A new study focuses on how cash payments to families in poverty affects recipients in the long term. (Mark Makela/Getty Images)

It has been more than four decades since the federal government dedicated hundreds of millions of dollars to a massive experiment on poverty. Thousands of families in several cities participated, entering a program that guaranteed as much as $25,900 a year for a family of four.

Economists are still debating the results of the federal Income Maintenance Experiments. This week, two researchers put out some surprising new findings from the Seattle and Denver programs that challenge existing theories about welfare and the poor.

On one hand, the results contradict two common conservative critiques of welfare: The researchers did not find that the payments broke up marriages or made the children more dependent on government help as adults. On the other, there is little evidence to support the hopes of some on the left that the payments would improve children's future prospects in life.

The researchers' draft paper also contains an unexpected and troubling puzzle. Participants in the experiment retired earlier than those who received less generous payments from regular federal welfare programs. Initial studies had predicted the experiment would have only a minimal effect on labor in the long term.

Largely as a result of these early retirements, participants made less money during their working years: For each dollar the government spent on the Income Maintenance Experiments in Seattle and Denver, it reduced participants' incomes over their careers by $4.50. Hence, if a person received $20,000 from the program, her lifetime earnings would be reduced by $90,000. In short, an experiment designed to help lower-income Americans might have instead cost each family that participated tens or even hundreds of thousands of dollars.

The draft — produced by David Price, a graduate student at Stanford University, in collaboration with Jae Song, a federal economist at the Social Security Administration — addresses a question that almost no one else has studied: the effects of welfare programs on recipients decades later.

That means there is little other data available to validate the results.

“It's important to be a little bit cautious in interpreting the results,” Price said.

Because the payments available through the Income Maintenance Experiments were more generous than those available through traditional welfare, it is no surprise that participants' earnings were lower than those of households who were not assigned to the program. Participants could cover their expenses without working as much, and there was no requirement that they work to receive benefits.

Price and Song found that husbands in the experiment worked 9 percent fewer hours, while wives worked 20 percent fewer hours. For single mothers, the figure was 14 percent.

However, it is unclear why participants in the experiment would retire earlier, long after the program ended. For some participants, the experiment lasted three years; for others, five.

Participants by and large returned to work immediately after the program ended, which gave researchers confidence that the payments would not have persistent consequences.

Price and Song offer a few possible explanations for why the program's effect on labor was delayed so many years.

It could be that, because participants spent less time on the job during the experiment than those not assigned to the program, they developed fewer skills, which reduced their potential to earn and made retirement more attractive many years later.

However, Charles Michalopoulos, the chief economist at the social-policy research group MDRC, argues that this interpretation is at odds with other research. Workers in occupations that pay poorly do not always develop their skills on the job, he said, so adults who were not in the experiment would not become significantly more skilled simply because they worked more hours.

Another possibility is that participants saved some of the money they received during the program, but it is hard to see how they could have saved enough to reduce their earnings so substantially later on.

Finally, Price and Song suggest that, during the program, participants might have acquired a taste for leisure that persisted into the final years of their working lives. This theory implies that participants planned an early retirement many years in advance rather than immediately working less.

There are major differences between the experiments the government conducted decades ago and the programs that exist for the poor today. Research on existing federal programs generally finds they have a limited effect on how much participants work. Since current programs barely affect participants' earnings in the short term, it seems unlikely that there would be an effect in the longer term.

To receive major forms of federal assistance — such as cash and food stamps — Americans must satisfy requirements that they work or look for work. The Earned Income Tax Credit, another important program for the poor, only rewards those who claim earnings in the formal labor market. For other programs, such as housing vouchers and Medicaid, assistance is provided in kind, so recipients often work to earn cash to spend on other necessities.

The differences between the experiments conducted several decades ago and those in existence today could also explain why Price and Song's results do not support conservative or liberal arguments about social policy.

For instance, they write, it could be that children of parents who participated in the program were not more likely to seek help from the government because the program they knew growing up was no longer in existence after the experiment concluded.

Likewise, it could be that the experiment did not generate benefits for children because not all of the parents in the experiment were desperately in need. While past research suggests that children do get important benefits from a variety of public programs, it is possible that public assistance does the most for children in the most indigent families.

At the very least, Price and Song's paper demonstrates the value of experimentation. Massive social experiments like the ones conducted in Denver, Seattle and other cities are infrequent, but they can produce results that confound experts' expectations and open up new possibilities for further research.

No steak, no seafood, no strip clubs: There's a logical gap in the recent laws that bash the poor who receive government welfare and food stamps. (Tom LeGro/The Washington Post)

More from Wonkblog:

How welfare reform changed American poverty, in 9 charts

What rich countries get wrong about poverty

One of America's most popular children's books has a secret political message