Two years ago Paul Ryan, speaker of the U.S. House, argued that Americans "are no better off today than they were before the war on poverty began in 1964." The poverty rate, he explained, stood at 15 percent — the same as in the mid-1960s.

Last month the White House's Council of Economic Advisers (CEA) reached a completely different conclusion. The war on poverty, it proclaimed, was "largely over and a success," with only 3 percent of Americans now poor.

Clearly, both cannot be right. In fact, neither is.

Ryan's preferred measure, the Federal Poverty Level (FPL), is perfectly orthodox, yet perfectly absurd as a guide to how poverty has changed over time. It was based on food costs in the 1960s, using a rule of thumb that these were one-third of a family's budget. Today's FPL is, in essence, the cost of food for a family in the 1960s multiplied by three and adjusted for inflation. This rickety measure undergirds America's entire welfare system, determining the flow of hundreds of billions of dollars a year.

The FPL is misleading partly because food now accounts for under 13 percent of the average household budget — much less than half a century ago. The measure also makes no adjustment for variations in the cost of living. The poverty threshold in New York City is the same as in eastern Kentucky. Worst of all, the measure relies on pretax income. It misses savings and government benefits like subsidized health care, food, rent assistance and tax credits.

In short, the standard measure of poverty ignores the effects of antipoverty programs. No wonder the rate has hardly budged.

Other (superior) measures are more encouraging. The Census Bureau's Supplemental Poverty Measure accounts for antipoverty programs as well as variations in living costs. It suggests that poverty has ebbed. Social Security has helped cut poverty among old people, while food stamps and other programs have filled children's bellies. In terms of living space and access to amenities like air conditioning and dishwashers, the poorest 20 percent of Americans live about as well as the middle class did a generation ago.

The CEA uses yet another measure, known as the consumption-poverty rate. This reports what households spend rather than what they earn. Similar measures are often used in poor countries, where most work is informal and hard to track. But how much consumption is too little? Researchers usually peg the measure to consumption levels in a particular year, known as the "anchor year," then estimate changes. Choosing different anchor years gives wildly different results.

The consumption-poverty measure that produced the heartening 3 percent figure, for example, was pegged to consumption levels in 1980. Had 2015 been used as an anchor year, the estimate would be a much less impressive 12.7 percent. "I don't think there's a lot of meaning to that 3 percent number," says Bruce Meyer, an economist at the University of Chicago who helped devise the consumption-poverty index.

Measuring poverty always entails tricky value judgments, as Adam Smith observed in "The Wealth of Nations" while meditating on a linen shirt. Though ancient Greeks and Romans lived comfortably without linen, he wrote, "a creditable day-labourer" of the 18th century "would be ashamed to appear in public without a linen shirt" for fear of betraying a "disgraceful degree of poverty." Today much the same could be said of cars or smartphones.

Poverty can be defined in absolute terms — as having too little to pay for life's essentials. For example, 4.9 percent of American households experienced "very low food security" in 2016, meaning that a family member went hungry at least once. But that measure seems too austere in a country as rich as America.

Despite their differences, the White House reaches the same conclusion as Ryan did in 2016: More poor people should work in return for benefits. The White House is encouraging states to impose work requirements on Medicaid, the health-insurance scheme for the poor. The government would like to apply similar rules to rent subsidies and tighten existing eligibility requirements for food stamps.

Such moves are unlikely to cut poverty much, if at all. Welfare reforms in the 1990s under President Bill Clinton, which tied payments to work, succeeded in boosting working, says Lynn Karoly, a senior economist at the RAND Corporation, who examined the effects of welfare reform for the Department of Health and Human Services. But because benefits tapered off as people earned more, poverty did not fall.

It would be foolish to remake the welfare system on such shaky statistical foundations. America has made great strides against poverty in the last half-century. That is an argument for carrying on, not reversing course.

Copyright 2013 The Economist Newspaper Limited, London. All Rights Reserved. Reprinted with permission.