The Trump administration has announced that it will allow states to impose work requirements on Medicaid recipients. Administration officials portrayed the policy change as an attempt to encourage workforce participation among the poor, characterizing it as a move away from “the soft bigotry of low expectations consistently espoused by the prior administration.”

Nevermind that most Medicaid recipients work already, many holding down two or three jobs in precarious low-wage sectors. Meanwhile, those who don’t work usually can’t, either because there are no jobs available, or because of circumstances that make employment infeasible — like untreated chronic pain, inadequate transportation, or lack of child care. By characterizing Medicaid work requirements as much-needed kick in the butt to unemployed poor people, conservative politicians are recapitulating mythical ideas about welfare reliance — conjuring what Ronald Reagan famously called the “spider’s web of dependency,” which he used to justify his administration’s relentless attacks on social programs.

But conservatives aren’t the only ones to blame for the consistent incursions on Medicaid. Work requirements are only possible, structurally speaking, because Medicaid is a means-tested program to begin with — that is, it targets a specific population with strict criteria and eligibility tests, unlike Social Security (or, theoretically, single-payer health care) which applies automatically to everyone and is therefore universal. Over the decades, many of the architects of means-tested social welfare policy have been liberals — specifically neoliberals.

Social programs that were established in the early twentieth century tended to be universal in their scope, as Theda Skocpol notes in her book The Missing Middle. In 1912, for example, the US Children’s Bureau was established explicitly to investigate and mediate “all matters pertaining to the welfare of children and child life among all classes of our people.” Children’s Bureau chief Julia Lathrop explained that it was “not an act of charity” but was instead “designed to emphasize public responsibility for the protection of life just as already through our public schools we recognize public responsibility in the education of children.”

Universal social programs are based on the principle, as single-payer health care advocate Dr Quentin Young would later put it, of “Everybody in, nobody out.” The goal is to establish a sense of mutual accountability and mutual benefit across society. The method is to enshrine cooperation and solidarity in the institutions by which the state distributes public resources.

In the latter half of the century, neoliberal Democrats became less amenable to expensive universal social programs and more smitten with means-testing, which allows the government to be more selective and therefore more flexible. But the means-testing frenzy produced an ideological phenomenon that Mike Konczal calls “pity-charity liberalism, where the conceptual project of the welfare state is to compensate the losers of society rather than broadly empower citizens.”

Instead of fighting for universal social programs, Democrats wound up being hostile to them. Think of Hillary Clinton slamming free public college tuition on the basis that Donald Trump’s kids would use it. Gone was the will to “emphasize public responsibility” through policy design, and the idea that government assistance is society’s collective responsibility to itself administered via the state. The implication instead was that public programs should be reserved only for the neediest: a contribution from the rich to the poor, meant to slightly alleviate inequality — and justify the hyper-marketization of the economy — while capitalism proceeds apace.

Today’s mainstream Democrats do believe the government should provide assistance; they just want it to be administered selectively rather than comprehensively. The problem is that means-tested programs are very vulnerable to attack.

For one thing, the income cut-off point is usually very low, as with Medicaid. This means it slices right through the working class, dividing them into the eligible and the uneligible, the deserving and the undeserving, those who need charity and those who must fend for themselves. This division guarantees widespread ambivalence about these programs within the very constituency of people who would — and in other countries do — mobilize to defend public programs when they come under attack: the working-class majority. The bifurcation creates a breeding ground for stigma and resentment, especially racial and cultural animosities, which elites can capitalize on to make a popular case for shrinking programs that should instead be expanded. This is exactly what the Trump administration is doing when it says that Medicaid recipients need to be disciplined into the workforce.

Another problem with means-tested programs is that they’re confusing for the public to understand, and therefore identify with and be moved to defend. The maze of state-by-state eligibility criteria leave even people who rely on means-tested programs puzzled about who qualifies for them, how they work, and what values or social commitments they’re premised on. This confusion makes it difficult to convince people of the basic underlying thesis that animates, for instance, mass anti-austerity mobilizations in Europe: that these programs are ours. They are paid for by society for the benefit of society, and anyone trying to take them away is an enemy of that society.

This lack of public understanding and support for public programs turns them into punching bags, since politicians know that they can screw with them with a far smaller risk of significant or coherent popular pushback. There’s a saying that programs for the poor make poor programs, as they tend to be over-complicated and underfunded. It might also be said that targeted programs make easy targets.