This PDF is a selection from an out-of-print volume from the National

Bureau of Economic Research

Volume Title: The Personal Distribution of Income and Wealth

Volume Author/Editor: James D. Smith, ed.

Volume Publisher: NBER

Volume ISBN: 0-870-14268-2

Volume URL: http://www.nber.org/books/smit75-1

Publication Date: 1975

Chapter Title: Introduction to "The Personal Distribution of Income

and Wealth"

Chapter Author: James D. Smith

Chapter URL: http://www.nber.org/chapters/c3748

Chapter pages in book: (p. 1 - 8)

 Introduction
James D. Smith

This volume contains papers presented at the annual meeting of

the Conference on Research in Income and Wealth held on the
campus of Pennsylvania State University in the fall of 1972.
The meeting and the one to follow it in 1974 reflect a renewed
interest in the distributions and determinants of income and
wealth. The papers in this volume focus on income and wealth
defined in traditional monetary terms. The 1974 follow-up
conference will extend the scope of inquiry to include non-
monetary dimensions.
In large measure, the resurgence of researcher interest in
personal distributions of income and wealth has resulted from
increased availability of microdata and the sustained method-
ological efforts by Orcutt and others demonstrating the superior-
ity of microdata approaches to the estimation of many socially
relevant or intellectually interesting models of behavior.
With respect to data, the decade of the sixties saw a rich harvest
of microdata, reflecting the desire of policymakers to estimate in
advance and measure in retrospect the consequences of social
programs. Early in the sixties, the Board of Governors of the
Federal Reserve System produced the Survey of Financial Charac-
teristics of Consumers and the Survey of Changes in Financial
Characteristics. By mid-decade, the Department of Labor was at
work on the National Longitudinal Surveys, and the Office of
Economic Opportunity had begun work on the Surveys of
Economic Opportunity (1966 and 1967). The Office of Economic
Opportunity, in conjunction with the Survey Research Center of
the University of Michigan, also began collecting data from an
ongoing panel of families in an effort known as the Panel Study of
Income Dynamics. Microdata from all of these studies were made
available to researchers. Near the end of the decade, Internal
Revenue Service (IRS) tapes of income tax returns (without names
or street addresses) became available to researchers who needed
them. By the beginning of the decade of the 1970s, the IRS had
made available microdata from estate tax returns, and the Census
Bureau was creeping toward the release of a standard version of

1
2 Introduction

the Current Population Survey which could be purchased by

researchers who were not among the superrich.
The release of public data by government agencies has not been
achieved easily, but the benefits, as evidenced by the papers in this
volume, justify the effort. The push to make microdata available
came from many researchers inside and outside of government.
Raymond Bowman in his many years at the Bureau of the Budget
contributed to this effort, as did Richard Ruggles, Guy Orcutt,
Joseph Pechman, Benjamin Okner, Robert Levin, Harold Watts,
and Ralph Nader, among others, outside of government.
Opposition to the release of microdata often has been argued in
terms of confidentiality. The principle of confidentiality has both
positive and negative aspects. We can easily agree that the
protection of individuals is often well served by priviledged
information, but when applied to the innocuous data of Census
and tax records after names and street addresses have been
removed, the argument rings false. Indeed, secrecy may harbor
worse evils than those it is intended to prevent. One of the serious
consequences of microdata privy only to the government agency
collecting it is that the situation offers a tremendous incentive for
"Watergating" errors which, understandably enough, occur in
large-scale data-collection efforts. The social benefits of accessible
microdata include an incentive to agencies to do first-rate work,
many eyes to uncover elusive, but often serious, errors which can
escape even very careful workers, and greater use of information
that is usually collected at great cost. In any event, it is clear that
research using microdata is on the ascendency, and that govern-
ment agencies and private researchers are finding ways to reap the
social benefits of microdata without injury to citizens who' provide
information.
The first chapter in this volume has very little to do with data
per se but, rather, is concerned with economic methodology in the
broadest sense—that is to say, with meta-economics. It is charac-
teristic Boulding, piercing through the fine polish of technique to
the important questions: What is it all about? What does it mean?
The author forces reflection:
Considering the enormous symbolic importance of the concept [equal-
ity], surprisingly little work has been done on it, philosophically,
theoretically, or empirically. Perhaps the reason for this is that it is too
painful, too contradictory, too confusing, and too important to be the
object of anything but rhetoric.
 Introduction 3

