[Economic Report of the President (1997)]
[Administration of William J. Clinton]
[Online through the Government Printing Office, www.gpo.gov]

[DOCID: f:erp_c5._]
Economic Report of the President - - - - - - - - - - - - H. Doc. 105-002
[From the online service of the U.S. Government Printing Office]
[wais.access.gpo.gov]


CHAPTER 5--Inequality and Economic Rewards

IT WAS OVER 30 YEARS AGO that President John F. Kennedy said, ``A
rising tide lifts all the boats.'' The decade preceding his Presidency
and the decade thereafter supported this optimism. Tremendous economic
growth raised the incomes of American families at all levels, including
the poor, and income inequality fell dramatically. Beginning in the late
1970s, however, this broad tide of equalizing growth turned, and
inequality began to increase. The gap between rich and poor continued to
widen through the 1980s and into the early 1990s, regardless of economic
conditions. In the last few years some signs have begun to emerge that
inequality may be stabilizing and perhaps even declining slightly, but
the gap in economic rewards between rich and poor is still much larger
than it was 20 years ago.
Economic inequality has several different dimensions. We begin by
looking at trends in earnings inequality across and among workers as
grouped by age, sex, and level of education. Earnings inequality is an
important indicator in its own right, because it helps characterize the
structure of the labor market. It is also an important contributor to
inequality in household incomes, a broader measure of economic well-
being that aggregates the resources of all household members and
incorporates other income flows besides earnings. Finally, we consider
some alternative measures of inequality that may better address
differences in lifetime income across households.
Concerns with inequality are inseparable from concerns about the
well-being of the poor, but a rise in inequality does not necessarily
mean the poor are worse off. A rise in inequality is consistent with a
scenario in which the circumstances of the poorest are improving, but
the richest are experiencing even greater gains. Such a state of affairs
is less troubling than one in which those at the top prosper while the
living standards of those at the bottom stagnate or decline. It makes a
profound difference to our understanding and to our policies which of
these depictions of rising inequality is the correct one. Therefore, in
addition to documenting trends in inequality, this chapter will focus
specifically on the well-being of those at the bottom of the
distribution.

RECENT TRENDS IN INEQUALITY

Before addressing longer term trends in inequality, we briefly
explore the record of the recent past. Although it is too soon to tell
whether a break in the long-term trend toward greater income inequality
has occurred, income statistics over the past few years do show some
reduction. From 1993 to 1995, income gains were observed throughout the
income distribution, but the percentage increases were the largest for
low-income households. One way to view these changes is to separate
households into five equal groups based on their income (called
quintiles) and estimate the increase in income received by each
quintile. Chart 5-1 displays the results of such an analysis for the
1993-95 period. It shows that this period has seen gains for each
quintile, which were largest for the lowest quintile and smallest for
the highest.


The ``rising tide'' theory might have predicted such results, given
the ongoing economic expansion. Yet recent historical experience
indicates that expansions do not always reduce inequality. Consider, for
example, three years--1979, 1987, and 1995--when economic performance
was similar: in all three years gross domestic product (GDP) grew by
about 2 to 3 percent, the unemployment rate was about 6 percent, and the
economy had been expanding for a few consecutive years. Yet whereas the
percentage of the population living in poverty (i.e., the poverty rate)
fell by 0.7 percentage point in 1995, it actually rose by 0.3 percentage
point in 1979 and fell by only 0.2 percentage point in 1987. The Gini
index of household income inequality (which ranges from 0, indicating
perfect equality across income quintiles, to 1, which would indicate
that all income is going to the top quintile) rose in both 1979 and
1987, but fell in 1995. Recent data show that inequality has been
reduced beyond what would have been predicted by cyclical factors.
Although these results are encouraging, it is too soon to tell
whether the longer term trends of increasing inequality have been
reversed. The remainder of this chapter focuses on these longer term
trends.

EARNINGS INEQUALITY

The incomes of most people consist mainly of earnings from labor. A
large component of income differentials across households can be
attributed to differences in the earnings of individuals. An examination
of earnings is also facilitated by the individual nature of the measure:
it is not necessary to adjust for the changes in household composition
that so complicate discussions of household income. This section
documents trends in earnings inequality in general, trends across
workers with different characteristics, and trends across workers with
similar characteristics, before attempting to identify the factors that
can help explain the observed rise in inequality over time.

