(Archived document, may contain errors)

710 June 6,1989 DISPELLING THE MYTH OF INCOME INEQUALJTY INTRODUCTION Capitol Hill has been debating what is alleged to be an enormous income gap between rich and poor in America. Some lawmakers claim that this income inequality is excessive and reflects a flaw in federal tax and spending policies.

Thomas J. Downey, the New York Democrat who chairs the House Ways and Means Subcommittee on Human Resources, even charges that the democracy.

Criticisms like Downeys are unfounded because they are based on faulty government data. Government income distribution statistics, for example, do not count as income a substantial portion of government assistance to low-income households. The latest U.S. Census Bureau report on income distribution fails to include a staggering 98 billion of government spending on low-income and elderly persons (equal to 50 percent of all such spending).

These omissions make an accurate comparison of the incomes of poor families with the rest of society virtually impossible and invite false charges about income inequality.

Remarkable Equality. If the full value of government spending were included in estimating the incomes of poor families, the figures would show not an al arming degree of inequity but a remarkable level of economic equality. When the full value of government assistance is counted, the average per capita income among households in the poorest one-fifth of US society turns out to be about 60 percent of the p e r capita income in median American households, much higher than generally assumed government statistics fail to reflect the degree to which income disparities are current income disparity in the United States threatens the health of a Yet the notion of an excessive income inequality is in fact a myth 2 -4 Another factor contributing to the myth of great inequality is that offset by differences in family size or are a result of differences in the number of workers in a family. Low-income families generally a re small. And among these families, work is quite rare, while total reliance on government is common. There are over six full-time workers among households in the fifth of the population with the highest income for each full-time worker in the lowest fift h. America is increasingly divided not between the rich and the poor but between families with two ormore taxpaying:workers and families dependent on government funds, in which no one works.

Future reports of income inequality by the federal government sho uld be corrected to include all government assistance to low-income and elderly persons. Such studies also should clarify the extent to which differences in household income are affected by reasonable variations in family size, in the number of workers pe r family, in education, and in productivity levels HOW THE CENSUS BUREAU UNDERCOUNTS ASSISTANCE TO THE POOR 1 Kate Walsh OBeirne, U.S. Income Data: Good Numbers Hiding Excellent News, Heritage Foundation Backgounder No. 667, August 19,1988 2 U.S. Bureau of Census, Current Population Reports, Series P-60, No. 164-RD-1, Measuring the Effect of Benefits and Tares on Income and Poverty: 1986 (Washington, D.C U.S. Government Printing Ofice, 1988 3 Means-tested programs are directed to persons With limited income . They constitute the public welfare system, in which recipients qualify for benefits based on need. Social insurance programs, such as Social Security and Medicare, have no income test for eligibfity and so are nonmeans-tested When the Census Bureau compil e s its annual statistics on income and poverty, it traditionally counts only before-tax cash income. These statistics constitute the official measurement of income and poverty in the United States and form the foundation of virtually all debate and policy. Yet the Bureau ignores the effect of such noncash programs as Medicaid, housing subsidies, and food stamps, which ow make up about 73 percent of all government assistance to the poor. Because of these flawed procedures, the Bureau systematically exaggerat e s the extent of income inequity and poverty Remedying Deficient Estimates. In a special study entitled Measuring the Effect of Benejits and Tares on Income and Poverty: 1986, released in December 1988, the Bure3u partially corrects this glaring omission i n its widely used annual report. This study analyzes the distribution of income amonghouseholds, using both the official Census Bureau definition and alternative income definitions that take account of other types of cash and noncash income and subtract ta x es. These alternative definitions partially count some government transfers, including some nonmeans-tested cash and noncash programs, such as Social Security and Medicare, and some means-tested cash and noncash programs, such as public assistance Medicai d , food stamps, and housing subsidies? When these 2 noncash benefits are included as income to the poor, the Bureau calculates that the official poverty rate of 13.6 percent for 1986 actually falls to 11.6 percent. This would reduce the number of Americans categorized as poor by 4.8 million. More dramatic, using this alternative definition, the poverty rate for the elderly would drop from 12.4 percent in 1988 to just 5.7 percent.

