Poor accounting…and the poor get poorer? Sure, if you leave out half their income

National Review, Nov 29, 1993 by Bruce Bartlett

... and the poor get poorer? Sure, if you leave out half theft income.

RECENTLY, the U.S. Census

Bureau released its latest estimate of poverty in the United States, indicating that 14.5 per cent of the U.S. population, some 37 million people, lived in poverty in 1992. In the years such data have been kept, the poverty rate has been as high as 22.4 per cent in 1959 and as low as 11.1 per cent in 1973. Since 1980, the poverty rate has hovered around 14 per cent, moving only slightly in either direction.

Looking at such data, it is easy to conclude that we as a nation are not doing much, if anything, about poverty. In truth, we are doing a great deal, but by definition virtually all of our efforts are excluded from the data. This is because poverty is measured exclusively on the basis of pre-tax money income, excluding capital gains. (Members of the Forbes 400 could technically qualify as poor if they chose to draw 100 per cent of their income in the form of capital gains or simply borrowed against assets rather than draw a salary or receive interest or dividends,)

More significantly, the data exclude a large share of the government's efforts to alleviate poverty; the Census Bureau counts only cash grants, and not such in-kind benefits as Medicare, Medicaid, food stamps, school lunches, and rent subsidies. Including these items reduces the poverty rate from 14.5 per cent to 11.7 per cent.

Another important factor inflating the poverty rate is the mismeasurement of inflation during the 1970s because of the way housing was treated in the index. Instead of measuring changes in the rental value of housing, the index assumed that people bought a new house every month. In 1979, the Bureau of Labor Statistics fixed the CPI to measure housing costs more accurately, by switching to a rentalequivalent measure. However, previous years' measures of inflation were not revised accordingly.

Although the Census Bureau has not altered the official definition of poverty (the poverty threshold varies according to family size and age; in 1992 it was $14,335 for a family of four), it has calculated what the effect on the poverty rate would be if inflation had been correctly measured all along. It did this by constructing an alternative inflation index, called CPIU-X1, which conforms to the current measure, and using it to adjust poverty thresholds back to 1968. The outcome of this adjustment is to push about 3.6 million people out of the artificial poverty category and to lower the poverty rate from 14.5 per cent to 13.1 per cent. The inclusion of in-kind benefits would lower the rate further, to 10.3 per cent. (See graph below)

The point of this is not to belittle the plight of those with low incomes, but to demonstrate that all measures of poverty are essentially arbitrary and subject to conceptual error and measurement error. Reducing the economic status of the poor to a single measure, the poverty rato, is an extremely superficial way of looking at the problem of poverty.

The same may be said of income distribution in general, which is often reduced to a single measure, the Gini index. The higher the Gini index, the greater the inequality of income distribution. In 1992 the Gini index stood at .425.

Unfortunately, many of the same problems which plague the definition of poverty also distort the measurement of income distribution. In particular, the distribution of income is based solely on money income and, most importantly, excludes all taxes. The table shows the important effect that the inclusion of taxes and in-kind government benefits has on the distribution of income.

The impact of the adjustments is to lower the Gini index from .425 to .380, a significant reduction in income inequality. The result of relying on one standard measure of inequality is to make income appear much more unequal than it actually is, increasing pressure on Congress for punitive taxation of the rich and more benefits for the poor, which, ironically, will have no effect on the incomedistribution statistics because all taxes and most government benefits are excluded from the definition of income. Thus no matter how much we raise taxes on the rich it will not lead to greater income equality unless it discourages them from earning income; the taxes themselves have no effect at all on the standard measure of inequality. The well-being of the American people cannot be summarized in a few statistics, however well computed. Unfortunately, some of those that we rely on most heavily--the poverty rate, the Gini index--are conceptually flawed because they exclude the vast bulk of the government's redistributive efforts. They make us feel worse off than we really are and encourage government policies that may make matters worse. At a minimum, policymakers should be aware of these problems and at least take into account broader, more conceptually sound measures of income and poverty, which are easily available.

  SHARES OF TOTAL INCOME BY QUINTILE, 1992

  Standard  Adjusted
Quintile  Measure   Measure* Difference

Lowest    3.8       4.9        28.9%
Second    9.4       11.0       17.0%
Third     15.9      16.7       5.0%
Fourth    24.1      24.0      -0.4%
Highest   46.8      43.3      -7.5%