Almost all the benefits of economic growth since [the 1970s] have gone
to a small number of people at the very top.
Financial Times, Jan. 29, 2008
Since the mid-1970s, however, income growth has been confined almost
entirely to top earners.
—Robert H. Frank, New York Times, March 9, 2008
The modern American economy distributes the fruits of its growth to a
relatively narrow slice of the population.
—David Leonhardt, New York Times, April 9, 2008
(Related article: Has Middle America Stagnated? The Region, Sept. 2007)
The economic progress of middle-income households over the past
generation is difficult to assess. Many recent reports portray
stagnation—household incomes increased little, wages increased even less
and rising expenses drove families into debt. In contrast, another set
of reports describe large gains—income per person almost doubled, people
are healthier and living longer, and the quality, quantity and variety
of goods and services being consumed are greater than ever. It seems
that life for middle America is stagnating at the same time it’s getting
This article is the second in a Region series that seeks to reconcile
the apparent conflict between statistics indicating stagnation in
standards of living and statistics indicating robust growth. The issue
addressed here is whether income growth over the past three decades
bypassed middle America and accrued almost entirely to the rich. I find
that—contrary to many reports—middle America did quite well.
A theme of this series is that much of the apparent contradiction can be
resolved by taking a closer look at the statistics being reported. A
better understanding of the underlying data provides a more accurate
assessment of the economic progress of middle America, which, in turn,
is a critical input in the formulation of sound public policy.
The first article in the series (September 2007 Region) focused on
hourly wages and compensation (wages plus benefits). The central
findings were that hourly compensation increased more than hourly wages
because nonwage benefits grew rapidly and that the price index used to
adjust for inflation has a substantial impact on growth rates. The
analysis showed that inflation-adjusted hourly compensation of middle
Americans rose by almost 30 percent from 1975 to 2005.
But hourly compensation is just one part of the story. This article
takes a closer look at annual household income growth over the past 30
years. Income here is from all sources, not just labor earnings. A
forthcoming article will examine other factors related to standards of
living—including consumption, debt and income volatility.
A quick note before proceeding. The recent slowdown in the growth of the
national economy has been extremely difficult for many individuals.
Unfortunately, such downturns are an inevitable part of a dynamic market
economy. Thirty years ago, the economy suffered a series of recessions
that lasted through the early 1980s; the economy was characterized by
widespread unemployment and high inflation.
While keeping in mind the hardships endured in such periods, I hope in
this series to address a broader, longer-term question: Is the average
standard of living for middle Americans still rising? More specifically,
are cyclical job losses and income fluctuations occurring around a
higher, or lower, average standard of living compared with a generation ago?
Let’s take a look at the data on income growth and preview the main
Income gains mostly to the rich?
According to the U.S. Census Bureau, median household income adjusted
for inflation increased a scant 18 percent over the past 30 years
Chart 1).1 In contrast, data from the Bureau of Economic Analysis (BEA)
indicate that income per person was up 80 percent, almost doubling (see
Chart 2). A widely reported explanation for these statistics is that the
rich reaped most of the benefits of economic growth over this period,
while middle-income households gained little. Findings on rising
inequality are consistent with this view.
These statistics appear quite compelling, but hiding in the background
are some key issues that might alter the story. Average household size
declined substantially during the past 30 years, so household income is
being spread across fewer people. The mix of household types—married
versus single, young versus old—also changed considerably, so the
“median household” in 2006 looks quite different from the “median
household” in 1976. Finally, the measure of income used by the Census
Bureau to compute household income excludes some rapidly growing sources
The goal of this article is to examine the income gains made by middle
American households over the past 30 years. The analysis requires a
careful look at data, with the payoff being a more comprehensive picture
of income gains by middle-income people. The analysis will also clarify
how the modest growth in median household income is consistent with the
large increase in income per person.
The main finding is that—after adjusting the Census Bureau data for
three key factors—inflation-adjusted median household income for most
household types increased by roughly 44 percent to 62 percent from 1976
to 2006. The only household types with substantially lower growth were
“working-age male householder without spouse present” and “male
householder with children but without spouse,” but these types
constitute just 10 percent of all households. Household income
inequality increased notably over this period; nonetheless, middle
American households had substantial income gains.
Here is a preview of the key data issues that lead to the higher
estimates of median household income growth.
The price index used by the Census Bureau overstates inflation, and
thus understates income gains, relative to a preferred price index.
A changing mix of household types leads the overall median increase
to understate the median increase of most household types.
The Census Bureau measure of household income understates income
growth by excluding some rapidly growing sources of income.