DOCUMENTING TRENDS IN EARNINGS INEQUALITY

Because earnings are a function of both the wage rate and the number
of hours worked, we concentrate here on full-time, year-round workers so
as to abstract from any biases due to changes in working hours over
time. Men's earnings are the focus of this analysis, because the
increasing labor force participation of women over time may have altered
the composition of the female workforce in ways that might distort the
results. For instance, if women with higher earnings potential have
entered the labor market at a faster rate in recent years, measured
inequality would appear to have increased, even if the underlying
distribution of wages for women continuously employed has remained
unchanged. After examining earnings inequality among men, we briefly
examine trends among women.
For male workers we examine two ratios that compare earnings between
workers at different points in the earnings distribution. One of these
is the ratio of the earnings of a male worker at the 90th percentile
(i.e., one whose wages exceed those of 90 percent of all male workers)
to those of a male worker at the 50th percentile (i.e., the median male
worker). This ratio is called the 90/50 earnings ratio. The other ratio,
called the 50/10 earnings ratio, is that between the median worker and a
worker who earns more than only 10 percent of workers. Estimating both
these ratios is more useful than the common alternative of estimating
the 90/10 ratio alone, because the 50/10 ratio provides more information
on the well-being of those at the bottom of the distribution. Because
the median male worker's wages have fallen somewhat in real terms, an
increase in the earnings ratio between the 50th and the 10th percentiles
indicates a larger reduction among those with low earnings. In 1995,
annual earnings at the 10th, 50th, and 90th percentiles were $12,920,
$31,497, and $70,314. (Note that the 90th-percentile figure is well
below the huge salaries paid to top corporate executives; see Box 5-1.)
Trends in the 90/50 and the 50/10 earnings ratios for full-time,
year-round male workers are shown in Chart 5-2. These data reveal that
the male worker at the middle of the earnings distribution earned about
2.4 times the wages of the worker at the 10th percentile in 1995,
compared with 2.2 times in 1979. The 90/50 earnings ratio rose by a
similar amount, from about 1.9 in 1979 to 2.2 in 1995. The overall trend
in both ratios is upward over most of this period, indicating increasing
inequality across the wage spectrum.
Another way to document increasing wage inequality is to calculate
the percentages of the workforce falling in each of several different
earnings categories at different points in time. Chart 5-3 shows that a
larger proportion of workers earned less than $15,000 in 1995 than in
1979 (when earnings are measured in constant 1995 dollars); at the other
end of the distribution, a larger share of the workforce earned in
excess of $75,000 in 1995 than in 1979. (The consumer price index, or
CPI, is used in both calculations to adjust for inflation; potential
biases introduced by using this index are described in Chapter 2.) These
increases at the top and bottom of the distribution are offset by a
reduction in the share of workers earning between $35,000 and $75,000.

BETWEEN-GROUP INEQUALITY

The trend in inequality may be better understood by first grouping
workers according to certain key characteristics (educational attainment
and age are two that are commonly used) and then separating observed
wage differentials into two components: the differential observed
between workers so grouped (between-group inequality) and the
differential observed among workers in the same group (within-group
inequality). Taking first the education dimen-

Box 5-1.--Executive Compensation

One much-publicized aspect of earnings inequality has been the
extraordinarily high level of compensation of top corporate executives.
In 1995 the average compensation package for chief executive officers
(CEOs) in a sample of 362 of the largest 500 U.S. firms was $1.5
million, and some CEOs received more than $10 million that year.
Defenders of current corporate pay scales argue that today's executive
compensation packages, with their moderate base pay and generous stock
options, encourage high-level management to act in the shareholders'
interests by providing greater rewards for good long-run performance.
Critics respond that it is unclear in practice how much executive
compensation is even designed to be ``performance-based.'' For example,
compensation in the form of stock options rewards executives for share
price increases even when these are attributable to market-wide price
gains rather than the executives' own actions. In addition, such
compensation practices may have adverse effects on worker morale, when,
for instance, a firm pays its top management very high salaries at the
same time that it is laying off workers.
However this debate is resolved, the effect of high executive
compensation on measured earnings inequality throughout the economy is
minimal, because top executives represent only a tiny fraction of the
workforce. As we saw in Chart 5-2, earnings disparities have been
growing even when measured by the 90/50 earnings ratio. The executives
whose compensation is the subject of this controversy receive a level of
earnings that places them well above the 90th percentile, and therefore
even a doubling of their salaries would have no impact on trends in this
measure. And executive earnings obviously have no influence at all on
the 50/10 ratio, which has been increasing as well.
sion, Chart 5-4 shows the trend in the ratio of the earnings of the
median male college graduate to that of the median male high school
graduate. The chart reveals that returns to education grew tremendously
during the 1980s and early 1990s. In 1980 the median male college
graduate earned roughly one-third more than the median male high school
graduate, but this wage premium grew to over 70 percent by 1993. Since
then that trend has slowed, and the ratio even declined slightly in
1995.
Experience on the job is another important dimension in studying
inequality. The premium paid to more experienced workers has also been
increasing over the past two decades or so. As shown in


Chart 5-5, the median 45- to 54-year-old male worker earned roughly 50
percent more than the median 25- to 34-year-old worker in 1995, compared
with a difference of less than 20 percent in 1979. The main reason for
the increase is that young workers were paid less in 1995, not that
older workers were paid more.

