667 August 19,1988 U.S. INCOME DATA GOOD NUMBERS HIDING EXCELLENT NEWS INTRODUCTION The Census Bureau later this month will release its yearly update of family income statistics and poverty data for the United States. These are the "official" figures used to analyze the economic conditio& of Americans. They will provide a picture of how poor and middle class and minority Americans fared in 19

87. If the new figures continue the pattern of the past four years, they will confirm the strong improvement in the condition of middle class American families and show another solid decline in the official poverty rate.

Ironically, as good as this pattern has been, deficiencies in the methodology used by the Bureau cause the statistics to understate the economic well-being of Americans.

In last year's update, covering 1986, the Census Bureau reported that the real median family income was $29,458, up a dramatic 4.2 percent from 1985 and the largest increase in 15 years. Real family income changed little between 1973 and 1978, but there h ad beenpal precipitous decline from the second year of the Carter Administration until the economic expansion began late in 19

82. The improvement in 1986 meant that U.S. median family in come was up 10.7 percent since 19

82. This has been the sharpest an d most sustained real im provement in U.S. family income in 20 years All Sections Gaining. This improvement has left no segment of America untouched. All regions of the country have gained, ranging from the Midwest's 7.2 percent to the North 1 By definiti o n, one-half of all American,families are above the.median income and one-half are below. This is a more useful way to analyze income distribution than by using "averages" because aver.ages are distorted by very large or small incomes. Median, however, fai l s to take account of differences in family easts 13.6 percent. The income of black fdlies rose at an even greater rate than for white families. Median black family income grew a startling 14.0 percent between 1982 and 1986 Some critics contend, however, t h at a dark cloud lurks behind this silver lining. They point to statistics that indicate that the number of American families deemed middle class is declining and proclaim this bad news. The truth is, however, that the middle class is shrinking because 4 m illion famjlies have moved to the upper income brackets between 1981 and 19

86. In fact, since 1969 the proportion of families in both the middle and lower classes has declined while the proportion of families with an income of $56,000 or above has more th an doubled Historic Low. There also has been a steady reduction in the proportion of Americans below the official poverty line. This line indicates the threshold income considered neces sary by the government for Americans to avoid poverty? In 1986,13.6 p e rcentof the population was categorized as poor, down from 15 percent in 1982 Percent of Americans in Poverty white IndMdUah t Among black families the poverty rate of two parent families in 1986 was almost one-third less than in 1982, the sharpest decline of any racial group. Indeed, the poverty rate among in tact black families was at an historic low.

Even this excellent news understates the im provement. For techni cal reasons, the median family income calcula tions of the Census Bureau are lower than I similar calculations by the Congressional Budget Office (CBO) and the.Commerce.z.

Departments Bureau of Econ&mic Analysis. fhese government agencies, using a more reli 2 3 The declining middle-class thesis: a sensitivity analysis, Month& Mor Review, May 1988.

The poverty threshold was designed in 1%5 by the Social Security Administration. The thresholds varied with the family size, the number of children under 18, and the sex and age of the family head. In addition adjustments were made if the family resi ded on a farm. Distinctions based on sex and farm residence were dropped in 19

80. The original threshold was based on the U.S. Department of Agricultures estimate of the relative cost of an adequate diet for each type of family 2 able system of measureme nt, report an even more rapid improvement in U.S. family in come. 4 In addition, it should Taking In-Kind Benefits Into Account. Similarly, major deficiencies in the official Cen sus Bureau measurement of income grossly overstate poverty in America. One o f the most serious flaws is that in-kind income, such as medical benefits, subsidized housing, and food stamps, is not counted as income. The Census Bureau recognizes this deficiency and has attempted to measure the cash value of the principal federal welf a re benefits. In 1986 the Bureaus calculations showed that including in-kind benefits as income would have reduced the poverty rate from 13.6 percent to either 11.6 or 9.0 percent, depending on the method used to value in-kind benefits. This means that fro m 2.6 million to 11 million Americans were erroneously identified as poor by the official method of determining the poverty rate. The most dramatic effect was upon the elderly poor, whose poverty rate plum Year sou- cenrlmBum4u I Herllage InfoChyt meted fr o m 12.4 per cent to a mere 3 percent when a market value was placed on in-kind benefits including medi cal assistance To provide a much more accurate picture of how poor families are faring, the Census Bureau report should count all cash and non cash incom e and allow for geographic differen ces in the cost of living Median Family Income Total 1 986 Dollars 24 S t A N D S THE PATTERN OF IMPROVEMENT The Census Bureau last year reported that by 1986 real median family income had risen 10.7 percent since 1982, w hen the last economic recession ended. There were comparable increases for all regions of the country, and for both men and women. There were also solid gains by all races: in fact, black families enjoyed the largest percentage increase in income 4 CBO ha s found family income up 11 percent from 1973-1986, and the Bureau of Economic Analysis calculated that per capita income was up 21 percent during that same period. Tmnh in Fumify Income 1970-1986, Congressional Budget Ofice, February 1988 3 Percent of Fam i lies in Poverty 11;1;.1 1 I-Famiiia All Families of any group in recent years. There has been an enormous increase in the number of black middle class families This up wardly mobile class of blacks grew by a third between 1988 and 1986, from 3.6 mil lion to 4.8 million. The black middle class now constitutes 40 percent of all black households, larger than either the black working class or the black poor.