Although Qese ,new Census .Bureau calculations.constitute a more accurate esti mate of the financial well-being of American families, the Bureaus latest report is still deficient. For instance, the Bureau attribyies only $49 billion in means-tested benefits and services to the poor in 19

86. By contrast, other government sources est imate that government at all levels spent a total of 126 billion on such assistance? Thus the Bureau dismisses $77 billion in government aid as having no income value to its low-income recipients Similarly 71 billion was spent on Medicare in 1986, yet the Bureau identifies only $50 billion of thig as having increased the share of national income received by the elderly. Overall, the government fails to count 98 billion in government aid to low-income and elderly persons 50 percent of the total. This uncoun t ed aid is given in four ways: cash assistance, noncash assistance, services, and Medicare benefits I Uncounted Cash Assistance One element of this missing $98 billion comes from deficiencies in the calculation of cash assistance In its recent, improved me a surement of income and poverty in 1986, the Census Bureau still ignores $14 billion or almost 38 percent of government cash aid given to low-income persons. The Bureau attributes 23 billion in means-tested cash assistance to the poor, including public ass i stance (Aid to Families with Dependent Children and state welfare payments) and Supplemental Security Income (federal payments to the aged blind, and disabled In a more comprehensive analysis, however, the nonpartisan Congressional Research Service of the Librairy of Congress (CRS) identifies $37 billion in government cash aid in 19

86. The difference arises in part because the CRS 4 Census Bureau expenditure figures throughout this study are calculated from the aggregate income data provided in U.S. Burea u of Census, op. cit, pp. 18-19 5 Vee Burke, Cash and Non-Cash Benejits for Persons with Limited Income: Eligibility Rules, Recipient and Expenditure Data, FY 1985-87 (Washington, D.C The Congressional Research Service of the Library of Congress, August 1 , 1988 p. 2The Congressional Research Service (CRS) report shows $148.7 billion in expenditures from all levels of government on federal means-tested programs in FY 1986.The CRS data however, include expenditures on persons living in nursing homes and menta l hospitals, while such persons are excluded from Census data. To provide a fair comparison to Census figures, therefore, all CRS data alluded to in this study have been adjusted to eliminate expendimes on such institutionalized persons. Expenditures on in s titutionalized persons in 1986 equalled approximately $225 billion. Deducting this figure from the published CRS totals on means-tested spendq leaves a total of $126 billion in spending on non-institutionalized persons 6 U.S. House of Representatives, Bac k pund Material and Data on Progmms within the Jutisdiction of the Committee on Ways and Means, 1989 Edition (Washington, D.C U.S. Government Printing Oftice, 1989 p 152 7 Burke, op. cit p. 2.The CRS report shows $41 billion in cash means-tested aid in FY 1 9

86. After deducting for expenditures on institutionalized persons, the total comes to $37 billion 3 report includes veterans pensions, the Earned IncomeTax Credit (EITC foster care, and emergency aid payments, which provided cash to needy families in 198 6 but are ignored by the Census Bureau.

The Bureau explains that its statistics do not include the effect of the EITC because of a technical problem that will be corrected in future reports. Yet this omission iq the latest study makes the studys conclusio n very misleading since the EITC is the governments most effective meanis of assisting low-income working families In 1987, some 6.3 million families claimed an average credit of 327 per family per year. This refundable credit added $2.1 billion to the in come of poor working families in 1987 Uncounted Noncash Assistance by the Bureaus failure to take into account $44 billion of means-tested noncash assistance to the poor included by the CRS in its analysis.

The Bureau assigns a value of 26 billion for Medicaid, food stamps, school lunches, housing assistance, and other noncash benefits for 19

86. In contrast the CRS reports that federal and state governments spent $70 billion on such noncash assistance in 1986.8 Why is there this difference expenditures? I t overlooks entirely other medical benefits provided to the poor, such as veterans medical benefits, help from the Maternal and Child Health Block Grant, and medical services from Community Health Centers.

These three programs alone represented about $4 billion of government-provided medical care in 19

86. Not only does the failure to fully count these medical benefits understate the real income of poor families, it The Census Bureaus 14 billion undercount of cash assistance is dwarfed One reason is that the Bureau includes only a portion of Medicaid 8 hid 9 Some argue that medical benefits should not be counted as income because the sicker you are, the richer you appear. But the value of medical benefits is not determined in that fashion.The Census Burea u calculates the income value of medical benefits based on the insurance or market value that represents the average cost of these benefits to an employer or the government. Example: In calculating the value of employee medical insurance for a family of fo u r, Census determines the average cost of such coverage to an employer and then adds that amount to the familys income. Similarly, in determining the income value of Medicare, Census calculates the average Medicare expenditures per elderly person in each s t ate and then adds that amount to the income of each elderly individual who is eligible for Medicare within the state. Census follows this procedure consistently except with low-income households where the value of medical benefits is counted at partial co st or ignored completely 4 accentuates the degree of inequality between families, since employer-provided medical coverage is included in the income of more fluent households."

