President Bill Clinton promotes welfare reform / The Washington Post

Ever since the welfare reform measure passed in 1996, its central tenet — that cash assistance should be dispersed only on the condition that recipients move toward employment — has become a near-consensus in Washington.

But the rule has its critics. The University of Michigan's Luke Schaefer and Harvard's Kathryn Edin have found that extreme poverty — families living under $2 a day — shot up after the bill passed, indicating that it pulled the rug out from under a number of very vulnerable people. But there's also a more general argument to be made that assistance to the poor shouldn't come with those strings.

A new working paper from economists Nicola Pavoni, Ofer Setty and Giovanni Violante adds some quantitative weight to that critique. They built a model designed to determine the "optimal" welfare system, the one that maximized the utility — that is, happiness, or degree of getting what you want — across the economy.

They compare a number of possible solutions. On one extreme, there's pure, unconditional cash transfers, in which the government simply distributes money without any requirements to needy persons. On the other, there's what the authors call "mandatory work", in which needy people receive public works jobs that they must keep to receive assistance.

But there are a lot of options in between, including unemployment insurance, part-time public works, full-time but temporary public works, and extensive government job search assistance. Which one is best for a given country depends on a number of factors, but the authors focus on two: the skill level of workers, and the amount of money the government has to spend. Here's what they find:

Data: Pavonni, Setty, and Violante. Graphic: Dylan Matthews

The numbers in the authors' data are particular to their model and don't tell us a whole lot about real-world policies, but there are some very interesting implications here. One is that when the government can't afford generous benefits, it's cheaper for it to provide assistance in the form of public works jobs rather than through cash assistance. This is something public works advocates have been saying forever. It was born out to some extent in the subsidized work program that the nation's stimulus funding provided to welfare recipients, which created jobs more efficiently than any other stimulus component. That's not a true public works program, but it's closer to it than any other recent policy, and it showed in the results.

The second takeaway is that the more skilled the population is, the more important it is to get them off government assistance and into private work. That's because high-skilled workers are more productive and thus can produce more output both for themselves and the economy at large when in private jobs, rather than in public works or under government assistance.

Designing effective welfare programs that help people out of poverty without creating perverse incentives is quite difficult, and Pavoni, Setty, and Violante do a strong job of illustrating that complexity.