This increase in black incomes has been due mainly to the explosive growth in jobs over the past five y ears. Although blacks account for 11 percent of the workingage population, they have filled 15 percent of the new jobs created, and the black jobless rate has been slashed by almost half since 1982 The Importance of Intact Families The economic boom in re cent years is also the prin cipal cause of the sharp decline in the poverty rate which had begun rising alarm ingly in 19

78. Yet although the statistics show that the rate is linked to the nations economy, they show some thing else that is just as impor t ant: poverty is related closely to family status. In 1986, for in stance, only 6.1 percent of the 51.5 million married-couple families in America were 20 18 16 14 12 Percent of Intact Families I W .v n 10 t 8 6 below the Bureaus poverty threshold. Among t he 10.4 million female householder families, however, the poverty rate was 34.6 percent. The great majority of these female headed families are single mothers with children under age

18. As a distinct group, these single mothers have a poverty rate of 46 percent, in comparison with a rate of 8 percent for married couples with children.

Since 1960, between 30 percent and 40 percent of female-headed families have been below the poverty line. But since that time the total number of such families has more tha n doubled in number, while the number of intact families has risen by only 30 percent, hold ing back the improvement in overall family income. Yet this great increase in the non-for 5 Joseph Perkins, Boom Time for Black America, Policy Review, Summer 1988 4 mation or breakup families over the past twenty years is not reflected in the Bureaus tables of family income statistics. These are misleading if used to gauge the income of some sup posed average family. The statistics do not trace the condition of an average family; they are an average of two groups of very different families one group consisting of intact families, who have been doing well, the other consisting of female-headed families, in which the poverty rate always has been appallingly high.

This difference be tween intact and broken families is particularly impor tant in distinguishing fact from fiction regarding the condi tion of the black family. The Bureau reported last year for instance, that 28.0 percent of black families in 1986 were offic i ally poor. This was more than three times the rate for white families and Percent of Black Familiesln Poverty 60 torBkklntact reinforced the popular idea of a wide divide between blacks and whites. Yet this was an families. Among black families with child r en, 54.2 percent in 1986 were headed by women while among white families the proportion was 16.5 percent. Only 10.8 percent of married couple black families were below the poverty threshold. This was the lowest level since 1974, when the Bureau began to k e ep statistics based upon family status, and less than double the rate for married whites. But an astounding 50.1 percent of female householder black families were poor. This latter category had a disproportionate effect both on the overall black poverty r a te and on comparisons with white families average. Closer examination of the income tables reveals a very different picture of black 66 Unless the numbers behind the widely-quoted averages are understood, federal efforts to combat poverty are doomed. Marr i ed couple families have an excellent chance.of escaping poverty, while female-headed families suffer from an enormous degree of poverty. Thus if poverty is to be reduced, the underlying causes of broken families must be addressed, and federal government p o licies to encourage families to form and stay together must be adopted How the Numbers Understate the Improvement The Census Bureau reports of substantial gains in family income during recent years un derstate the actual economic growth enjoyed by America n families. Between 1973 and 1986, for instance, the Census Bureau indicates that real family income suffered a small net loss, with the improvement since 1982 more than offset by earlier declines. This has led 5 some critics of the Reagan Administration t o claim that Reagan policies have left American families no better off than 15 years ago.