Another reason for the disparity in CRS and Census Bureau figures is that despite counting the value of food stamps and school lunch subsidies, the Bureau neglects 0the;;supplements that impqove the economic well-being of poor families. For example, the S pecial Supplemental Food.Program'for Women Infants, and Children (WIC) provides nutritional aid, generally in the form of vouchers, to low-income mothers and young children.The average benefit was $396 a year in 19

86. Similarly, the CRS reports that in 19 86 the federal government spent a total of $13.2 billion on housing subsidies for low-income families through welfare programs as wells federal housing programs. Yellthe Census Bureau counts only 6.5 billion in housing benefits for the poor Uncounted Serv i ce Assistance The CRS recognizes the huge amount of government aid in services to the poor that the Bureau ignores, including day care, job training, education, and legal assistance. The federal and state costs of these programs amounted to 19 billion in 1 986.12 At least a portion of this assistance should be counted as income, since these setvices have economic benefit to their recipients and would be counted as routine expenditures by middle-class and fluent Americans. Thus including the value of some of these services is essential not only for an accurate picture of the condition of the poor, but also for meaningful assessment of income inequality. Example: If a low-income family receives free day care services while a middle-class family must pay for th o se same services out of its income, then the free service received by the low-income family should be considered part of its income when compared to the middle-class family. The Census Bureau ignores this discrepancy 10 One of the reasons the Census Burea u undercounts or ignores the value of Medicaid stems from a confusion between measuring "poverty" and measuring "income distribution The Census Bureau feels that an individual who has expensive medical coverage should be counted as poor if he has not enoug h money to pay for food and rent, even if the total economic resources available to him (including medical coverage) exceed the "poverty income threshold The Census Bureau diswunts the value of medical coverage to such an individual to assure that he conti n ues to be counted as "in poverty But income distribution measures the overall economic resources available to one individual in comparison to others; arbitrarily undercounting the value of medical benefits to low-income persons while counting those same o r similar benefits at full value for more fluent persons inherently misrepresents the distribution of income in the U.S. Deliberate undercounting of government benefits is incorrect in measiring income distribution and problematic in measuring "poverty."

W hile this study deals with measuring income distribution, technical problems in the measurement of poverty will be examined in a future paper 11 U.S. Bureau of the Census, op. cit p. 207 12 Burke, op. cit p. 2 5 I Uncounted Medicare Benefits The fourth ma j or missing element in the Census Bureaus calculation of government assistance involves medical services to the elderly. The Bureau understates help for poorer elderly households while counting most benefits for richer households. Specifically, the Census R eport considers the value of Medicare benefits as being equal to the insurance value or average.cost of such benefits to the government. However, while the Census Bureau counts Medicare benefits at their full value for affluent families, it counts the sam e Medicare benefitsply partially to low-income elderly families or leaves them out entirely. This procedure artificially increases the apparent level of inequality. Over& the Bireau fails to count $21 billion in Medicare benefits to low-income elderly pers o ns Undercounting Spending on Low-Income and Elderly Families I (in billions of dollars Burke, op. cit p 2. Expenditures on insthtionalized persons have been deducted from the CRS totals Census of the Census, op. cit pp. 18-19 Overall Impact of Uncounted B e nefits calculations of the Census Bureau and those of the CRS By counting virtually every possible benefit as having income value to higher-income families while failing to count billions of dollars in government transfers to the poor, the Census Bureaus r eport gives a false impression of the condition of low-income Americans, and thus of the financial differences between income classes. It is this indefensible treatment of family resources by the Bureau that has ignited charges in the media and in Congres s that America is an unacceptably unequal society and that government transfers have been The table indicates the cumulative effect of these disparities in the income 13 Example: If the Census Bureau determines that a particular elderly couple does not hav e enough income to fully meet the cost of food and rent, then the Medicare coverage received by that couple is deemed to have no income value. For a more affluent elderly couple, the Census Bureau will add the full cost of Medicare coverage for the couple i n calculating the household income 6 insufficient to reduce income inequalities. This conclusion bolsters calls for further taxation and income redistribution methodology, the government could double its spending on many low-income programs and have absol u tely no effect on income distribution as measured by the Census Bureau.The Bureau thus should not even profess to be measuring the distribution of !aggregate household income? if it includes certain benefits as part of the income of the upper and middle c l ass but excludes the same benefits when counting the incomes of poorer families. It cannot arbitrarily assume that billions of dollars in government spending on the poor have no income value and can therefore be ignored. The Census Bureaus faulty model th u s tells policy makers very little about the relative well-being of U.S. families. An accurate measure of income distribution in the U.S. must represent the share of total economic resources made available to individuals at different economic levels Tellin g Policy Makers Little. In fact, using the Bureaus faulty ASSESSING THE DEGREE OF INCOME INEQUALITY Charges of excessive inequality arise not only from these flaws in the measurement of income considerable as they are -but also because the Census Bureaus c omparison of families fails to take into account key differences in the characteristics of rich and poor households. This failure adds further erroneous ammunition to support the inequality myth.

