The Census Bureau tomorrow will release the latest statistics on poverty in the United States, the income level of an average household and the number of Americans still lacking health insurance.

Don't believe the numbers.

A growing chorus of experts and politicians is raising questions about the data that frame Americans' understanding of their nation's well-being. From poverty levels to health insurance, inflation to personal savings, widely accepted statistics are overstating some problems and understating others, miscounting people, and sending policymakers down blind alleys.

"We're getting at best an impressionistic sense of what's going on in the economy," said Rep. Rahm Emanuel (D-Ill.), who recently introduced legislation to establish an independent commission aimed at overhauling government economic statistics. "Major policy decisions are being made based on data that is inadequate to the task."

This seemingly technical problem has real-world consequences, allocating federal assistance to some who don't need it while cutting off others who do, raising the costs of programs like Social Security, or pushing policies for problems that may not exist.

For example, since poverty levels are not adjusted for regional costs of living, the working poor in expensive urban centers like Washington are routinely excluded from federal programs because their income lifts them above the official poverty line. The rural poor in low-cost states like Arkansas often can afford considerably higher standards of living than their urban compatriots. Yet they may be eligible for food stamps, housing aid, free school lunches and other programs that exclude the urbanites.

In March, Michelle Billups, 42, began working full time in the dining hall of the Washington charitable group So Others Might Eat. She earns $8 an hour to support herself and her 17-year-old daughter Shannon. This month, Billups was told she is no longer eligible for $225 in food stamps, a program available for District residents with incomes up to 130 percent of the federal poverty line. Billups' $16,640 annual income is $153 higher than that threshold for a family of two.

In a study to be released Thursday, the liberal Economic Policy Institute estimates that a family like Billups's, with one parent and one child, requires an annual income of $47,460 to meet its basic needs in Washington. That family in Fayetteville, Ark., would need $24,096.

Perhaps no statistic has more critics than the poverty rate, which in 2003 stood at 12.5 percent, the latest Census data available. University of Chicago economist Robert T. Michael, who chaired a National Academy of Sciences panel tasked to update poverty statistics a decade ago, called the current poverty data "truly awful."

"The poverty statistics are absolutely wrong," agreed Rebecca M. Blank, dean of the University of Michigan's school of public policy, who served on the panel.

Officially, the poverty rate has drifted upward since 2000, from 11.3 percent to 12.5 percent in 2003. But a more sophisticated measurement that the Census also publishes, which accounts for variable costs of living, rising medical expenditures and more accurate price inflation, shows the official rate has consistently understated poverty. By that alternative measure, the percentage of Americans below the poverty line has risen from 12.8 percent in 2000 to 14.2 percent in 2003. Using such measurements, last year the Democratic staff of the Joint Economic Committee found poverty rates nearing 16 percent in the late 1980s.

At the same time, household incomes may be understated because they do not include non-cash income like food stamps. The earned income tax credit was created during the Reagan administration specifically to raise the working poor out of poverty. But by government counting, the program has not lifted a single person above the poverty threshold, Michael said. Since poverty rates are based on pre-tax income, refunds like the earned income credit do not count.

Another problem? Double counting. With a $10-an-hour clerical job at Howard University, Marilyn Bryant, 41, earns too much money to qualify for any federal assistance, yet is poor enough to live in the Thea Bowman House, which is heavily subsidized by So Others Might Eat. Her three-bedroom apartment rents for $349 a month, but her $400 weekly income is deceptive: A quarter of it is extracted from her paycheck as child support for her daughter, Natasha Carter, 15, who lives with her father. The Census Bureau still records Bryant's income as $400 a week, Michael said, but it also includes $100 of that in Carter's household income.

"I don't know how I do it," Bryant said. "I really don't."

For more than 20 years, the Census Bureau has been developing alternative poverty measures, many of which answer these criticisms, said Charles Nelson, the Census's assistant division chief for income, poverty and health statistics. But it is up to the White House budget office to change the official measurement, and successive administrations have declined to do so.

Scott Milburn, spokesman for President Bush's Office of Management and Budget, said "a consensus has yet to emerge in the scientific and policy communities on the critical statistical elements of a new methodology."

The Census Bureau on Tuesday will also update its count of the uninsured, and any change from the 2003 figure of 45 million likely will be slight. But recent studies by the Urban Institute and the Annandale-based Actuarial Research Corp. concluded the number is overstated by 4 million to 9 million people, largely because it includes Americans already enrolled in Medicaid and other federal health care programs.

Joseph Antos, a health policy analyst at the American Enterprise Institute, estimates that the true policy problem of the uninsured may center on about 20 million truly needy individuals. Out of the Census Bureau's 45 million, nearly 15 million report household incomes of more than $50,000 and could afford to buy at least catastrophic health plans. Another 4 million are already eligible for government programs but have not enrolled. Millions more actually receive government-funded health coverage.

From the conservative Heritage Foundation to the more liberal Brookings Institution, economists agree the government's basic measurement of consumer price changes is overstating inflation. As a result, tax collection has been depressed, since tax brackets rise with inflation. Government spending on programs like Social Security has been excessive, since such programs enjoy annual cost-of-living adjustments based on the current consumer price index. And labor contracts have been distorted by built-in inflation protections.

The Labor Department's standard consumer price index measures the cost of a basket of goods in urban areas as they rise over time. But since 2000, the department's Bureau of Labor Statistics has also tracked more realistic spending patterns, allowing for the substitution of products when prices spike. This "chained" CPI, for example, might substitute a pound of chicken for a pound of beef one month if steak prices have shot upward, said David S. Johnson, assistant BLS commissioner for consumer prices and price index.

Switching to this more sophisticated measurement from now through 2014 would cut $70 billion from Social Security payments while raising income tax collection by $83 billion, according to Brookings Institution economists. Yet Congress has made no effort to change the official inflation measurement, in part because lawmakers have no desire to slow the growth of either tax bracket increases or Social Security benefits.

"This is a political decision, and no one wants to make it," said Fritz Scheuren, president of the American Statistical Association.

More recently, a debate has begun over the nation's savings rate, which officially hovers just above zero. When Congress returns in September, the House Ways and Means Committee will try to put together legislation to raise personal savings through tax credits and other incentives. But according to David Malpass, chief global economist at Bear Stearns & Co., the United States is accumulating savings hand over fist. The country's pool of liquid savings grew by $1.5 trillion last year, he said, and U.S. households remain the world's largest creditor, with $37 trillion in financial assets.

The problem, Malpass said, is that the official savings rate measure does not consider economic gains from patents, innovation, capital gains or land appreciation.

"We may be throwing billions of dollars at a problem that isn't there," said Emanuel, who has advocated savings proposals.

Marilyn Bryant, right, with daughter Renee, earns too much money to qualify for any federal assistance, yet is poor enough to live in subsidized housing.