WITHIN-GROUP INEQUALITY

Within-group inequality is also on the rise and in fact accounts for
about two-thirds of the total increase in earnings inequality. For
instance, among male high school graduates both the 90/50 and the 50/10
earnings ratios have risen since about 1970 (Table 5-1). Although the
upward trend in the 50/10 ratio apparently stopped in the late 1980s,
that of the 90/50 ratio continues. Similar findings emerge for groupings
of workers by age. Table 5-1 also shows the 90/50 and 50/10 ratios for
25- to 34-year-old full-time, year-round male workers. Within this
group, the 90/50 ratio increased from about 1.6 to about 1.9 between
1979 and 1995.

EARNINGS INEQUALITY AMONG WOMEN

Women have experienced increases in earnings inequality similar to
those of men. The 90/50 and 50/10 ratios of earnings for women working
full-time, year-round began rising in the late 1970s and have continued
upward through the 1980s and 1990s, as have


those for men. The results presented in Charts 5-2 through 5-5 and Table
5-1 for men, with respect to overall, between-group, and within-group
inequality, generally find parallels in the patterns for women. For
instance, the wage premium received by college-educated women roughly
doubled between 1978 and 1995, from 38 percent to 70 percent.

EXPLANATIONS FOR INCREASING EARNINGS INEQUALITY

Alternative explanations for the observed increase in earnings
inequality can be categorized into three broad groups: supply-side
factors, demand-side factors, and institutional factors. (A provocative
alternative hypothesis is presented in Box 5-2.) Although no clear
consensus has emerged regarding the relative strength of these
alternatives, demand-side explanations play a large role.
A simple model of the labor market for more skilled, usually higher
paid workers and for relatively low paid, less skilled workers will help
clarify the role of supply- and demand-side factors. Supply-side factors
can increase inequality if they cause the supply curve in the market for
less skilled workers to shift outward by relatively more than the supply
curve in the market for more skilled workers. As shown in Chart 5-6,
such shifts would lead wages to fall by a greater amount in the less
skilled labor market than in

Table 5-1.-- Earnings Ratios for Male High School Graduates               and 25- to 34-Year-Old Male Full-Time,
Year-Round Workers
----------------------------------------------------------------------------------------------------------------
Male high school  graduates       25- to 34-year-old male
--------------------------------             workers
Year                                                       -------------------------------
90/50 ratio     50/10 ratio     90/50 ratio     50/10 ratio
----------------------------------------------------------------------------------------------------------------
1967............................................            1.62            1.89            1.64            2.03
1968............................................            1.57            1.86            1.58            2.01
1969............................................            1.61            1.76            1.64            1.84

1970............................................            1.61            1.80            1.65            1.81
1971............................................            1.64            1.85            1.65            1.85
1972............................................            1.62            1.91            1.65            1.91
1973............................................            1.66            1.90            1.65            1.92
1974............................................            1.68            1.94            1.66            1.97

1975............................................            1.62            1.89            1.64            1.82
1976............................................            1.59            1.89            1.65            1.83
1977............................................            1.62            1.99            1.63            1.98
1978............................................            1.61            2.02            1.64            1.94
1979............................................            1.60            1.98            1.65            1.94

1980............................................            1.62            2.00            1.68            1.88
1981............................................            1.63            2.02            1.68            1.98
1982............................................            1.69            2.08            1.72            1.98
1983............................................            1.69            2.12            1.73            2.03
1984............................................            1.69            2.13            1.72            2.10

1985............................................            1.74            2.16            1.81            2.09
1986............................................            1.73            2.22            1.83            2.08
1987............................................            1.71            2.21            1.81            2.12
1988............................................            1.71            2.17            1.84            2.15
1989............................................            1.79            2.18            1.87            2.11

1990............................................            1.79            2.15            1.83            2.16
1991............................................            1.77            2.19            1.84            2.17
1992............................................            1.78            2.19            1.91            2.16
1993............................................            1.87            2.11            1.96            2.13
1994............................................            1.88            2.17            1.96            2.14

1995............................................            1.83            2.16            1.93            2.16
----------------------------------------------------------------------------------------------------------------
Source: Council of Economic Advisers tabulations of the March Current Population Survey.

the more skilled labor market, increasing inequality. What might cause
such an asymmetry? The increasing numbers of immigrants in the labor
market, and the increasing labor force participation rates of women, who
tend to have less work experience, could have led to a disproportionate
supply shift in the market for less skilled workers.
In analogous fashion, demand-side factors could have influenced the
relative wages of more and less skilled workers if they caused the
demand curve in the market for more skilled workers to shift outward by
more than that in the market for less skilled workers, or (especially)
if the demand curve in the latter shifted inward. As shown in Chart 5-7,
these changes would increase wages in the more skilled labor market and
reduce them in the less skilled labor market, increasing inequality.
Technological developments favoring skilled workers (called skill-biased
technological change) could have led to such shifts. The integration of
new production technologies may have increased firms' demand for workers
capable of using these technologies. Evidence indicates, for instance,
that workers