A glaring problem with the recent Census Bureau study of ine quality in 1986 is that it fails to recognize the effect of differences in family size and in the number of workers per family on income inequality. Example: Suppose the Smiths have an income of $40,000, while the Jones have an income of 20,0

00. The conv entional Census Bureau analysis would conclude that the Smith household must be high income, while the Jones household was low income. Some policy makers might even conclude that such inequality represents aprima facie case for income redistribution to co r rect inequality. But if the Smith household consists of a husband, wife, and three children, while the Jones household is a recent college-graduate living alone, very few Americans would feel that the Smith household should be regarded as the more affluen t . Indeed, the per capita income of the Smith family would be $8,000, compared to $20,000 for the Jones.14 The Census Bureau income inequality data are riddled with paradoxes of this kind. Because they do not take into account differences in family size 14 Some analyses adjust for family size according to a complicated scale in which each added family member is given a reduced weight See Committee on Ways and Means, op. cit This procedure is unnecessarily complex and misleading. It confuses the objective me a sure of the distribution of income with a comparison of subjective economic utility. The distribution of income should measure the share of total economic resources allocated to an individual; for that purpose, the measures used in this study, per capita h ousehold income and per capita household income/per worker, are decidedly superior 7 conventional Census Bureau measures of income inequality exaggerate its level in the U.S Unrealistic Impression. Similarly, most Americans would agree that it is quite re a sonable and fair for a household with two workers to have a higher income than a household with one worker. But the Census Bureau analysis ignores the fact that much of the apparent income inequality in America is caused by households having different num bers of workers The American people and their congressional representatives need a more accurate impression of the degree of differences in income levels of families.

Unrealistic impressions lead to bad public policy. Fortunately, the Census Bureau numbers could yield a far better indication of income differences in the U.S. by adjusting for a few family characteristics, such as family size and number of workers. The resulting figures still would exaggerate inequality because of the 98 billion in governmen t assistance to low-income families omitted from the Census data. But despite these deficiencies, the high level of equality in America, after making only limited adjustments for differences in family characteristics, is remarkable. And if the undercountin g of government benefits to low-income persons were corrected, the level of economic inequality would be even smaller than the following analysis indicates How the Number of Workers in a Family Explains Inequality. A Chart 1 ranks all U.S. households accor d ing to total after-tax household in come for 1986, dividing them into five equal groups, or quintiles, according to the income of the households. This distribution then is compared with the number of workers in each household As the chart shows, the numbe r of workers in the household is a major factor influencing the degree of income inequality. The lowest fifth of U.S. households has only 19 full-time workers per hundred households; the most affluent fifth, on the other hand, has 134 full-time workers for each 100 householdsu Working vs. Non-Working. The gap between the high-income and low-in come Americans thus is primarily a gap between families that work to sup port themselves and families both elderly and nonelderly that perform litr tle or no work and depend on the government. Full-time work among the lowest fifth of households in fact is very rare. As Chart 2 indicates, over half of the households in the lowest quintile are elderly or nonelderly with no workers. Only 18 percent of households in the lo w est fifth have even one full time worker. In the next to lowest fifth of households, by contrast, work in creases dramatically as half of these households have one or more full-time workers 16 15 Throughout the text a full-time worker is defined as one wh o worked primarily at full-time jobs for 50 oi more weeks during the year. Data are from U.S. Bureau of the Census, op. cir column 11 on pp. 21,?7,79 16 hid, pp. 21,33,77,79, and

