U.S. Rep. Paul Ryan's discussion paper on "Expanding Opportunity in America" has much to admire. It puts attention where it belongs — on poverty and upward mobility for those at the bottom of the income distribution. It argues for an expanded earned income tax credit, the best way to reduce poverty in work. And it correctly explains what's wrong with existing federal programs. When it comes to reforming those programs, though, Ryan's proposal leaves a lot of questions unanswered.

The essential challenge is how to provide financial support to the poor without reducing their incentive to work. As earnings increase, means-tested benefits are withdrawn, creating an implicit tax on higher earnings. Paying a universal benefit regardless of income would deal with that, but it would be prohibitively expensive. Reducing the benefit at every level of income would lessen the disincentive, too — there'd be less to tax away — but only by making the poor worse off.

One way to mitigate the trade-off is to deny benefits to people who aren't working. The EITC supplements wages; if you don't have a wage, you don't qualify. This only gets you so far, though. Even with a work test in place, higher incomes reduce benefits, so there's still an implicit tax on working harder (full time rather than part time, say) or on seeking a better-paid job. The best you can do is to simplify and coordinate the benefits so that their interactions don't cause this hidden tax to reach punitive levels.

Ryan's idea is to bundle some programs — assorted housing subsidies, the Supplemental Nutrition Assistance Program, Temporary Assistance for Needy Families, the Weatherization Assistance Program and others, 11 different schemes in all — and replace them with a single block grant to states wishing to participate. These states would contract with competing one-stop providers to deliver, in total, the current level of funding for payments and services to the needy.

They'd have to meet certain requirements. Except for the elderly or disabled, beneficiaries would have to work or "engage in work-related activities." And the states would have to monitor various measures of success — families lifted out of poverty, percentage of people finding work or getting off assistance, growth in wages, high school graduation rates and so on. All this, it's hoped, would spur innovation. "For too long, the federal government has tried to supplant, and not to support, the people fighting poverty on the front lines — families, neighborhoods, community groups."

The standard objection to this approach is to recoil at the very idea of block grants. Progressives see a predictable conservative ruse to convert programs into forms that can be more easily cut or diverted to other purposes. Desirable as it may be to foster policy experiments in the states, this suspicion seems amply justified. Many Republican state governments have refused to expand their Medicaid programs even though federal taxpayers under the Affordable Care Act would foot nearly all the bill.

I also see other problems. Yes, Ryan is simplifying the Washington end of the federal income-support programs by folding many of them into a single payment to states — but so what? Whether this makes things better for beneficiaries and the taxpayers funding their benefits is a different question. The answer depends on how the providers design their interventions.

Ryan isn't designing a better system — he's delegating the design of a better system to the states and their contractors. There's no guarantee in his proposal that the providers would come up with something more cost-effective, or more conducive to effort and ambition, than what you see in the chart.

Maybe competition to design better schemes would discover good models — that's the hope. But this will depend on exactly how the rules for providers are specified; Ryan leaves this vague. Would the best models, once discovered, be rolled out across the country? That's unclear as well. In any event, the underlying trade-off between means-testing (for the sake of cost control) and work disincentives would remain, just as before.

Another big problem with Ryan's state-led approach is that it works against a prompt response to cyclical contractions of the economy. Federal taxes and spending (especially payments automatically tied to rising unemployment) play a key role in stabilizing the economy, nationally and regionally. Ryan, to his credit, recognizes the issue and suggests ways of making the block grants countercyclical — for instance, by linking the grants to each state's unemployment rate. That would help, but the risk of weakening the automatic fiscal stabilizers — which, if anything, ought to be strengthened — is serious.

Still, it's great that Ryan is discussing the right subject. It's also great that he frames the issue so broadly to include criminal-justice reform and innovation in education. He says he wants to start a conversation. In that, let's hope he succeeds.

Clive Crook is a Bloomberg View columnist and a member of the Bloomberg View editorial board.