Roughly 38 million Americans, about one of every seven, are officially classified as poor by the U.S. government. But this figure is about 19 million more than are actually poor, according to some economists.

This poverty-rate overstatement—which comes about largely by the failure to count Medicaid, food stamps, school lunches, and other noncash federal welfare payments as income—can potentially trigger the spending of billions of extra taxpayer dollars a year on antipoverty programs.

The official poverty rate— the latest figure is 15.2 percent for 1993—is released every year by the Census Bureau and has become one of the most important economic statistics in the United States. It is a number that attracts much attention in part because it is a vital factor in determining how billions of welfare dollars are distributed in the country but, more important, because it is interpreted widely as a barometer of fairness in our society

On the spending front, the magnitude of poverty is often cited to justify increases in welfare outlays. Thus, if the level of poverty is exaggerated, a stronger argument is provided for greater spending. It’s difficult to quantify this impact, because the relationship between spending and poverty runs both ways.

Increasingly, economists recognize that higher levels of welfare spending create labor-market disincentives that may lead to increases in the official poverty rate, especially if the additional outlays are in the form of noncash benefits.

It is something of a vicious cycle, according to these experts. More official poverty creates more spending, which, in turn, creates more official poverty.

Unfortunately, many economists agree, the poverty statistic has through the years become an increasingly flawed measure.

At the inception of the poverty rate in the 1960s, defining who is poor and who not was done by estimating how much money would buy the food to provide a minimally adequate diet.

To take into account a person’s other expenditures, the cost of such a diet was then multiplied by three to arrive at the final poverty-level income.

Thus, initially, the determination of what constituted poverty was a matter of expenditures, not just money income. However, it was convenient to use money income, which is easier to measure, as a proxy for expenditures.

As time passed, the initial poverty-threshold level of money income was updated each year to take account of price inflation. Today’s estimates of the poverty threshold are merely constant-dollar versions of the 1960s numbers. This updating process is one source of the deterioration in the quality of the poverty-rate statistic.

This could occur because the current Consumer Price Index overstates the rise in prices since the 1960s by about 10 percent, a figure that is both significant and widely accepted, even by the Census Bureau itself. This means that both the current poverty thresholds and the poverty rates that go with them are overstated. The real link between income and expenditures has been altered.

In addition to the bias in the official price index, the income measure has been distorted by the exclusion of noncash forms of income.

Currently, less than one-fourth of all government benefits for the needy are in the form of cash aid. For the most part, the programs that provide noncash benefits did not exist when the original definition of poverty was developed.

It is possible to assign numerical values to these biases, although some of them are more speculative than others. The firmest numbers come from the Census Bureau, which has been mandated by Congress to calculate alternative estimates of poverty that take into account the effect of noncash forms of income on poverty status.

The Census Bureau also employs an alternative price index that is free of the distortions mentioned earlier. Making both of these adjustments reduced the poverty rate by an average of 5.2 percentage points a year over the period 1989—1992.

The Census Bureau made these calculations on the basis of its workers’ interviews with individual citizens, who were asked to estimate how much cash and noncash income they received from the government. It’s here that another major error has crept into the calculation of the poverty rate. The amount reported in the interviews, when compared with government records, was found to be significantly understated.

In terms of cash aid, the surveys show only about 70 percent as much as government records. And the under-reporting of noncash benefits is even more severe. The Census Bureau’s valuation of these is less than 40 percent of what government data say is the total spending in this category.

If all benefits were counted as income at their full value, the official poverty rate might be overstated by another 4.5 percentage points. That, however, is a maximum estimate. A very conservative approach would be to take one-third of that amount—1.5 percentage points—and add it to the 5.2 percentage-point overstatement already identified. The combined error, 6.7 percentage points, is not quite one-half the mean annual poverty rate of 13.75 percent for the period 1989—1992.

In fact, a good rule of thumb to get an accurate reading on the poverty situation might be to take one-half of whatever official rate is reported.

Others have made the argument that the official poverty rate is overstated, although not by as much as indicated here. Based on work done in conjunction with Professor Dale Jorgensen of Harvard University, Daniel Slesnick, a professor at the University of Texas at Austin, recently calculated poverty-rate estimates that use a poverty definition that is in the spirit of the original approach. They emphasize patterns of consumer expenditures as the measure of well-being and report a poverty rate for 1989 of 8.4 percent, slightly less than two-thirds of the official rate.

Because this estimate exceeds the Census Bureaus minimum 1989 estimate of 7.7 percent, the Slesnick poverty rates would seem to represent a more conservative estimate of the extent to which official poverty data err in an upward direction.

A few final words: Official poverty statistics in the United States, most economists agree, are quite suspect and should be approached and interpreted with great care, especially when used to formulate policy. There are times when it is better to know nothing than to know something very imperfectly. This may be one of those times.