81. Due to overlap in some categories, the number of households with part-ti me workers in Chart 2 may be slightly underestimated 8 Chart 1 Average Number of Workers per Household 1986 Number of FulCTlme Workerr Lowest Second Mlddle Fourth Hlghert Quintile of Families ource: U.S. Bureau of the Census, op. cit., column 11 on pp. 21 , 77, 79 The disparity in the level of work performed by the lowest- and ughest-income quintiles is particularly striking. Among the lowest-income oup, less than one household in five has a full-time worker. In the est-income group, no less than 88 percent of homes have at least one Chart 2 Types of Households by Income Class: All Households 1986 I 2 FulCtlmo Worker 1 Full-The Worker Put-Tlmo Workere 0 Workwe Nonolderly h Eldorly Houroholde Lmot Second Mlddle Fourth Hlghoot Quintile of Families Horltage Inf o Ghart ource: U.S. Bureau of the Census, op. cit column 11 on pp. 21,33,77,79, and 81 9 Ill-time worker, and nearly half have two or more. The number of workers er family, in fact, is the single most important factor explaining the conomic status of a fami ly.

Since there are very few workers among the quintile of households with the west incomes, it should not be surprising to find as Chart 3 indicates, that ss than 25 percent of the income of these households comes from wages veri Census Bureau data indica te.75 percent of the income of these ouseholds comes directly from the government. If corrections were made Chart 3 Sources of Pre-Tax Income per Income Class: All Households 1986 I 100 0 0 76s 601 Y 8 25 ros 0 U ource: Calculated from U.S. Bureau of the C ensus, op. cit pp. 15-19 3r the government transfers that are omitted or undercounted in the Census bureau figures, at least 90 percent of the income of the lowest fifth of ouseholds would be attributable to government transfer programs louseholds at the m edian family income level, by contrast, rely'on the overnment for only 10 percent of their income. Ninety percent of the icome of these families comes from wages or other nongovernmental ources such as interest or dividends. Among the top fifthfif househo l ds, 98 lercent of income comes from nongovernmental sources 17 Calculated from Bureau of the Census, op. ciL, pp. l8-19 10 How Family Size Explains Inequality Another factor heavily influencing the well-being of various households is family size. Census B u reau data tom 1986 show that the average post-tax household income among the lowest fifth of households wk $6,805, while the median household income in the same year was $21,186.18 But in contrast with the conventional Wisdoa.thelypical lower-income -hous ehold is on average quite small; a majority of these households have only one or two members. Households at the other income levels generally are larger, with many more members to be supported by the family income.

Adjustment for the smaller size of households at the lower end of the income scale further reduces the apparent income disparity between individuals in the lowest fifth of households and those in the rest of society.

According to Census Bureau data, post-tax income per capita in the lowest fifth of households was $3,600 in 1986, or 42 percent of the post-tax per capita income in the median American household in the same year which was 8,585. l9 Moreover, since the Census Bureau figures systematically exclude half of the cash and noncash benefits and the targeted social services provided by the government to low-income households, this comparison still understates the overall levelof income equality. If the true size and value of government transfer programs are included, the average after-tax per capita income among the lowest fifth of households would be closer to $5,000 or nearly 60 percent of the per capita income in the median income household.20 Income Equality among Working Families tion of U.S. households with at least one full-time worker. Again, a superfi cial interpretation of the data suggests a relatively high level of inequality.

Among those households with at least one full-time worker, the average post tax income of the lowest fifth is only $1&766.The average post-tax income of Chart s 4 shows the Census Bureau figures on the post-tax income distribu 18 Unless otherwise noted, all income figures in the text are taken from the Bureau of the Census, op; cit. using income defintion #lO.This definition of income includes eamhgs, interest, dividends, capital gains, and pensions, and the value of private medical insurance. It also includes a partial count of government cash and non-cash transfers. Finally, it subtracts federal and state income taxes and Social Security taxes to determine pos t-tax family income 19 U.S. Bureau of the Census, op. cit, p.

