(Bloomberg View) —When Bill Clinton signed welfare reform into law, the outrage from the left was incandescent. Peter Edelman, a prominent official in Health and Human Services, resigned in protest and wrote an article for the Atlantic calling it “The Worst Thing Bill Clinton Has Done” and declaring that “it will hurt millions of poor children by the time it is fully implemented.” And that was one of the nicer responses.

Senator Daniel Patrick Moynihan was really critical: Sixty years of social policy, he says, is being mindlessly dismantled by “the monstrous political deception embodied in the term ‘welfare reform.’” Millions of children will join the ranks of the homeless trying to get a little warmth by sleeping on the grates in our city streets. “The defenders of the old activism toward the poor,” complains Moynihan, “surrendered willingly, with the shrugs and indifference of those who no longer believed in what they stood for.”

As the 1990s wore on into the 2000s, some of the critics began to moderate. Poverty didn’t rise as the doomsayers had predicted. Other programs, like the earned-income tax credit, were increased to offset declining income from Temporary Assistance for Needy Families. Even in the depths of the Great Recession, the poverty line never exceeded the peak it had reached in 1992, so it was hard to argue that welfare reform had somehow dramatically worsened the material circumstances in which poor people live.

Hard — but not impossible. Averages can conceal a lot of variation. A lot of people on the left have worried all along that poverty was making poor people better off on average, but hurting the very poorest among them — people with mental illness, substance abuse problems, and other issues that made it difficult for them to work. And many of those people have children.

This is approximately what Jason DeParle showed in his path-breaking book “American Dream,” which followed three mothers receiving welfare through the process of reform: Two of the women improved their circumstances somewhat, but the third got addicted to crack and ended up much worse off than she probably would have been if the government had still been sending her a check.

It is also what’s suggested by “$2 a Day: Living on Almost Nothing in America,” a recent book by Kathryn Edin and H. Luke Shaefer, which has received a lot of attention for its chronicle of people living on what seems like an impossibly low sum.

With another Clinton now running for president, the issue of whether welfare worked obviously has great political salience.

Even if worried progressives are right that the poorest of the poor are doing worse since welfare reform, that doesn’t settle the question. A life of work offers cumulative benefits: You gain good work habits and experience, which enables you to earn more money in the future, and sets an example of self-reliance for your children to follow. The experience of welfare reform suggests that welfare was enabling people to make a rational short-term decision — welfare benefits are better than working an entry-level job — that was a disastrous long-term decision, because the longer you go without working, the harder it is to get your foot on the first rung of that vocational ladder, and the less time you have to enjoy any cumulative benefits.

A government policy that helps people with bad long-term decision-making was probably not “the greatest good for the greatest number.” On the other hand, making the badly off even worse off doesn’t sound like great policy either. Which problem you think is more important is a value judgment, not an empirical question.

But what if we don’t have to choose? Scott Winship of the Manhattan Institute has just released a paper arguing that “Children — in particular, those in single-mother families — are significantly less likely to be poor today than they were before welfare reform.” How can he come to such a radically different conclusion from other authors? Because official measures of poverty aren’t very good.

That’s a known problem, which Christopher Jencks, now an emeritus professor of social policy at Harvard, exhaustively explored last year for the New York Review of Books.

The measure of inflation used to calculate the poverty line is known to be flawed in a way that pushes up estimates of poverty. People underreport income to survey takers, a problem that may have gotten worse in recent years. Social changes have also affected the reliability of the measure: For example, more people cohabit instead of getting married, but the Census Bureau counts two unrelated people living together as essentially living in two separate households. Noncash benefits have also vastly increased, allowing poor families to stretch their limited funds further, but those too are excluded by the Census. Our poverty rate is calculated as if we were still living in 1960.

The question, of course, is how we ought to calculate it. The one thing that can be said for the official measure is that it’s been the same for a long time. That makes it easy to compare to past periods, and also means that we don’t have to squabble about what number to use. Those are great benefits when you’re getting into a policy argument — though I think Winship makes a pretty thorough case that at this point, the drawbacks of inaccuracy outweigh the benefits of consistency.

But Winship will probably be less successful at convincing liberals that his measures are the right ones. The charts he provides offer multiple ways of calculating poverty, and while these are extraordinarily useful for policy wonks, they also highlight the fact that there is no one way of looking at poverty.

For example: Should we treat cohabiting couples as if they were married? Cohabiting cuts down on expenses, yes, but both incomes might not be available to other members of the household in the way that we’d expect in a family with two married parents.

Or: Should we count health-care benefits? Medicaid probably frees the poor up from some spending, which means they have more cash for other things, but how much? Some of the health benefits they consume would have in any case been provided for free; prior to Medicaid, the U.S. had a fairly elaborate network of charity care, particularly in urban areas. Or the patients might simply have gone without. In either case, we can’t fairly say that a Medicaid benefit with a market value of hundreds of dollars a month is exactly the same thing as getting cash income. Winship settles for calculating the value of Medicaid benefits at a quarter of their costs, which is not unreasonable, but will probably also be not uncontroversial.

But before we can get to these questions, we need to get to an even more fundamental one: “What does it mean to be poor?” And that’s a surprisingly hard question to answer. Most poor people in America — even those in deep poverty — are not starving, freezing to death, forced to go without shoes or clothes, or in danger of dying from some easily treatable disease. Those poor Americans who do have those problems are often suffering from substance abuse or mental health issues that affect their ability to care for themselves even when a safety net is available. And America’s rather unique view of civil liberties makes it more difficult than it is in other countries to, say, forcibly institutionalize someone who is simply incapable of keeping themselves sheltered and fed even with cash assistance. In the 19th century sense of the word, we have already largely won the war on poverty. And yet, looking around, we still seem to have a lot of poor people.

One problem is that we’re often talking about “relative poverty”; Jencks suggests that this sets in roughly when someone’s income falls below half the median income in their community. But the idea is not new. Adam Smith laid out similar thoughts in “The Wealth of Nations”:

A linen shirt … is, strictly speaking, not a necessary of life. The Greeks and Romans lived, I suppose, very comfortably though they had no linen. But in the present times, through the greater part of Europe, a creditable day-labourer would be ashamed to appear in public without a linen shirt, the want of which would be supposed to denote that disgraceful degree of poverty which, it is presumed, nobody can well fall into without extreme bad conduct.

“The poor you will always have with you,” said Jesus, and by a relative measure, that’s tautologically true: there will always be a bottom of the income distribution, even if that distribution is very narrow and much higher than it used to be. Should we really try to assuage this by government intervention? Can we? This is really the question we’re debating when we argue over whether food stamps should cover steaks.

Or should we try to assuage the constant financial anxieties that come with having a low income? The lack of capital that makes it hard to buy a decent car or appliance, the months in which inflow won’t quite cover outgo, so you have to choose between fixing the car, paying the gas company, or buying your daughter a prom dress? Is freedom from financial worry a basic component of a decent life? Should kids have to skip the prom because their parents are broke?

None of these questions have a simple answer that everyone can agree on. But before we can talk about a better measure of poverty, we need to at least establish the criteria for what we’re calling poverty. Given the divisions over these questions, we may end up with several measures that describe different things. But that’s probably more useful than one measure built to gauge the poverty of the last century, and currently gauging not much.

This column does not necessarily reflect the opinion of the editorial board or Bloomberg LP and its owners.