“If anyone is not willing to work, let him not eat,” reads the New Testament injunction. An overwhelming majority of Americans apparently agree, at least when it comes to work requirements for welfare recipients. And this is most definitely an issue in which public opinion has shaped policy.

A requirement to demonstrate a willingness to work was a key feature of Temporary Assistance for Needy Families, the centerpiece welfare reform of Bill Clinton’s administration. The Supplemental Nutrition Assistance Program (SNAP, or, as everybody still calls it, “food stamps”) allows states to limit benefits given to able-bodied working-age adults, and more have begun doing so. Work requirements for Medicaid may be next.

Some — among them, the “tough love” liberals who supported welfare reform in the 1990s — hope that nudging and cajoling people into jobs will reduce stubbornly high poverty rates. Others hope work requirements will spur economic growth by increasing the effective size of the labor force. Still others see it as a means of curbing the lesser angels of our nature: if safety net benefits are to be had without stigma just for showing up, they fear that people who are perfectly capable of helping themselves will instead flock to welfare as a lifestyle.

These hopes and fears are understandable, but they put the proverbial cart way before the horse. The first task for evidence-based policymakers (if not moralists) is to focus on the actual effects of work requirements. And as the evidence discussed here suggests, they just don’t accomplish much. Thankfully, there are ways to make the social safety net fairer and more efficient than tying it to work.

The Facts, Please

The best evidence about the effects of work requirements probably comes from a set of controlled experiments that were conducted in conjunction with the welfare reforms of the 1990s. From 1991 to 1999, the U.S. Department of Health and Human Services oversaw 11 real-world experiments in seven cities, known as the National Evaluation of Welfare-to-Work Strategies. Each trial compared a group of single-parent welfare recipients who were subject to work requirements (called the “program group”) with a control group that faced no such demands. Several thousand people took part. And to get a sense of what factors mattered most, the work rules imposed on the program groups and the help given in finding and retaining jobs differed somewhat among the 11 experiments.

As expected, the experiments found that households in the must-work program groups were more likely to work during at least a part of the study period than those in the control groups. However, the differences were often small, and the number of welfare recipients who worked even when not required was greater than many people expected.

The highest rate of employment was achieved in an experiment in Portland, Ore., where 85.8 percent of those in the program group found work during at least one of the 20 quarters of the study period. But fully 81.7 percent of those in the control group also found jobs in at least one quarter. In the experiment that generated the lowest employment rate, in Riverside, Calif., 66.9 percent of the program group worked, compared with 61.9 percent of the control group. Moreover, in 5 of the 11 experiments, the positive impact of work requirements on employment was not even statistically significant — that is, the difference recorded had a reasonable chance of being caused by the random division of test subjects between the program groups and the control groups.

The NEWWS experiments did demonstrate that work requirements tend to reduce the number of people on welfare, which was, of course, one of the goals of reform. But here, too, the differences between the program groups and the control groups were disappointingly small. For example, in Portland, the city with the most positive results, participants in the program group still claimed welfare benefits for roughly 20 of the 60 months in the five-year experimental period. In four of the experiments, even the program group continued to receive benefits in more than half of the 60 months.

The findings that most people work whether required to or not, and that many continue to receive benefits even while working, are consistent with more recent observations. In particular, many low-income working families remain eligible for SNAP and Medicaid benefits because their earnings are meager.

A 2013 report from the nonpartisan Center on Budget and Policy Priorities reviewed data for households that received SNAP benefits and included at least one non-disabled working-age adult. At any one time, more than 60 percent of those who were able to work did so. And more than 80 percent worked in at least one quarter in the year before and after they received benefits. That share reaches 90 percent for families with children. Consider, too, that the number of food stamp recipients who worked actually rose during the Great Recession, suggesting that many people who apply for SNAP did so because of intermittent employment or low wages, rather than as a substitute for working.

A similar pattern is evident with Medicaid. A study by the Kaiser Family Foundation analyzed 2015 Census data for Medicaid households that included at least one non-disabled working-age adult. Some 79 percent of these households included at least one person who worked; 64 percent included at least one person who worked full time. All told, 59 percent of the non-disabled working-age adults in those households worked, with a majority working full time for the entire year. Work requirements were not in force for Medicaid in 2015. Full-time workers in such families qualified for Medicaid because their wages were low.

As we have seen, to the extent work requirements have any statistically significant effects, they both increase time worked and decrease public benefits received. As a result, they have an ambiguous effect on total household income, since additional income earned may not offset the loss of public benefits. This may count as a policy success if the goal is reducing public outlays. But it makes work requirements a problematic tool for combating poverty among households with able-bodied but low-skill workers.

In the NEWWS experiments, the reduction in welfare benefits outweighed the increase in earnings, on average, resulting in a reduction of $339 per year, about 1 percent in combined household income. However, as the lower chart on page 43 shows, the results varied greatly from one experiment to another.

Similar results were reported in Kansas and Maine, which recently cut off SNAP benefits to able-bodied adults without dependents. According to an analysis of state data, over the year after the cutoff, earnings of former SNAP recipients increased, but by barely enough (or not quite enough) to offset the value of former benefits.

Why the Weak Effect?

Three factors explain the feeble effect that work requirements for safety net programs have on employment and incomes.

First, there is a failure of incentives. Adding work requirements to conventional safety net programs like TANF, SNAP and Medicaid produces a system in which policy carrots and sticks work at cross purposes. The carrot, in this case, is the added disposable income that participants can obtain by working, and the stick is the loss of benefits imposed on those who do not work.

The incentive failure arises from the fact that in order to save taxpayer money and to limit recipients to the truly needy, most safety net programs aggressively reduce benefits as household income rises. Economists analyze the benefit reduction rate to measure this effect. A household’s overall benefit reduction rate is the sum of the marginal benefit reduction rates on all social programs for which it is eligible.