This contention hides the fact that families now are gaining ground lost in the 1970s. As important, it overlooks some of the deficiencies of the Census Bureau statistics. Other government studies suggest the gains reported in recent years by the Census Bureau are too modest, and point to a considerable improvement over 19

73. For instance, an alternative Commerce Departmen t study, released this year and undertaken by the Bureau of Economic Analysis, reports that real per capita income rose 21 percent from 1973 to 1986 Moreover, recent updates by the Bureau of Economic Analysis show that between 1982 and 1987 the rise in in c ome was an astounding 12.9 percent. The reason for this startling diver gence from the Census Bureau calculations is that the Bureau of Economic Analysis ex amines individuals rather than families. This thus corrects for changes in the size and nature of f amilies. The Bureau of Economic Analysis takes into account noncashbenefits, such.as employer-paid health insurance, food stamps, and Medicare in measuring well-being. And in place of the Consumer Price Index (CPI this study uses the personal consumption ex penditure deflator to measure purchasing power in constant dollars. The CPI has been widely criticized as having greatly overstated housing costs in the 1970s thereby lowering the rate of income growth during that time!

Using a still different method of measurement, the Congressional Budget Office (CBO recently concluded that family income rose about 11 percent from 1973 to 19

86. Like the Census Bureau, it measured income growth for all families, but this study took into account economies of scale in l arger families and the declining size of the average family since the early 1970s. To avoid overstating the inflation rate, the faulty CPI was replaced by an alter native real dollar conversion. The CBO, like the Census Bureau, ignored all noncash in come or benefits, citing difficulties in obtaining hard data with respect to private sector non cash benefits. Yet the CBO explicitly recognizes that in not including in-kind benefits, real gains in family income are understated, and it notes that this underst a tement has become progressively greater as in-kind income increased from 11 percent of personal income in 1980 to 18 percent in 1984.7 Thus the CBOs figures for the changes in family income since 1973, although far rosier than the Census Bureaus, still ar e a low estimate.

Though Census Bureau calculations of the number of Americans in poverty indicate the trends in recent years, the serious shortcomings mean that the official poverty rate over states dramatically the number of Americans actually in poverty . Among the..most serious flaws are 6 7 CBO, op.cir pp. 9-11 The Politics of Numbers, for the National Committee for Research on the 1980 Census (New York Russell Sage Foundation, 1987 pp. 88-92, CBO, op. cir pp. 6-9 6 1) Ignoring In-kind Benefits The Cen s us Bureau states in Poverty in the United States, 1986, that [its poverty statistics are based on money income only and do not include the value of noncash benefits such as employer-provided health insurance, food stamps, or Medicaid. This is a very signi ficant omissiod About 70 percent of all poor families receive at least one of the following means-tested, in-kind benefits: food stamps, Medicaid, public housing, or school lunches.

Federal assistance to low-income families in the form of non-cash programs had an es timated market value in 1986 of $59 billion. Yet not a dime of this federal assistance is in cluded by the Census Bureau in determining income and thus poverty rates. If a family with a cash income of $1 below the poverty line, for instance, qu alifies for $5,000 worth of housing and food assistance, the family is counted by the Census Bureau as in poverty.

Recognizing this enormous shortcoming, in recent years the Bureau has been trying to find a way to value noncash or in-kind benefits? By usin g a market value concept noncash benefits would be valued at the cost of the specific goods or services in the private market place. By this method, the cash value of food stamps could be assessed reasonably easily since the cost of food can be determined by supermarket prices. But subsidized housing and medical benefits would be more difficult. Among other things, the market value of real es tate in many poor neighborhoods is distorted by the very existence of subsidized housing.

Medical assistance is eve n more difficult to evaluate as a contribution to income; few Americans would consider that $20,000 worth of hospital care to a poor family struck by a serious illness makes that family $20,000 better off. The Bureau points to difficulties in find ing har d data on comparable goods in the private sector.

The second method being explored is the recipient or cash equivalent value; this reflects the .value the recipient places on the benefit. Again, the major problem the Bureau identifies is the lack of hard data.

While officials grapple with the problem of valuation, however, policy makers and the American people are being seriously misled regarding the extent of poverty in America. If the cash income of a family is below the Cens.us Bureaus poverty threshold , as calculated by the Bureau, it is simply declared to be officially poor. Yet these thresholds reveal very little about the economic well-being of any given family 8 9 U.S. Department of Commerce, Poverly in the United Stutes, 1986, Current Population R eports, Consumer See The Census Bureaus Estimates of Poverty Including the Value of Noncash Benefits: 1986, Technical Income Series, P-60, number 160, page 1.

