Welfare reform is pretty popular. A Rasmussen poll last July found that 83 percent of American adults favor the work requirement placed on welfare by the 1996 Personal Responsibility and Work Opportunity Act, commonly known as the welfare reform law. In 2002, closer to the bill's enactment, a Pew poll found 46 percent of Americans thought the bill changed the welfare system for the better, and only 17 percent thought it changed things for the worse. Mitt Romney certainly thought it was popular, as he purchased airtime for an ad that (falsely) alleged that President Obama has gutted the law:

And some of the numbers seem to suggest this support is justified. The average number of people receiving cash benefits from Temporary Assistance for Needy Families (TANF), the name welfare has gone by since 1996, has fallen from 12.6 million that year to 4.6 million in 2011. "Caseloads declined by 54 percent. Sixty percent of mothers who left welfare found work, far surpassing predictions of experts," President Clinton wrote in a 2006 op-ed in the New York Times. "Child poverty dropped to 16.2 percent in 2000, the lowest rate since 1979, and in 2000, the percentage of Americans on welfare reached its lowest level in four decades."

But it's not that simple. Indeed, the health consequences of the change, a new study suggests, are potentially quite large, and quite negative. The Health Affairs study, written by Columbia's Peter Muennig and Zohn Rosen, along with the Wallace Foundation's Elizabeth Ty Wilde, finds that welfare reform increases mortality among recipients, reducing life expectancy by about nine months.

They don't look at the 1996 act but at the Florida Family Transition Program, a precursor welfare-to-work program that operated from 1994 to 1999. Helpfully, the program involved a randomized trial. Participants were randomly assigned either to Aid to Families with Dependent Children (AFDC), the program TANF replaced, or to the Family Transition Program, in which they received training, education and job placement services but were limited to either 24 months of cash aid per 60 months, or 36 months cash aid per 72 months; people in the latter group were judged to lack work experience and need more time to prepare to reenter the workforce.

The people in the experimental welfare-to-work group were much likelier to obtain employment, but they weren't likely to make more total income than those in AFDC, once you take cash assistance and other forms of income AFDC recipients received into account. Of the 1,611 experiment participants in the county the study focused on, 75 died by November 2011. Of the 1,613 members of the control group, by contrast, 67 died by November 2011. That means the welfare-to-work had 16 percent higher mortality than those receiving normal cash assistance, a result that was highly statistically significant and, because of the study's random design, can be attributed to the different welfare program. That amounts to a nine-month reduction in life expectancy between the ages of 30 and 70.

One can't generalize from that to say that national welfare reform cost lives, since the programs are sufficiently different, though it's worth noting that the Florida program was, if anything, less stringent than the federal law. But it would hardly be surprising if the same were true of the 1996 law. As the Center for Budget and Policy Priorities has noted, welfare reform made TANF a much less comprehensive anti-poverty program than AFDC was. In 1979, under AFDC, 82 families were on welfare for every 100 families with children in poverty. When welfare reform was passed in 1996, that had fallen to 68 families per 100. By 2010, it had plummeted to 27 per 100:

Welfare, for better or worse, simply wasn't supporting poor families as much as it had. That showed up in the poverty numbers. Research from Kathryn Edin at Harvard and Luke Schaefer suggests that extreme poverty — defined as households living on less than $2 per day per person — grew greatly after welfare reform's passage.

That's not a coincidence. The yellow line shows what extreme poverty would be each year but for AFDC/TANF, and the dark blue line shows the extreme poverty rate taking that and other cash assistance into account. In the 1990s, the distance between the two is great, suggesting that AFDC/TANF kept a lot of folks out of extreme poverty. But by the late 2000s, it had narrowed greatly:

It's hardly a big leap to think that increases in extreme poverty could lead to increases in mortality, but other studies on the health effect of welfare reform are mixed. In 2008, UC  Davis's Hilary Hoynes and UC  Irvine's Marianne Bitler summarized the literature as concluding, "welfare reform led to reduction in health insurance coverage, with small and often insignificant impacts on health care utilization and health status."

But there's also literature suggesting that anti-poverty programs improve health outcomes,   literature for which Hoynes is largely responsible. Along with Davis's Douglas Miller and David Simon, she found that a $1,000 increase in the earned income tax credit (EITC) is associated with a 6.7 to 10.8 percent reduction in the rate of low birth weight for infants. That's backed up by a study by Notre Dame's Williams Evans and Northwestern's Craig Garthwaite, who found that the expansion of the EITC included in Clinton's 1993 budget reduced the probability of recipient reports of high blood pressure by 3.2 percentage points and the probability of reporting biomarkers that predict stroke, heart attacks and mortality by 9.6 percentage points. Reagan Baughman at the University of New Hampshire found similar results for state EITCs.

Same goes for non-EITC programs. Along with Davis's Marianne Page and Ann Huff Stevens, Hoynes found that the aid for Women, Infants and Children (WIC) program increases birth weight by between 18 to 29 grams. That's actually lower than other researchers; Bitler and Princeton's Janet Currie found that it increases birth weight by 62 grams. And along with Columbia's Douglas Almond and Northwestern's Diane Schanzenbach, Hoynes found that food stamps increase birth weight by between 15 and 20 grams for whites and between 13 and 42 grams for blacks.

None of those are directly relevant to welfare reform per se, but they do suggest that benefit cuts implemented through welfare reform could hurt health outcomes, potentially swamping any gains from increased employment. For her part, Hoynes calls the new mortality study "potentially very important," but emphasizes that the effect may not hold in all states, and that we know very little about why mortality increased. "What were the causes of death?" she asks. "Internal (e.g. heart disease) or external (accident, homicide)? Who was dying? Which age ranges?" All of that's important for interpreting the results.

So it's too soon to conclude that the national law actually cost lives. But if the Florida results hold, it's probably worth taking another look.