For example, suppose a household receives SNAP, subject to a benefit reduction rate of 24 percent (that is, 24 cents in food stamps lost for an extra dollar earned), and TANF, subject to a rate of 35 percent. The overall reduction rate for that household would be 59 percent, meaning that its net income would rise by just $41 for each $100 earned. Suppose that earnings come from a minimum wage job that pays $7.25 per hour and is subject to a 7 percent payroll tax. In that case, after-tax earnings — net of the benefits lost — would rise by just $2.47 per hour worked.

The combined effect of benefit reductions and taxes tends to be highest, sometimes reaching 80 percent or more, for households just below and just above the poverty line — the very families with the greatest potential to benefit from welfare-to-work policies. Subjecting them to work requirements puts them in a catch-22 situation. If they don’t work, their benefits are cut off. But if they do work their benefits are nonetheless reduced.

Second, there are no bright lines between “unwilling to work,” “willing and able to work, but not working” and “unable to work.” Many willing welfare recipients face practical problems such as lack of child care or lack of transportation that make it difficult to hold a job. Even if those practical difficulties can be overcome, a high proportion of safety net users experience other employment barriers that fall short of full disability, including physical or mental health problems, criminal records, intellectual impairment, substance abuse and low skills.

People with such problems tend to move in and out of work frequently, whether they are subject to work requirements or not. When they are placed in welfare-to-work programs, they do not move through them in a simple linear fashion from unemployed to trainee to permanent job holder. Work requirements may induce them to make a greater effort to work, but frequent failures and backsliding undermine the statistically measurable effects of well-intentioned policies.

Administrative problems constitute a third factor that can render work requirements ineffective. Running a successful welfare-to-work program requires adequate funding and well-trained staff. Case workers and other administrators must do more than simply monitor eligibility and compliance. They need to serve as life coaches who stress self-sufficiency — a task that may require cultural change on the part of the social workers as well as the beneficiaries.

In the most successful of the NEWWS experiments, the one in Portland, case workers worked intensively with individual participants to cajole them into jobs or training programs, or to coerce them to make greater efforts by threatening withdrawal of benefits. But that modest success suggested that welfare programs are generally ill-designed in this regard. The HHS report directly attributed poor results in Oklahoma City and Detroit to limited staff and program funding.

The need to coordinate government agencies, the private contractors who supply training and education services and potential employers, whether public or private, poses further administrative problems. Even in Portland and Grand Rapids, which reported the best results, delays of one to two weeks in processing information about compliance problems from employers or trainers and contacting participants were considered normal. In Detroit and Oklahoma City, which invested less in program administration, delays of two to three weeks, or even more, were common. The NEWWS report notes that, at any given time, some participants were “simply ‘lost track of’ by program staff.”

Congress Take Note

Taken as a whole, the evidence on work requirements discussed here supports three sobering takeaways.

• A majority of non-disabled working-age adults who receive public assistance work at least some of the time whether they are required to or not. Their incomes nevertheless remain low because their wage rates are low and their work tends to be part time or intermittent. Thus, adding work requirements to existing safety net programs should be expected to have modest effects at best on labor force participation.

• A large portion of the non-disabled adult population needing safety net assistance who do not work still face barriers to finding and holding jobs other than their willingness to work. One’s heart need not be prone to bleeding to recognize that poverty is rarely solely caused by lack of motivation to achieve. To take the most concrete example: the millions of people who have been convicted of non-violent crimes face daunting barriers to stable employment for the rest of their adult lives. Work requirements that target the marginally employable thus cannot be expected to have much effect on actual employment unless they include adequate funding and staffing to overcome myriad social and health problems.

• Work requirements layered on top of the existing safety net system are never going to be an effective tool for combating poverty. Although well-administered work requirements can produce moderate increases in earned income, the effects on total incomes are eroded by high effective benefit reduction rates for the working poor — and, of course, by termination of benefits for those who are unwilling or unable to comply with the requirements.

A focus on work requirements does, however, suggest some ways in which policy could be tilted in constructive directions. For a start, efforts to increase labor force participation among public assistance recipients should focus on increasing incentives to work by lowering benefit reduction rates and payroll tax rates. Consolidating separate safety net programs into a single negative income tax is one of the oldest ideas. The Earned Income Tax Credit, which provides greater work incentives than a negative income tax for workers in the lowest income brackets, is really a modern version of the NIT.

There are several ways to build on this approach. A universal basic income would effectively cut benefit reduction rates to zero, since the benefit would not be linked to earnings. Note that a UBI would provide an especially big boost in incentives for those just below and just above the poverty threshold, who are helped less by the current formulation of the EITC. No matter how you manage it, though, the key is the same: lower benefit reduction rates would encourage greater labor force participation.

Even if work incentives were improved, other barriers to finding and holding jobs — I won’t repeat the litany — would remain daunting for many. This implies that beefed up support from public or private agencies would still be needed to maximize the effectiveness of any sort of work incentives. And while the prospect goes against the conservative grain in American politics, it’s hard to dispute that welfare reforms would have a much greater prospect of success within a supporting context of reforms to health care, education, drug policy, criminal justice and the tax system.

* * *

“If anyone is not willing to work, let him not eat” might accurately reflect the ethos of a hardscrabble Mediterranean culture two millennia ago and may ring a moral bell for some today. But it’s not a useful guide to policy. If unwillingness to work were the main reason for poverty, work requirements might be an effective remedy. But that is just not the case. Moving people from welfare to work is a worthy goal. However, getting there requires much more than issuing a command and waiting for the salutary results.