And what of the complement of equality: inequality? Is there a

meaningful quantitative interpretation of that? Bouldmg suggests
that we have named the measure rather than measured the named.
Gini coefficients, standard deviations, means of algebraic sums or
absolute differences, ranges—bah, humbug! It is all a myth. The
inequality of a distribution is a psychological phenomenon, a
measurable but unmeasured perception of beholders. He suggests
mental mapping may be the path to a better understanding. Social
psychology in the Sherifian tradition may or may not provide the
conceptual basis of quantitative measures, but, in any event,
reading the essay is delightful medicine for cerebral sclerosis.
Part II of the volume contains three papers on redistributive
mechanisms. The first paper (Chapter 2) by Robert Lampman,
deals with measuring the redistributive impact of tax and transfer
systems within the National Income Accounts framework. Starting
with the basic proposition that every economy has within it
mechanisms for redistributing its national product among con-
sumers, he asks how the redistributive process can be more
meaningfully monitored within the National Income Accounts
than it currently is. In answer, he suggests that the household
sector be subdivided to show insurance and pensions, philan-
thropic organizations, and families' subsectors, and that direct
interfamily transfers be identified. These changes would permit
the accounts to quantify the relative importance of transfers
between years and between nations.
He urges government to make such changes in the official
accounts, but suggests the profession should be prepared to move
forward on its own.
In Chapter 3, Ben. Okner imaginatively manipulates a synthetic
population (the MERGE File) created from the 1966 personal
income-tax model and the 1967 Survey of Economic Opportunity.
Using an assortment of income concepts, Okner addresses a
number of questions dealing with the distribution of tax burden a
la Musgrave ("The Distribution of Tax Burden by Income Group:
A Case Study for 1949," National Tax Journal, Vol. 4, March
1954), and with the combined redistributive impact of taxes and
transfers.
Looking only at the tax burden, his results are similar to those
of Musgrave, which appeared two decades ago: essentially, there is
little progressivity in federal, state, and local tax systems taken as
a group. Such progressivity as does exist is to be found far out in
4 Introduction

the right tail. Thus, the combined personal tax structure tends to
be proportional with respect to income and, consequently, does
not per se contribute significantly to income redistribution.
Transfers, on the other hand, significantly improve the income
position of persons in the lower tail of the pretransfer income
distribution, but they also contribute importantly to the upper
tail.
Using an income concept which comes close to being consump-
tion minus taxes plus changes in net worth, he found a 13 percent
reduction in the Gini coefficient due to the combined influence of
taxes and transfers.
In his early work on tax burdens, Musgrave had been forced to
work with aggregated data, which precluded estimation of the tax
burden for subpopulations. Using the microdata developed from
the tax and survey data mentioned above, Okner was able to
demonstrate that there are substantial differences in the effect of
tax-transfer systems on subpopulations. For instance, where he
found a 13 percent reduction in the Gini coefficient for all
families, a 30 percent reduction for families with a head age 65 or
over was discovered.
In Chapter 4, Harold Watts and Jon Peck use the MERGE File
to simulate the redistributive consequences of a number of
variants of a tax function combining a constant marginal rate and
a fixed credit. They compare the before and after
set of six "family" types. The family typology, obviously selected
for its relevance to the problem of welfare reform, consisted of:

1. aged individuals;
2. families with aged heads;
3. families with female heads;
4. nonaged individuals;
5. families with nonaged male heads and 2 to 5 members; and
6. families with nonaged male heads and 6 or more members.

The authors examine the before and after tax, income status of
these units, and more interestingly, the before and after tax
distribution of the welfare ratio (income/poverty threshold).
They conclude that the extant tax structure is differentially
beneficial to the aged and to 'female-headed units, and that
 Introduction S

working poor families headed by males are significantly overtaxed

or undertransferred. This is a finding of both economic and
political significance.
Part III deals with accounting periods of lengths other than the
traditional one-year period. It is noteworthy that two of the
chapters in this part use large microdata sets assembled outside the
federal government, and that the third uses microdata produced
cooperatively by the government and Herbert Parnes.
Martin David and Roger Miller, relying on a decade of
professional investment in the development of a file of Wisconsin
tax returns which links families between years and individuals to
their parental families, examine the importance of capital gains on
the size distribution of income. They find little difference in long
versus shorter accounting periods, which can be explained by the
inclusion of capital gains. The reason adduced is that capital gains
tend to be a recurring form of income for those who receive them.
Kohen, Parnes, and Shea (Chapter 6) examine income instabil-
ity among two groups of men: those aged 14 to 24 in 1966 and a
group aged 45 to 59 in the same year. The microdata used by
them was collected in a joint effort by Parnes, the Department of
Labor, and the Bureau of the Census and has become known as
the National Longitudinal Surveys. They develop a measure of
relative income instability (RIC) which is intended to measure the
relative instability of an individual's income change vis-á-vis the
mean change in income of his cohort.
They found that the instability of income rank was greater for
blacks and for younger men generally than for whites and men in
the 45 to 59 year age group.
Among whites, earnings were more stable than total family
income. While the same relation was found among the older black
group, the reverse was true for young blacks.
As has been found in other studies, a regression toward the
mean was found when changes in income rank were compared to
initial rank.
Jacob Benus and James Morgan use three sets of panel data
collected by the Survey Research Center to study the influence of
the unit of analysis, the accounting period, and the income
concept on the size distribution of income. The first of these data
sets was produced in a study of the response of consumers to the
1964 tax cut. Families were interviewed quarterly in that study
and thus provided a basis for looking at short-term changes in
6 Introduction