Paper 57, pages 2-10 7 The average food benefit received by an eligible family has an estimated market value of 1,400 per year, and medical benefits a market value of $2,8

90. Housing assistance has been valued by the Census Bureau at $1,780.1 A family receiving all three of these benefits would not have its poverty status affected in the least, as f ar as the Census Bureaus official poverty statistics are concerned. Yet if these benefits were counted as income at their market value, the Bureau estimates that 11 million fewer Americans would have been categorized as poor in 1986, reducing the official poverty rate from 13.6 percent to 9 per cent. Failing to include these estimates in the published poverty numbers thus introduces an enormous upward bias in the rate used in most policy debates.

The U.S. Senate noted in 1980 that it was essential, at the earliest possible date, for the official poverty statistics to be corrected to reflect the effects of in-kind benefits. Without such information, said the Senate, Congress and the Executive Branch cannot be certain that Government transfer programs are pr o perly targeted. Yet when the Bureau issues its report this month, eight years after the Senate expressed its frustration, the numbers will continue to reflect only certain forms of cash income. It still describes its estimates of the cash value of in-kind benefits as experimental 2) Ignoring unreported income The Bureau cautions that there is a tendency in household surveys for respondents to un derreport their income.12 This underreporting also seriously detracts from the reliability of these numbers in m e asuring poverty. Compared with some independent estimates, only about 72 percent of aggregate money income is reported in the Bureaus surveys. About two-thirds of the probable unreported income is imputed to surveyed families by the Bureau. As a result, t o tal income used in the Bureaus annual statistics, comprising both reported and imputed, still is only 90 percent of independent estimates of income Attempts to take some account of underreporting of income do not even consider inten tional non-reporting o f income from,the widespread underground economy. This income would include those who work for cash to avoid taxes in addition to the income from illegal activity 10 In determining the market value of Medicaid benefits, total medical benefits paid are divi d ed by the number of persons enrolled in the program. Under this method the average value of medical benefits was 2,890 per year. Under the recipient value concept, the value of medical benefits to a recipient family is assessed according to the average am o unt spent for such benefits by unsubsidized families with the same level of income. Under the recipient value method, medical benefits were assigned a cash value of just $700 per year. Technical Paper 57, op. cit., pp.2-10 11 Technical Paper 57, Appendix A, op. cit., p. 152 12 Current Population Reports, Series P-60, number 160, op. cit., p.

2. Data used in poverty analysis come from the Bureaus March Current Population Survey Supplement which obtains detailed information on inco me and work experience from some 60,OOO interviewed households. Answers to survey questions may be provided by any member of the household who is 15 years of age or older 8 3) Ignoring Regional Living Costs No distinctions are made in the Census Bureau po v erty statistics between the living ex penses of a family in an expensive inner city area, such as Manhattan, and a family in a lower cost rural setting; such as Appalachia. Families in both places are assumed to need the same income to escape poverty. Tho ugh it is useful to have national data, geographic variations must be taken into account to measure deprivation 4) Incorrectly Indexing for Inflation.

Indexing is used to revise the poverty income threshold to account for increases in the cost of living. T he index converts income into constant dollars. Since 1969 the Census Bureau has used the Labor Departments Consumer Price Index (CPI But this index has been widely criticized for its inability to measure certain important factors affecting the pur chasin g power of the p00r.l The CPI measures increases over time in the cost of a specific basket of goods and ser vices, but does not allow for the widespread practice of substitution, where a family elects to buy a less costly alternative if the cost of an ite m increases. For instance, a family might buy chicken if the price of beef rises. Until corrected in 1983, the CPI had overestimated the cost of housing substantially; this caused about 4 million people to be categorized er roneously as poor. The published values of the CPI for years prior to 1983, however, have not been corrected. Thus the Census Bureaus figures continue to exaggerate the rate of in flation during the 1970s, artificially increasing the number of families below the poverty threshold during that period.