Box 5-2.--Earnings Inequality and the Winner-Take-All Society

One provocative hypothesis offered to explain part of the increase
in within-group inequality is the expansion of ``winner-take-all''
markets, where top performers reap far greater rewards than do others
whose ability is only slightly inferior. For example, it is not uncommon
to see a star professional athlete making millions of dollars a year
while another, only slightly less talented athlete earns far less. It
has been argued that markets such as these have become more pervasive in
the American economy, with the result that ours is increasingly a
winner-take-all society.
Huge wage premiums for small differences in performance may now be
observed in law, medicine, investment banking, academics, and other
professions. Windfalls to the top producers in these fields have become
increasingly common as computing and telecommunications technology have
advanced, facilitating the flow of information, and as transportation
costs have been reduced, increasing mobility. These factors increase
competition to hire the best performers, increasing their wages. How
large a share of the observed increase in earnings inequality may be
attributed to the expansion of winner-take-all markets remains unknown.
who use a computer on the job earn significantly more than those who do
not.
The expansion of international trade could also have produced the
hypothesized shifts in demand curves. Because import industries tend to
employ relatively less skilled workers, it is argued that the wages of
less skilled American workers are coming under pressure either from
direct job loss or from more intense wage bargaining with their own
employers, who are now forced to compete internationally. Of course, the
demand and supply shifts just described may occur simultaneously,
compounding the effect on earnings inequality.
Within this framework, demand shifts appear to play the larger role
in explaining growing inequality. Trends in the returns to education
provide perhaps the most accessible evidence of the influence of demand
shifts, if the assumption is valid that more education translates into
higher levels of skill. The returns to a college education rose
throughout the 1980s, as noted earlier, even though the college
enrollment rate among recent high school graduates grew dramatically
over this period. If relative demand for more and less skilled workers
had remained constant, the greater supply of college-educated workers
should have led to a decline in the college wage premium. The fact that
the college wage premium instead


rose sharply suggests that demand shifts must have more than outweighed
any concurrent supply shock. This framework is useful in explaining
within-group inequality as well, because skill differentials will remain
within broad demographic categories.
Evidence shows that skill-biased technological change is probably
the main contributor to these demand shifts (many experts support this
view; see Box 5-3). Some evidence suggests that international trade may
be responsible for only a relatively small share of the increase in
inequality. For example, even manufacturing firms whose products face
little foreign competition have reduced their demand for less skilled
workers. Nevertheless, direct evidence of the importance of skill-biased
technological change in explaining trends in within-group inequality is
difficult to come by. Some studies avoid this difficulty by treating
technological change as a residual, attributing rising inequality to
this factor when their findings have excluded all other likely
candidates.
A final set of explanations suggests that changes in institutional
arrangements in the labor market, such as the declining influence of
unions and a reduction in the real value of the minimum wage, have led
to lower returns for workers in the lower tail of the earnings
distribution. Unions have long provided wage premiums to such workers.
But the share of employed workers belonging to unions has eroded from a
peak of roughly 30 percent through much of the 1950s and 1960s to about
15 percent in 1995. Although research indicates that the decline of
unions may indeed have played some role in increased earnings
inequality, it probably can explain only a small share of the increase.
This finding is consistent with the fact that inequality also increased
among groups of workers, such as college graduates, who are unlikely to
belong to unions.
The eroding value of the minimum wage also could contribute to
earnings inequality. A minimum wage truncates the earnings distribution
at its lower end. If more than 10 percent of workers receive the minimum
wage, inequality on such measures as the 50/10 earnings ratio will be
less than it would be otherwise. Inequality on this measure could even
be reduced if the fraction receiving the minimum were less than 10
percent, if ``ripple effects'' exist whereby workers who would otherwise
earn slightly over the minimum instead receive higher wages because of
greater competition for their labor. The decline in the real value of
the minimum wage through the 1980s is similar in its timing to that of
the increase in inequality. It is unlikely to be a leading explanation
of rising inequality, however, because inequality also increased within
groups of workers, such as older workers, who are unlikely to be
affected by the minimum wage.

Box 5-3.--The Experts' Consensus on Earnings Inequality

Possible explanations for the observed rise in earnings inequality
during the 1980s and early 1990s include skill-biased technological
change, trade liberalization, demographic shifts, declining
unionization, and rising immigration. Although the relative importance
of each of these is difficult to determine precisely, some leading
economists generally agree as to which are the main culprits.
Participants at a recent colloquium on this topic at the Federal Reserve
Bank of New York--a group that included many prominent labor
economists--viewed technological change as the strongest contributor.


INCOME INEQUALITY

Household income is a broader measure of economic well-being than
individual earnings, because it aggregates the incomes of all household
members and incorporates other flows of income besides earnings.
Although labor earnings are typically its largest component, household
income also includes interest and dividend receipts, cash transfer
receipts, and rental payments. Household size and composition are
clearly important factors in determining observed household income. In
this section we document the increase in inequality since the late 1970s
and explore its possible causes.

