House Republicans want to cut spending as a condition for raising the debt ceiling — but they have proven unwilling to make major cuts to the three biggest components of the federal budget: Social Security, Medicare, and the military. And so their just-passed spending plan focuses heavily on what’s left: mostly, programs for the poor.

The Lift, Save, Grow Act, the House GOP’s opening bid in the debt ceiling drama, would add work requirements to Medicaid — the health insurance program for low-income Americans — and expand those in Supplemental Nutrition Assistance Program (SNAP, or “food stamps”) and Temporary Assistance for Needy Families (TANF, often called “cash welfare”).

Requiring people to work to be eligible for social program benefits doesn’t save much money in the scheme of things; the Congressional Budget Office found it would save about $120 billion over 10 years, or 0.2 percent of the 10-year budget. But they could have a major impact on the low-income families that rely on these programs. The White House (hardly a disinterested party, but still) projects the Medicaid provisions could put 21 million people at risk of losing their health insurance. (Roughly 90 million Americans currently receive Medicaid benefits.)

Defenders of work requirements for social programs tend to point to evidence that they, well, “work” — that is, that when work requirements are in place, more benefit recipients wind up employed, which they argue is better than the unemployed passively receiving benefits.

One response from opponents is to dispute the evidence. Arkansas, the first state to introduce Medicaid work requirements in 2018, saw no increases in employment as a result, even as the share of the population without health insurance surged. As my colleagues Alvin Chang and Tara Golshan noted a few years ago, a review of several randomized experiments with work requirements for cash welfare found that while they boost employment noticeably early on, the effects fade quickly with time:

Chart by Alvin Chang and Tara Golshan

Let me make a slightly different argument. I think it’s plausible that work requirements modestly increase work in the short or even long term. But I think they are still a bad idea, because of the effects on the people for whom they do not “work.”

Work requirements inevitably leave in their wake a large group — maybe 20 percent, maybe 30 — who do not or cannot work after their implementation. Those people are then left without either wages or support from the government program that’s now kicked them out. Applied to food stamps and Medicaid, that means creating a group of people who have no cash income, no means of buying food, and no health insurance.

The prospect of abandoning a large group of Americans to that fate should trouble us greatly.

Who is left behind when you add work requirements?

Work requirements are not new. SNAP currently has work requirements for able-bodied adults age 18 to 49 without dependents. The Limit, Save, Grow Act would apply these requirements to people from age 50-55.

Medicaid currently has no work requirements (the Arkansas experiment was blocked by a court). The House bill would change that. Adults aged 19 to 55 without dependents would have to work, do community service, or engage in work training for at least 80 hours a month (about 18 hours a week) to receive Medicaid.

Let’s suppose these requirements, put together, would be startlingly effective at raising 50 to 55-year-olds’ work participation. Let’s say out of the population of adults that age, without dependents, receiving both SNAP and Medicaid, the share holding a job would go from 55 percent (the current level for childless people on Medicaid) to 70 percent. That would be an enormous effect in the context of past work requirements, a huge success.

My question is: What would happen to the 30 percent of people who didn’t find work (or a training program, or a community service opportunity)? If the work requirements were vigorously enforced (as they would have to be to generate the budgetary savings the GOP wants), these people would lose their monthly food benefit. They would no longer have health insurance if they got sick. And they would not have a job. They would have no source of cash income or government support whatsoever.

We don’t really have a recent precedent for so thoroughly abandoning a group of Americans. In the 1990s, Bill Clinton signed a welfare reform bill establishing strict work requirements on cash welfare — but he did so only because no such requirements were imposed on parents’ access to food stamps or Medicaid. He had vetoed two earlier GOP bills that attempted to limit those programs, later telling journalist Jason DeParle, “I thought there ought to be a national guarantee of health care and nutrition.” After welfare reform, the share of people with no income except for food stamps rose sharply.

What the House GOP is now proposing is to pull that safety net out from under the very poorest. If you are someone who can’t meet the new requirements in the world the House GOP is contemplating, you will not have a guarantee of health care and nutrition to fall back on. You will have nothing but private charity and desperate hope.

One sees a similar dynamic at work with disability insurance. The Social Security Disability Insurance program absolutely discourages people from working. There are excellent, rigorous studies proving this, and it annoys me when otherwise like-minded friends try, in defending the program, to pretend that this isn’t true. The reason to keep SSDI, and to not cut it, isn’t that it has zero effect on work. It’s that efforts to reform it and kick participants off will inevitably throw out people who will still be out of work, disabled, and now extremely poor.

A famous study measuring the effect of disability insurance found that in a group of applicants they examined, 52.2 percent of people denied benefits wound up working and earning at least $1,000 after two years; that compares to only 14.8 percent of people granted benefits who ended up working. That’s a big negative effect on employment.

But think about the 47.8 percent of rejected applicants who still were not working. They’re not getting any earnings, and they’re not getting any disability benefits. They’re just very, very poor. Don’t we owe them something? Would toughening up eligibility to force those who could be pressured to work to do so be worth impoverishing this other population, who wouldn’t work either way? Who perhaps couldn’t, physically, work either way?

Work is a good thing. But mercy is a good thing too. There has been a rough consensus, reflected in government policy, in the United States that poor people should not starve, whether or not they work. They should not die for lack of medical care. There should be a (patchwork, imperfect) safety net to prevent the absolute worst possible outcomes.

The moral case against work requirements isn’t that they don’t work, but that they can never work perfectly. There will always be people kicked off benefits who also do not or cannot work — and they will be without any economic resources at all in one of the richest nations the world has ever known.