19. Note that the Census report does not provide data on the median household.Throughout the text of this Backgrounder the term income of the median household refers to the mean income of the m iddle quintde of households.This figure is virtually identical to the real median, but is slightly higher. This substitution leads to a slight understatement of equality throughout the text of this Backgrounder 20 This calculation assumes, conservatively, that only half of the missing $98 billion h government assistance is allocated to households in the lowest 6fth 11 Chart 4 Comparison of Household and Per Least One Full-Time Worker Capiblncorrie~1086 Households With at n I Ss4000 Quintile of Families our ce Calculated from U.S. Bureau of the Census, op. cir p 7 and unpublished Census Bureau data households in the highest fi$p is $76,591, or six times greater than the income of those in the bottom fifth.

Disappearing Disparity. But the average household in the top quintileaof working house- holds contains nearly 50 percent more members than the bot tom quintile. Many of the households in the lowest group are single in dividuals, often young workers who have not yet started families. Nearly all of the househ olds at the top of the income scale, on the other hand, are large families with several children. After adjusting for family size, the apparent in come disparity between the top and bottom quintiles is greatly reduced.

Chart 4 shows that, among families with at least one full-time worker, the average per capita income of the bottom fifth for 1986 is $5,479 after taxes 21 Calculations based on the Bureau of the Census, op. cir p. 77,79, and unpublished Census tables.The households on pages 77 and 79 were redivided into new equal size quintdes based on additional unpublished Census data 12 the average gr capita income of the middle fifth is $9,598; and $22,330 for the top fifth.

But these figures still exaggerate the degre e of underlying economic ine quality among working families. Among those households in the bottom fifth shown in Chart 4, only 8 percent have two full-time working members while 54 percent in the top fifth have two or more full-time workers. Comparing the post-tax ixicome &stiibution of working families after fully adjusting for family size and number of workers per family, the most affluent fifth of households have a post-tax per capita income of 14,499 for each full-time worker.= The least afEluent fifth of households has an average post-tax per capita income of $5,069 for each full-time worker The remaining gap be tween the top and the bottom is primarily the result of differences in the productivity levels of the workers in each family. Workers in the m ore af fluent families generally are highly skilled professionals or managers.

Workers in the lower-income households tend to be younger and predominantly low-skilled blue-collar and white-collar workers.

A superficial interpretation suggests great income differences among work ing families: the average income of working households in the top quintile of 76,591 is six times greater than the average income of $12,766 among work ing households in the bottom quintile. But after adjusting for family size and t he number of full-time workers, much of this income disparity disappears the post-tax per capita income of the top fifth of households then is only 2.8 times that of the least affluent households. On close examination, in fact, the remarkable thing about U.S. society is not the alleged gaptween the rich and the poor but the astonishing overall level of equality.

Income Differences between Black and White Families comparisons between racial groups within U.S. society. For instance, raw Census Bureau data sh ow that the average black household has an income of just 57 percent of the average white household.This difference arises mainly This analysis of income distribution has important implications for 22 Ibid 23 The Census Bureau provides data only on the nu m ber of full-time workers in each family. A more complete picture of economic equality could be gained by including part-time and part-year workers.This should not however, affect significantly any of the major conclusions in this study 24 Calculations fro m U.S. Bureau of the Census, op. cif pp. 77,79, and unpublished Census tables 25 The per capita and per capitdper worker 6gures in each quintile in Chart 4 give average income figures for families currently ranked in those quintiles by the Census Bureau ba sed on aggregate household income. An alternative procedure would be to re-rbk all families after adjustment for fady size and number of workers.

Reranking based on per capitdper worker income would move some families out of the top quintile into lower qui ntiles and vice versa. Such reranking is beyond the scope of this study but would be possible using the computer capacity of the Census Bureau.The resulting income distribution figures for each quintile would be somewhat less equal than those shown in Cha rt 4, but still vastly more equal than conventional Census numbers.

Reranking also emphasizes the misleading nature of Census figures: many families presented as affluent in the current Census Bureau reports are far less so once adjustment for family size and number of workers are made 13 Chart 5 m'IbtJHolrrrhdd mlnoonrmFMlly Incslr mkr Married Couple Households With Children 1986 I I from the disintegration of the black family structure. Among both blacks and whites, families headed by single women have i n comes lower than the those of families headed by intact couples. The black illegitimate birth rate now equals 60 percent, and over 42 percent of black families are headed by single mothers. Among white families, only 13 percent are headed by single mother s apparent large income disparity between blacks and whites is greatly reduced.