DOCUMENTING THE INCREASE IN INCOME INEQUALITY

One way to trace changes in income inequality is to separate
households into income quintiles and estimate the share of income
received by each quintile. (Box 5-4 discusses some problems in income
measurement.) Increasing inequality would be manifested by a fall in the
share of income going to the lowest quintile and a corresponding rise in
the share going to the highest quintile. Chart 5-8 shows just such a
pattern in household income quintiles for 1979 and 1995: since 1979 the
shares going to the bottom four quintiles have declined, while the share
going to the highest quintile has increased.
Changes in income shares over time may mask how well those at the
bottom of the income distribution are doing. For instance, if the
richest quintile is getting richer but the incomes of all other
quintiles are holding constant, the shares of total income received by
the lower quintiles would fall, misleadingly suggesting that they are
becoming worse off. An alternative approach to documenting changes in
the distribution of income, one that examines levels of income for those
in different segments of the distribution, may prove beneficial.
Chart 5-9 displays this sort of information for 1979 and 1995.
Households are divided into four categories: those with incomes less
than $15,000, those between $15,000 and $35,000 (roughly the median in
1995), those between $35,000 and $75,000, and those over $75,000.
Incomes are converted into 1995 dollars using the CPI. The chart shows
that the share of households in the highest income bracket increased
from 10.9 percent to 14.8 percent between 1979 and 1995, while the share
in the lowest income bracket remained unchanged. These statistics
suggest that some middle-income households have moved up into the higher
income categories, but the number of households toward the bottom of the
income distribution has remained nearly constant.
This approach may be misleading, however, because the unit of
analysis is the household, not the individual. Because household
composition has been changing over time, the observation of an unchanged
number of households lying below a particular income cutoff may overlook
the reality that more people are residing in these households.
One way to focus more directly on the well-being of individuals near
the bottom of the income distribution is to examine trends in the
poverty rate. Throughout the 1960s and early 1970s the poverty rate fell
dramatically, from 22.2 percent in 1960 to 11.1 percent in 1973. (Chart
5-10 shows the trend since 1967.) It remained low throughout the 1970s,
ranging from 11.1 percent to 12.6 percent over the decade. In the 1980s
the poverty rate rose dramati-

Box 5-4.--Shortcomings of Household Income Measures

Household income is a useful indicator of economic well-being
because it is relatively easy to measure and interpret. It has its
shortcomings, however. For instance, it does not incorporate taxes or
payments made in kind, such as food stamps and housing subsidies. To the
extent that the tax system is progressive and that in-kind transfers are
means-tested, use of an after-tax-and-transfer definition of income
would reduce the measured level of inequality. Although some analysts
have adapted the standard income-based measures to include the value of
in-kind income, economists have not agreed on the best method for doing
so. Some value in-kind benefits according to the cost of providing them,
and others according to what an individual would be willing to pay to
receive the benefit. In any case, research incorporating taxes and in-
kind payments shows trends in inequality that are similar to those
reported by standard measures.
Another problem is that differences in household size will lead to
different assessments of the economic well-being of individuals with the
same household income. Attempts to abstract from differences in
household size have proceeded by developing ``equivalence scales'' that
adjust household income for the number of household members. Other
approaches scale the incomes of larger households by progressively
smaller amounts for each additional member. Even after making these
adjustments for differences in household size, however, income
inequality appears to be increasing.
Despite these obstacles, alternative measures of income are being
tested by the Bureau of the Census, and others have been proposed by the
National Academy of Sciences (NAS). The Census Bureau produces a series
of 17 experimental estimates of income in an attempt to gauge the
effects of various noncash government benefits and taxes on income
levels and on poverty. The NAS proposes another definition of income to
be used in the measurement of poverty that adds noncash benefits to
money income and subtracts taxes, some work expenses, some child care
expenses, child support payments, and medical out-of-pocket expenses. It
would also adjust the equivalence scale currently used in poverty
calculations. Measures such as the Census experimental series and those
proposed by the NAS are intended to reflect the effects of government
policy initiatives. Nevertheless, no clear consensus exists regarding
certain complex methodological issues, including valuation of some
benefits such as medical and child care.


cally and has fallen below 13 percent only once since then, in 1989
following 6 years of economic expansion.
The composition of the impoverished population has also changed over
time, especially with respect to age. The percentage of children living
in poverty rose from 14.4 percent in 1973 to 22.7 percent in 1993, but
has fallen somewhat since then. On the other hand, the poverty rate for
those over 65 used to be considerably higher than that for the
population as a whole (24.6 percent compared with 12.6 percent in 1970),
but mainly because of the Social Security system, poverty among this
group has actually fallen below the overall poverty rate since 1982. The
elderly poverty rate reached an all-time low of 10.5 percent in 1995,
falling significantly below that for the 18- to 64-year-old population
for the first time ever.


The transition from a poverty population that is disproportionally
elderly to one that is more heavily weighted toward households with
children suggests that the household size of the low-income population
has increased over time. This is consistent with the coexistence of a
rising share of low-income individuals and a constant share of low-
income households. The effect of changes in household composition on
income inequality is explored more fully below.

