weissmann makes no mention of the fact that receipt of food stamps without income rose for groups unaffected by welfare reform, so it is not at all clear how to interpret these numbers.
Nonetheless, Weissmann insists that “more likely than not, severe poverty has been rising.” There is no basis for that statement. I find very little change in deep child poverty, no $2-a-day poverty, and—in other evidence not discussed by Weissmann—similar 1996 and 2014 levels for a variety of food problems. Food problems—unlike, supposedly, extreme cash-income poverty—became rarer during the second half of the 1990s before turning up in the 2000s. The trend largely follows the business cycle. And at a Heritage Foundation event two weeks ago, economist Bruce Meyer noted that data on housing quality shows improvement for the poorest fifth of single-mother families over the post-reform period.
Weissmann’s piece contains other problems. He elides the detailed arguments I make justifying my choice of a cost-of-living adjustment for the poverty line by calling it “nonstandard.” It is narrowly true that there is not a standard for measuring inflation. But Weissmann seems to be appealing to the idea that I’m doing something a bit kooky without addressing the arguments I make (which take head-on the arguments made in the source to which he links). I’m no kookier than the Congressional Budget Office or the Federal Reserve Board, who also use my preferred adjustment.
Casting doubt on the propriety of including health benefits in some of my estimates, he writes, “The value of health insurance partly just reflects the absurd escalation of medical costs in this country.” But the intention of adjusting for increases in the cost-of-living is to account for the extent to which this escalation represents pure price increases rather than more and better health care. If Weissmann believes that the cost-of-living adjustment is deficient in that regard, he offers no reasons to counteract the evidence I marshal in my paper. I note that a Bureau of Economic Analysis study concludes that our available cost-of-living adjustments currently overestimate health care inflation rather than underestimating it.
These technical controversies—which each get their own appendix in my paper—are often not even relevant. Many of my most important conclusions—such as that child poverty is at an all-time low—are apparent even before I modify the cost-of-living increase in the poverty line or add in health benefits.
Weissmann suggests the consumption poverty research I discuss is tainted by predatory lending, though I show in yet another appendix that credit and debt are not an important part of the consumption trends story. He cites one study to debunk the idea that consumption estimates are better than income estimates—a study that I address in detail in an end note that Weissmann fails to engage.
Finally, there is also a fair amount of trolling in Weissmann’s piece. The title is “The Odd Conservative Argument that Food Stamps and Medicaid Saved the Poor from Welfare Reform.” Similarly, he says this argument is “a bit odd coming from a Republican.” (I am an Independent, not that that matters.) In fact, my paper makes no claims about why child poverty evolved as it did, though Weissmann and others seem to believe it does. Its very title—“Poverty After Welfare Reform”—signals description rather than a causal argument, and that was intentional. My conclusion discusses the complications around making a causal claim about the importance of welfare reform.
However, I have argued elsewhere that welfare reform itself was at least partly responsible for the decline in child poverty after 1996, and I have the first of two pieces for National Review up that will make this case again. Suffice it to say here that my paper does not imply that welfare reform harmed the poor or that the harm it did was simply compensated by food stamp and Medicaid expansions. That argument is in Jordan Weissmann’s head, not mine.
This piece originally appeared on Forbes