Mitt Romney has accused the Obama administration of trying to “gut welfare reform by dropping the program’s work requirements.” That may be a slight exaggeration, but the administration’s plans do raise serious concerns. As the CEO of one of the nation’s largest workforce-development companies and the president of the think tank that developed the intellectual arguments behind welfare reform, we urge the administration to reconsider.

The welfare reform signed into law by President Bill Clinton in 1996 transformed the nation’s welfare program from a lifetime entitlement to a system that aims to move people off welfare and into the world of work. The new name, Temporary Assistance for Needy Families, symbolized the change.

The program is still administered by the states within federal rules. But the reform set time limits (two years consecutively; five years lifetime) on benefits and required that at least half of a state’s able-bodied welfare recipients engage in 30 hours of “defined work activities” each week. And, lest states evade the law’s intent, Congress defined “work activities” in great detail.

But last month the Department of Health and Human Services issued an “Information Memorandum” inviting states to apply for waivers from some of the work requirements in order to run “demonstration projects” that would consider “new, more effective ways to meet the goals of TANF.”

The Obama team insists it’s merely tweaking the program to offer states more “flexibility.” But the memorandum includes troubling changes that would effectively water down the requirement that welfare recipients work, actively search for work or seriously train for work.

For example, it encourages projects that “improve collaboration with workforce and/or post-secondary education programs to test multi-year career pathway models; . . . test systematically extending the period in which vocational education training or job search/readiness programs count toward participation rates” and similar jargon. All translate to a reduced emphasis on rapidly moving people from welfare to work.

These types of efforts came into fashion in the 1960s — and studies have shown they don’t work.

In another project, a state could “count individuals in TANF-subsidized jobs but no longer receiving TANF assistance toward participation rates.” Translation: The state could count individualsno longer getting welfaretoward its work-participation rate, effectively lowering the work requirement foractualwelfare recipients.

All these “projects” represent a return to the failed, fuzzy-headed thinking of the decades preceding the landmark 1996 reform, when welfare policy focused on endless training and education programs, treatment programs and other non-work-related activities.

America Works was founded in 1984 on a different set of ideas — the belief that the best social program is a job — a real job. Over the years, it’s found employment for 300,000 hard-to-place workers, including welfare recipients, providing them with structure, dignity and hope, as well as a paycheck.

Also in 1984, Charles Murray, then a Manhattan Institute senior fellow, published the book “Losing Ground,” which forever changed the national discourse on welfare reform by empirically illustrating how no-strings cash assistance fostered a dependency that hurt the poor.

It took more than a decade, but Washington eventually recognized that the “work first” approach works best, and reformed welfare around that principle.

The 1996 law is generally heralded as a historic, bipartisan success. Welfare rolls have plummeted, and the poverty rate is still lower than it was in 1996, despite the prolonged economic slump. Even in this tough economy, the number of New Yorkers on welfare is about 350,000 — a far cry from the 1.1 million (one out of every seven city residents!) on welfare in the mid ’90s.

In the wake of worries that it’s rolling back reform, the Obama administration has suggested that it will offer waivers only to states that commit to moving at least 20 percent more people from welfare to work. But if it’s going to let states water down work requirements in the ways suggested above, their “work participation” rates could easily rise 20 percent without any more recipients actually working.

If the administration wants to let states undertake “demonstration projects,” these experiments must not decrease or replace the current work mandates.

Welfare reform is working: Why would we want to roll it back?

Lee Bowes is America Works’ president and CEO. Lawrence Mone is president of the Manhattan Institute for Policy Research.