For example, black married couples with children have post-tax incomes only 12 percent lower than the incomes of similar white families.='When differences in the number of pers ons per family are included, black married couple income falls to 75 percent of white married couple income. But this gap would be reduced if education and age level differences were taken into account.

Economic Inequality between Generations significant implications for the comparison of the economic conditions of young and old Americans. Conventional wisdom has it that the elderly are among the poorest, most economically vulnerable groups in America. Yet When these differences in family structure are ta k en into account, the Adjusting the Census Bureau statistics for family composition has similar 26 U.S. Bureau of the Census, op.cit pp. 41,43 14 I Chart 6 Income in All Households 1986 t I n I n m C 0 0 ource: U.S. Bureau of the Census, op. ut column 11 o n pp. 17,41,43 lnalysis of Census data strongly indicates otherwise (see chart 6). While the lost-tax median income of elderly households 17,875) is less than the nedian income of the average household 23,220 elderly households are maller than other househ olds. Adjustment for family size gives a better pic ure of the standard of living of the elderly: the post-tax per capita income of ilderly households is, in fact, higher than that for any other type ofhousehold n the U.S.

It k out that families with child ren, not the elderly, are the least affluent herican households.These families have a per capita post-tax income of 7,907 -well below the national average.n This amount is less than two hirds of the per capita post-tax income of the average elderly househ o ld.The ler capita income of families with children under age six is even lower; at 7,096 it is only 57 percent of the per capita income of the average elderly lousehold.28 Moreover, most elderly households already have paid off lousehold mortgages. Thus, t he living expenses of the elderly are relatively 27 hid, p. 29 28 hid p. 31 15 low in comparison to younger families with children who are trying to furnish homes, pay mortgages, and save for their childrens education IMPROVING THE CENSUS BUREAU MEASUREME N T OF INCOME DISTRIBUTION As this analysis has indicated, the impressions of family income and income equality provided by the Census Bureau differ markedly from reality. If the Census Bureau reports were obscure documents, used only by scholars who unders tood their shortcomings, that might not matter. But the Census Bureau data form the basis for media reports and for policy making in Congress Thus the deficiencies in the Census data often lead to deficiencies in public policy.

Good policy making requires that the Census Bureau measurement of inc.ome distribution be reformed in the following manner The recent inclusion by the Census Bureau, on an experimental basis of capital gains and employee health benefits, and the subtraction of federal and state inco me taxes and Social Security taxes, are welcome improvements.

These calculations should be made a permanent feature of the annual income statistics The full value of Medicare, means-tested cash aid, and means-tested noncash transfers and services should be incorporated in the Census Bureaus income data. The Congressional Research Service and other government sources have estimated that these benefits were worth 197 billion in 1986 Census Bureau procedures currently count a government benefit as having a ce rtain income value to affluent families, while the same benefit is counted as having less or even no income value to lower income families.

These procedures should .be reformed When comparing the income of American families, the Census Bureau should emphas ize comparisons of per capita household income and per capita household income per worker.The Bureau should downplay the more misleading figures of total household income CONCLUSION The Census Bureau has published reports on household income distribution f or over forty years.These reports are meant to measure the share of total economic resources allocated by the market and the government to various households. But the reports seriously underestimate the incomes of lower-income families because they omit n e arly $98 billion in government payments, noncash benefits, and services to low-income individuals. By undercounting the cost of government support to low-income individuals, the Census Bureau procedures exaggerate the level of income inequality in America 16 Correcting An Unflattering Image. Census Bureau data on income inequality also are misleading because they do not take into account variations in the number of workers per household. If income disparities attributable to differences in the number of wo r kers are discounted, the result is a relatively level distribution of income. Example: among families with at least one full-time worker, the post-tax per capita income of the highest quintile, adjusted for the number of workers per household, is only 1.8 times greater. than the average post-tax per capita income of the bottom quintile society with a high level of economic equality. While efforts to improve data over the past few years have been largely commendable, the current income distribution figures a re still inadequate and misleading. They present, an erroneous, unflattering image of U.S. society to policy makers and social commentators. Improvements in the Census Bureau income distribution data are long overdue and should be undertaken immediately M ore accurate income distribution data would show that the U.S. is a Robert Rector Policy Analyst Kate Walsh OBeirne Visiting Fellow 17