EXPLANATIONS FOR INCREASING INCOME INEQUALITY

Because measurements of income inequality incorporate all sources of
a household's income, including labor market earnings, it should come as
no surprise that a major contributor to increasing income inequality
across households is rising earnings inequality across workers. In fact,
about half of the increased inequality in household incomes over the
1980s can be explained by trends in earnings inequality among men.
Part of the remaining share can be attributed to changes in
household composition and, in particular, to the increase in female-
headed households. The share of family households headed by women has
risen rapidly, from just over 10 percent in 1970 to about 18 percent in
1995. These households are more likely to receive lower incomes because
they lack a second wage earner, because women earn less on average than
men, and because some of these women do not work at all. Therefore the
growing share of this type of household has worsened income inequality.
In fact, the rise in the percentage of children in poverty over the past
25 years is strictly due to the increase in the number of children
residing in female-headed households, whose poverty rates are higher
than those for children living in other circumstances. The poverty rate
among children in female-headed households has actually decreased over
time.
Research suggests that the rapid rise in female labor force
participation has also contributed to growing inequality. This finding
is not obvious, however, because in some ways a rise in the number of
working women serves to reduce inequality. For instance, the
distribution of women's earnings is more compressed than that of men, so
that increasing female labor force participation should reduce overall
earnings inequality. If all men and women lived alone, this reduction in
earnings inequality and the reduction in the number of people with zero
earnings (because of increased employment) would also reduce income
inequality.
Inequality may nonetheless increase in response to greater female
labor force participation because people tend to marry persons whose
earnings potential is similar to their own. For example, more educated
men may be more likely to marry more educated women. The increase in
employment among married women could therefore increase household
inequality in one of two ways. First, if women in high-income households
are joining the labor force in greater numbers than women from low-
income households, their earnings will push their household incomes even
further beyond the middle of the distribution, and income inequality
will increase. This hypothesis is not supported by the data, however, as
labor force participation rates for women have risen roughly equally
across households ranked by the husband's earnings level. Second, for
working couples, rising earnings inequality will be compounded at the
household level if men with high earnings are married to women with high
earnings. Taken as a whole, the evidence suggests that women's
increasing labor force participation has contributed somewhat to growing
income inequality during the 1980s.
Income inequality can also be affected by changes in unearned income
across households. The source of the unearned income determines whether
or not it increases or decreases the income inequality that would occur
from earnings alone. For example, property income is more likely to be
received by individuals with higher earnings, and therefore an increase
in property income would tend to worsen inequality. Transfer payments
are more likely to go to individuals with lower earnings, and an
increase in transfers would therefore tend to reduce inequality.
Research suggests that, on balance, nonlabor income tended to increase
inequality during the 1980s. The effect of these changes is still
significantly less than that caused by growing earnings inequality,
however.

ALTERNATIVE MEASURES OF INEQUALITY

This discussion, like much of the economic literature on inequality,
has focused on inequality in annual earnings and household income.
However, appropriate borrowing and saving behavior can smooth year-to-
year fluctuations in income, making consumption less variable, provided
households have appropriate access to credit markets. Therefore
differences in lifetime income across households may offer a more
valuable perspective on differences in well-being.
Of course, one cannot reliably measure lifetime income when much of
that income has yet to be received. Lifetime income is thus an
inherently unmeasurable concept, and analysts must resort to using
related measures as a basis for estimating it. One such measure is
consumption, on the theory that households set consumption levels
according to their own assessments of their lifetime income. A potential
problem here is that a household may have large asset holdings,
indicating the potential to raise its consumption in the future, but
choose to limit its consumption for the present. Therefore, another
indicator used to examine lifetime income inequality across households
is household wealth.
Another way to address differences in lifetime income across
households is to examine income mobility--the extent to which households
move across the income distribution over time. Increasing annual income
inequality is more meaningful as an indicator of lifetime income
differences across households if income mobility does not increase as
well.

CONSUMPTION INEQUALITY

If consumption decisions are based on households' assessments of
their lifetime income, then inequality in consumption can be used as a
proxy for inequality in lifetime income. For example, a middle-income
household that suffers a brief spell of reduced income will not change
its consumption habits much, whereas a household with regularly low
income will consume considerably less. Therefore we can expect to see
less inequality in consumption than in annual income.
Some evidence supports this proposition: studies have found that the
distribution of consumption is more concentrated than that of income. In
other words, individuals do appear to prefer to smooth their consumption
levels across their lifetimes through borrowing and saving. One
difficulty in comparing the distributions of income and of consumption
is that income is measured before taxes and in-kind transfers, whereas
consumption is based on after-tax income and includes in-kind transfers.
To the extent that taxes and in-kind transfers reduce inequality (an
issue that is discussed below), one would expect consumption inequality
to be less than income inequality. During the 1980s, consumption
inequality rose along with income inequality, but in the early 1990s the
two diverged. Between 1989 and 1993, consumption inequality leveled off
while income inequality continued to rise. Some demographic groups,
particularly households headed by a high school graduate or dropout,
experienced large declines in consumption inequality over the period. No
obvious explanation for the timing of the turnaround in consumption
inequality or its comparison to income inequality exists.