In developing the CPI in 1917, the Bureau of Labor Statistics attempted to measure the cost of goods and services bought by the typical American urban, blue-collar family. It has been modified several times since 1969, to take into account wor king and unemployed single individuals as well as families. Yet, it continues to measure the average consumption of only the urban population, ignoring the different consumption patterns of the rural p0pu1ation.l~ In studying family income, both the Burea u of Economic Analysis and the CBO have rejected the CPI in favor of more comprehensive price measurements which do not overstate the rate.of inflation IMPROVING THE MEASUREMENT OF POVERTY Despite these deficiencies of the Census Bureau poverty rate calcul a tions, the Bureaus poverty statistics are taken as a hard measurement of poverty in America. In fact, all that they provide is a guide to changes in poverty levels, though this is very important l3 77ie Politics of Number, op. cit pp. 88-105, and CBO, op. cit pp. 3-9 14 The Politics of Numbers op. cit pages 88-92 9 The largest federal welfare programs, Aid to Families with Dependent Children (AFDC and Medicaid, do not, in fact, use the federal governments official poverty threshold as the basis for eligibi l ity. Recognizing the shortcomings of the Census Bureaus measurement of poverty, these programs use standards of need set by the states. Under the eligibility criteria for these programs, assets and in-kind assistance are counted, and geographic varia tion s are taken into account. That is why many who are officially poor receive no welfare assistance Though it is impossible, using the current official statistics, to measure the impact of government spending on the poor, Congress continues to commit billions of dollars to anti poverty programs. The long-overdue revision of the Census Bureaus methodology would give policy makers and taxpayers a clearer picture of the extent of poverty in America and to assess the success or failure of government programs desig n ed to reduce it. Among the needed reforms are 1) Include in-kind benefits in the official income and poverty figures The Census Bureau should not release a greatly exaggerated poverty rate if the true rate is dramatically lower when noncash benefits are i ncluded. The Bureau should include the value of noncash benefits in calculating the official poverty numbers for its annual report.

Alternative estimates excluding such benefits, should be published in an appendix, as should a discussion of various alterna tive methods of placing a value on in-kind benefits 2) Include assets in the calculations In defining poverty the Bureau should not confine itself to measuring only cash or non cash income. Relevant also to a familys economic well-being is the home that i t may own as well as savings accounts, orstocks or bonds. The Bureau should experiment with a method to incorporate the value of assets in the financial status of families, and it should include in its annual report a discussion of methods of evaluation 3) Improve the correction for underreporting Income which is widely underreported no longer should be ignored. Experts agree that declared income is underreported by 10 percent, even after the Bureau makes its own revisions for legitimate income overlooked b y recipients. The Bureau does not include any estimate for income from the legal and illegal underground economy. The Bureau should increase its estimates for unreported income and experiment with methods of measuring in come gained from the underground ec o nomy 4) Use a consumption standard for assessing need and cost of living In place of the CPI, a minimum market basket of goods should be established which would fluctuate based on family size. This would be used to calculate changes in the cost of living. The market basket would represent goods and services typically consumed by the poor with allowance made for geographic differences in consumption patterns. The con sumption standard approach would give a much more accurate picture of what the poor ac 10 c tually need for a minimally decent standard of living. States then would use this new stand ard to determine the poverty threshold income for different areas and to set eligibility standards for programs. The Census Bureau should calculate the total numbe r of poor ac cording to these state thresholds 5) Create a Dependency Index The annual Census Bureau reports should include an estimate of the additional number of Americans who would be in poverty in the absence of means-tested cash or noncash assis tance . This Dependency Index would allow policy makers to gauge the impact on the poor of the more than $100 billion spent each year by states and the federal government on cash and b-kind means-tested programs. In particular, it would enable Congress to deter m ine how many Americans escaped poverty because they became truly independent, and how many depended on government assistance to avoid poverty. If the goal of federal wel fare policy is to help families achieve economic independence, the official estimates must measure progress or lack of it in reaching this goal CONCLUSION The Census Bureaus annual income and poverty statistics are assumed by most Americans and lawmakers to provide an accurate picture of the economic conditions of families and to serve as a reliable guide. They do not. They are able only to present a pic ture of the trends in these conditions, and in recent years those trends have been excellent.

But because the statistics systematically underestimate income, they give an unduly pes simisti c impression of poverty in America. And because the official numbers hide the most significant factor leading to poverty broken families the Census Bureau statistics are an unreliable guide for public policy. Fundamental reforms of the Bureaus methodology are required, and lawmakers should take careful account of the need for those reforms when they try to interpret the data for 1987 when they are released later this month.

Kate Walsh OBeirne Visiting Fellow 11