Whatever you might think of President Bush's faith-based initiative, you have to give him credit for ecumenical ingenuity. The whole scheme builds on a plan promoted by conservative Protestants. The head of the program, John DiIulio, is a Catholic. His chief adviser, former Indianapolis Mayor Stephen Goldsmith, is Jewish. Much of the experiment will be focused on Philadelphia, whose mayor, John Street, is a committed Seventh-Day Adventist.

It is not a huge program, but it is a big deal. Bush's promise to funnel more federal money through faith-based organizations ranks with tax cuts and education as a signature item in his budget. It represents Bush's determination to show that the "compassionate conservatism" he talked so much about in last year's campaign was more than just a slogan and that the "armies of compassion" will actually be called out.

Nothing has incited more dispute in Bush's young administration, and very likely nothing will. Traditional liberals and libertarian conservatives are horrified at the whole faith-based idea. Liberals say the proposals amount to government support of religion and thus violate the constitutionally required separation of church and state. On the other hand, how could organizations that practice motivation by faith be asked to change their approach just because they are getting money from the federal government?

Then there is the criticism that, religious freedom aside, government might be dumping social problems onto charitable organizations unable to bear the load. And there is the fear that community organizations might become so dependent on government cash--and the strings attached to it--that all the ingredients that previously made them successful would be stifled.

But while the enemies of the faith-based program have been raising those arguments, a coalition of strange bedfellows on the other side has been taking shape. Neo-liberals and neo-conservatives have joined forces in defense of the initiative, reasoning that, given the sorry record of so many government-managed social programs, faith- and community-based efforts are worth a try for purely pragmatic reasons.

At the core of these battles are three big questions:

First, is this really new? We crossed the line a generation ago in funding community-based and religious organizations through Model Cities and the Great Society programs that followed it. The Johnson administration launched that initiative for reasons similar to those that created this one: zeal to return power to the people and a pragmatic hope that community-based programs would succeed where others have failed. So Bush's proposal builds on more than 30 years of federal practice--and the continuing disputes that have dogged it.

That raises the second question: Do faith-based efforts really work better? The only thing that can be said with conviction is that no one knows for sure. The evidence is all anecdotal. Philadelphia's efforts seem impressive. The city has new community-based truancy programs and an effort in which religious congregations adopt convicts on their release from prison. In other cities around the country, from Boston to Indianapolis, leaders of faith-based efforts tell remarkable stories of success.

But they are stories, not proof. There has been some empirical research on the faith-based phenomenon: Studies by DiIulio at the University of Pennsylvania and by the Manhattan Institute in New York have found that African-American teens who attend church are much less likely to use drugs. But there isn't enough research yet to prove that managing social service programs through faith-based organizations succeeds more often over the long haul than more conventional programs. Critics, in fact, have complained that DiIulio (Disclosure: He is a sometime scholarly collaborator of mine) is doing little more than shouting "Geronimo!" and hoping people follow him.

The third question is this: If faith-based programs do work, can they be expanded without undermining the forces that drive success and without violating the constitutional church-state separation? This is the toughest question of all.

The fairest way to evaluate this experiment will be to judge it by what it accomplishes, rather than how it does so. There are numerous ways to structure faith-based federal assistance: through performance contracts that purchase job training, tutoring, truancy reduction, and other techniques. How the organizations work out the details should be up to them, within constitutional guidelines and requirements to report their results.

Just such a strategy is already being pilot-tested in Florida, Maine and Kansas. These tests should help us focus government money on buying outcomes, not funding religion.

However it is structured, faith-based social policy will be a daunting managerial and political stew. We'll be fighting about it for a long time. But if stay focused on results, we can at least improve the odds that we'll be fighting over the right issues.