WEALTH INEQUALITY

Another shortcoming in using annual income as a measure of
differences in economic well-being is that it does not capture the
purchasing power of a household's asset holdings. Therefore differences
across households in terms of net wealth (which consists of cash
savings, financial assets, and the value of physical assets such as a
house or a car, less any outstanding debt) provide an alternative
indicator of inequality.
Data on wealth are limited, but one source, the Survey of Consumer
Finances, sponsored by the Federal Reserve Board does provide comparable
data for 1983, 1989, and 1992. Over these years median family net wealth
(estimated at $52,000 in 1992) has been fairly stable. Wealth is
concentrated in the hands of a small number of families, and the degree
of that concentration has remained fairly constant. The wealthiest 10
percent of families have owned roughly 67 percent of total net wealth
since the early 1980s. The top 1 percent of families did increase their
wealth holdings from around 30 percent of total net wealth in 1983 to 37
percent in 1989, but their share fell back to 30 percent by 1992. The
stock market boom of the 1980s might have led one to predict increasing
concentration, but stock ownership has become more widespread over time.
In addition, home values increased over the period, and home ownership
is far more common than stock ownership.

MOBILITY

If a household's income varies widely from year to year, annual
measures of inequality may provide a very inaccurate picture of lifetime
inequality. If the increase in annually measured income inequality over
the past 20 years or so has been accompanied by an increase in income
fluctuations, it is possible that lifetime incomes have been unaffected.
For instance, if new labor market entrants make less than previous
entrants, but their wages grow more rapidly as they gain experience,
then annual measures of income inequality will be greater, as will
income mobility, but lifetime income may be unchanged. Therefore the
degree of mobility through the income distribution is another means of
examining the difference between annual and lifetime income. (A related
issue of mobility between parents and children is explored in Box 5-5.)
Studies of mobility have compared household incomes over varying
periods, such as 1 year, 5 years, and 10 years. One-year changes in
income are likely to reflect short-term changes, such as temporary job
loss, as well as measurement errors in reported income that are not
perfectly correlated between years. Longer term changes will also
incorporate these events but are more likely to identify more permanent
changes in incomes, which are particularly large among younger
households. Therefore one might expect mobility over longer periods to
be greater than that from year to year.
A standard approach in estimating income mobility is to rank
households by their income in each of two years, separate them into
quintiles in each year according to their rank, and then see to what
extent households have moved from quintile to quintile between the two
years. Results from these studies indicate a reasonably high degree of
mobility over time. One study finds that about 3 out of every 10
households move between quintiles from one year to the next. As one
would expect, mobility is greater over longer periods: almost two-thirds
of households change income quintiles over 10 years. These mobility
rates do not appear to be increasing over time. The probabilities of
making a transition between income quintiles over periods of varying
lengths have remained roughly

Box 5-5.--Intergenerational Mobility

Another issue relating to income mobility is the extent to which
income is transferred between parents and children. If the correlation
between parents' income and their children's income as adults is high,
then a child is likely to experience a level of economic well-being
similar to that of his or her parents (i.e., intergenerational mobility
will be low). If the two generations' incomes are not correlated,
children will have no greater probability of ending up in one income
quintile than in another. Early studies found a low correlation:
intergenerational mobility was quite high. The son of a high-income
father would have the same probability as anyone else of residing at any
given point in the income distribution after only two generations.
An important problem with these studies, however, is that they
ignored measurement error in reported income. If reported levels of
income of either the father, the son, or both were inaccurate, the
resulting estimates of the correlation in income would be biased toward
zero. More recent studies have paid careful attention to the measurement
error issue and the bias it introduces. These studies found a
considerably higher correlation and thus a considerably smaller degree
of intergenerational mobility than did previous work. Their results
indicate that it would take four generations before the son of a high-
income father had a roughly equal probability of residing at any point
in the income distribution.
steady through the 1970s and 1980s. The evidence thus does not appear to
support the proposition that rising income inequality has been offset by
increasing income mobility.
One issue in interpreting these studies is that transitions over
time between income quintiles may occur because of changes in the flow
of income (mainly earnings) or changes in household composition. A
person who marries is likely to experience a significant increase in
household income if his or her spouse works, even if that person's
earnings remain constant.
An alternative approach that some researchers have taken in
examining mobility is to focus exclusively on individuals' earnings and
transitions that occur between earnings quintiles over time. Again,
mobility rates are reasonably high, with higher transition rates over
longer time periods. Roughly 3 in 10 individuals change earnings
quintiles between one year and the next, and almost half make such a
transition over 5 years, according to one study. As


with income mobility, no trend over time is apparent in earnings
mobility (Chart 5-11).

GOVERNMENT POLICY AND INEQUALITY

Without government intervention, the distribution of income would be
even more dispersed than it is. A progressive Federal income tax and a
variety of Federal and State transfer programs have for decades worked
to reduce inequality. More recently, several new policies have been put
in place to reduce inequality further, particularly by improving the
conditions of those toward the bottom of the income distribution.