income. The second set of data were collected from about 1 ,400
respondents in four interviews spaced approximately a year apart.
The third data set was the Panel Study of Income Dynamics, a
panel study now in its fifth year and still going on. The authors
find that the unit of analysis and the concept of income have
greater influence on the size distribution than does the length of
the accounting period. In terms of the influence of other factors
on income stability, occupation, age, and race—in that order—were
found to be important.
There were no formal discussants in the session from which
these papers came, but each participant was asked to provide,
within the time allotted to him, such observations upon the papers
of the other participants as he felt would be useful. Martin David
provided a set of comments reflecting on his own joint effort with
Roger Miller and on the other two papers presented in this part.
David's comments are included as Chapter 8.
The three chapters of Part IV are concerned with the
distribution of wealth. Lee Soltow presents estimates of the
distribution of wealth, income, and social class of men in large
northern cities in 1860 (Chapter 9); A. B. Atkinson writes of the
distribution of wealth in England in 1968; and James Smith
presents a study of the distribution of wealth in Washington, D.C.,
in 1967.
The Soltow piece starts with the microdata of the 1860 Census
and incorporates it into a model which permits one to predict the
probability of escape from a state of propertylessness—a state
which was not as uncommon as one might suspect. According to
Soltow's data, slightly over half of the adult males were without
assets; and according to his model, the probability of remaining in
a propertyless condition in a given year was .96.
Soltow also presents a model which employs a Pareto-
rectangular income distribution and a conventional savings func-
tion to determine the 1860 wealth distribution.
Atkinson (Chapter 10) provides the latest available statistics of
the distribution of wealth in Britain. He uses the estate multiplier
method of estimating the number of wealth holders and their asset
holdings in 1968.
His paper is particularly valuable not simply because it extends
a long series of wealth estimates for Britain which have been made
using the estate multiplier method, but because it examines the
sensitivity of the estimates to a methodology which has been
 Introduction 7

increasingly employed in the United States and in other parts of

the world.
Atkinson presents evidence that the concentration of British
wealth is not decreasing: the conventional wisdom that the richest
1 percent own one-third, and the richest 5 percent one-half, of
the total wealth is more nearly correct than some recently
published estimates which suggest that there has been a significant
decrease in inequality of wealth in Britain. A discussion of
Atkinson's paper by Kathleen Langley, herself one of the pioneers
in British wealth estimates, is included at the end of Chapter 10.
Using the same technique as Atkinson, but concerning himself
with Washington, D.C., Smith (Chapter 11) estimates that the
total wealth of all residents of the District of Columbia was $5.5
billion in 1967. However, his focus is not upon the aggregate level
of wealth or its size distribution among all residents, but upon the
differences in wealth holdings of whites and blacks. He found
blacks to have an average net worth of $1,000, while whites
averaged $19,000. He also found that 96 percent of the black
population had a net worth of under $5,000, about the same
figure that Soltow's work suggests for the population as a whole in
1860.
A discussion of Smith's paper by Vito Natrella is included at the
end of Chapter 11.
In Part V, two papers which served as the focal point of a
session on the quality of data on income and wealth are presented.
One of the papers deals with a retooled model of the former OBE
income distribution series, which for many years presented
family-income size distributions aligned to personal income
aggregates in the National Income Accounts (NIA), but which was
subjected to a major overhaul in 1 962. The other paper is a test of
the quality of transfer income data reported in the Current
Population Reports. The two papers are related by the fact that
they both have the Current Population Survey (CPS) as a focal
point; the latter paper directly and the former indirectly, because
the CPS is the basic source of income information in the
redesigned OBE series.
In Chapter 1 2, Dorothy Projector and Judith Bretz present the
results of tests of the reliability of three types of transfer reported
in the CPS: (1) Social Security and railroad retirement; (2) public
assistance; and (3) unemployment and workmen's compensation.
They compared the incidence of reported incomes of each type of
8 Introduction

the CPS against a simulated incidence based upon the character-

istics of families and individuals in the CPS file. They also
compared the CPS reported receipts of transfers to administrative
tallies of the numbers of persons receiving such transfers and their
value.
Projector and Bretz found the characteristics of CPS respon-
dents reporting transfer incomes to be consistent—or at least not
inconsistent—with eligibility criteria established for specific pro-
grams. They concluded, using personal income in the National
Income Accounts as a standard, that all types of transfer payments
were understated in the Current Population Survey. Comparing
program aggregates, the CPS was found to understate the value of
transfer payments by about 8 percent. The understatement was
found to be greater for younger age groups than for those over 65.
In Chapter 13, Edward Budd and Daniel Radner provide a
guided tour of the statistical labyrinth they negotiate in producing
the BEA income size-distribution series. The BEA series is
intended to produce income size-distributions which are concep-
tually consistent with personal income in the National Income
Accounts and such that the sum of the distributed incomes is
equal to the personal income aggregate in the NIA.
The old OBE (now BEA) income size-distribution series was
discontinued in 1962 because benchmark data used to relate tax
and survey data had become suspect, and because the art of
income size-distribution estimating was trailing far behind the
science of data processing.
Budd and Radner have spent several years incorporating
microdata techniques into the BEA series. They have been
hampered by lack of access to direct matches of tax return and
survey data, but they have made imaginative use of statistical
matching procedures to overcome this impediment.