ASSESSING THE IMPACT OF GOVERNMENT POLICY

Incorporating the effect of tax and in-kind transfer policies into
income measures poses two challenges. First, a household's tax burden
and the value of noncash benefits such as food stamps and Medicaid need
to be calculated, and this calculation is subject to ambiguities (Box 5-
4). Second, calculating income in the absence of government as
conventionally measured income less transfers assumes that the
availability of the transfers has no impact on recipients' other income.
Still, after taxes and transfers have been taken into account to the
extent possible, government policy is shown to reduce inequality
significantly. The progressivity of the Federal individual income tax
system, together with all payroll taxes and State income taxes, reduces
the Gini index by about 5 percent. Transfer payments account for an even
larger reduction in the Gini index, of around 20 percent. The program
that contributes perhaps the most to reducing inequality is Social
Security, as one might expect from the relatively low poverty rate among
older Americans.
The incidence of poverty is similarly affected by government
policies. The officially reported poverty rate of 13.8 percent in 1995
would have been 21.9 percent if cash transfers were not included in
income. Moreover, when incomes are measured according to the most
comprehensive measure, which includes all taxes and the earned income
tax credit (EITC) as well as the valuation of in-kind transfers, the
poverty rate is estimated to be only 10.3 percent.

ADDITIONAL POLICIES TO REDUCE INEQUALITY

Both short-run and long-run policies are needed to help reduce
income inequality. In the short run, the EITC can help raise the incomes
of workers with low earnings. The EITC is a refundable tax credit of up
to 40 percent of earnings, depending on family size. The credit was
expanded in both 1990 and 1993, with both an increase in its value and a
broadening of the covered population to include very low wage workers
who do not reside with children. The number of families receiving the
credit rose from 12.6 million in 1990 to an estimated 18 million in
1996. Between 1990 and 1996 the average credit per family more than
doubled, from $601 to an estimated $1,400. In 1995 almost 3.3 million
people were lifted out of poverty by the EITC, more than twice as many
as only a few years before.
The recent increase in the minimum wage may also play a part in
reducing inequality. Between 1981 and April 1990, the minimum wage
remained constant at $3.35 per hour even as inflation eroded its value
by 44 percent. The 27 percent increase in the minimum wage in April
1990, to $4.25 an hour, did not restore it to its real 1981 level.
Inflation then eroded the value of the minimum wage another 23 percent
up to October 1996, when it was increased to $4.75; that increase is to
be followed in September 1997 with a further increase to $5.15.
Although even these raises will not restore the purchasing power of
the minimum wage to its 1981 level, the minimum wage and the EITC
together do more to reduce inequality today than they did then. For
example, a single parent with two children earning $5.15 per hour for 40
hours per week, 50 weeks per year in 1998 would make $9,775 (in 1996
dollars) before the EITC and $13,343 including the EITC. Without the
1996-97 minimum wage increases, this family's income including the EITC
would have been only $11,294. The combination of the recent rise in the
minimum wage with the expansion of the EITC makes the returns to work
for minimum wage workers greater than in 1981, when the minimum wage was
higher in real terms. In that year the same family, with the parent
working the same hours but earning the minimum wage of $3.35, would have
received $11,336 (in 1996 dollars) before the EITC and $12,034 including
the less generous EITC available at that time.
In the long run, greater access to education and training programs
should reduce inequality by reducing the wage premium associated with
additional training. In terms of the simple labor market model presented
above, as more workers obtain additional education, the supply of more
highly skilled workers shifts outward and that of less skilled workers
shifts inward (again assuming that more education translates into higher
levels of marketable skill). These shifts increase the wages of the less
skilled relative to those of the more skilled, reducing inequality
between the two groups.
Improved access to education and training can also reduce inequality
if it allows individuals from lower income households to make
investments in their human capital that they could not make otherwise.
Programs such as Head Start can provide disadvantaged preschoolers the
opportunity to begin formal schooling with the intellectual tools they
need to flourish. The recently inaugurated Federal direct student loan
program has also provided benefits to students and schools. The Federal
Government now issues loans directly to students through the financial
aid offices of their colleges, rather than through commercial financial
intermediaries, and offers four different repayment options, including
an income-contingent payment plan. In the 1996/97 academic year, 1.9
million students will have participated in the program, which is widely
viewed as successful in providing more timely, flexible, and accessible
service to both students and universities.

CONCLUSION

Income inequality in the United States has risen over the past two
decades. Its very persistence means that this trend will be difficult to
change. Even recognizing the reversal when it does occur will be
difficult enough, because statistical analysis cannot easily distinguish
a decisive turnaround in inequality from a relatively brief pause in its
rise. It is still too soon to tell whether the promising statistics
reported in the past few years represent a true reversal or just such a
pause.
Because of this uncertainty, continued vigilance is required to find
ways to help alleviate inequality, particularly to the extent that it
can reduce hardship for those at the bottom of the economic ladder. Some
changes have already been instituted, such as the increase in the
minimum wage and the expansion of the EITC. Improved access to education
and training is also essential. Although these represent useful first
steps, much remains to be done.