The remaining difference between the 44 percent to 62 percent increase
in median household incomes and the 80 percent increase in BEA personal
income per person appears to be largely attributable to an increase in
income inequality. The findings in this article are consistent with
recent research showing that the largest income increases occurred at
the top end of the income distribution. However, the findings here are
not consistent with the view that the incomes of middle American
households stagnated over the past 30 years. Income for most middle
American households increased substantially.
The first step in most analyses of income growth is to adjust the income
data for inflation. This procedure is meant to account for the declining
purchasing power of the dollar over time. The phrase “adjusted for
inflation” suggests that economists have agreed on one adjustment
procedure. In fact, there are several measures of inflation, and the
selection can have a substantial impact on reported growth in
inflation-adjusted variables. (See a more
complete discussion below.)
To measure economic progress over time, a central issue is adjusting the
data for changes in the overall price level—that is, inflation. The goal
in measuring inflation is to answer the question, How much more income
do people need with today’s (this year’s) prices to be just as well off
as they were with yesterday’s (last year’s) prices? This is referred to
as the change in cost of living (COL).
Calculating a single inflation number that summarizes the change in the
prices of hundreds of thousands of items—one Wal-Mart store has over
40,000 differently priced items—is extremely difficult. In addition,
inflation measures must also take into account a continuous stream of
new products and quality improvements in existing products, differences
in prices across stores and across days of the week, consumer
substitution away from products with rising prices and a host of other
A widely used measure of inflation uses the consumer price index for
urban consumers published by the Bureau of Labor Statistics. Extensive
research has concluded that the CPI overstates inflation, even after
recent changes to improve its accuracy. However, no government agency
publishes a price index that attempts to correct for all the identified
shortcomings in the CPI. So economists have turned to other measures of
inflation that address some, though not all, of the problems. The
personal consumption expenditures deflator published by the Bureau of
Economic Analysis is perhaps the alternative most widely used by
economists, but other measures exist.
Economist Michael Boskin summarizes the research on CPI inflation in
recent research, and he provides a rough estimate of how much the CPI
has overstated inflation since the early 1970s.* His work thus provides
a rough estimate of how large inflation-adjusted growth would be under
“correctly measured” inflation.
The accompanying table lists the growth rate in median household income
as reported by the U.S. Census Bureau using competing measures of
inflation. The Price Index columns report income increases using four
available measures of inflation. The last column uses Boskin’s estimate
of correctly measured inflation.
Two results stand out. First, Boskin’s estimate of correctly measured
inflation results in substantially higher income growth than any of the
standard measures of inflation. If this estimate is anywhere near
correct, inflation-adjusted growth rates of income, wages and so on are
substantially higher than has been reported.
Second, income growth is strikingly different even across the currently
published measures of inflation. Studies of economic progress that use
CPI inflation will report notably lower inflation-adjusted growth rates.
Readers should look carefully at what price index is being used when
they see the generic phrase “inflation-adjusted.”
—Terry J. Fitzgerald
Since there is no one “correct” measure of inflation, a choice must be
made on which price index is most appropriate. Throughout this analysis,
I use the price index for personal consumption expenditures (PCE
deflator) for all inflation adjustments. This index uses the basket of
goods and services that people consume each year, and it avoids some of
the shortcomings of the consumer price indexes produced by the Bureau of
Labor Statistics.2 The PCE deflator is widely used by the Federal
Reserve and macroeconomists. Using the PCE deflator, inflation-adjusted
median household income rose by 26 percent from 1976 to 2006, compared
to 18 percent using the CPI-based measure of inflation used by the
Key point: Using the PCE deflator to adjust for inflation adds 8
percentage points to median household income growth, raising the median
increase to 26 percent.
The next step in the analysis—and it is a much bigger step—is to look at
the impact of changes in the mix of households. Using households as a
unit of measure (income per household) creates difficulties. Household
size has changed over time; the average number of people per household
declined from 1976 to 2006, so household income is being spread over
fewer people.3 The mix of household types (for example, married versus
single) changed substantially. Finally, there is the odd statistical
fact that almost 60 percent of people live in households with
above-median income, and this percentage changes over time. Interpreting
household statistics presents challenges.
I address these difficulties by examining how median household income
has changed for specific types of households. This allows separation of
the income gains made by each type from the impact of the changing
Breakdown by household types
A basic characteristic of a household is whether or not a married couple
is present, and, if not, whether the “householder” is male or female.4 Here households are divided into four types: “married couple,” “female
householder with no spouse present,” “male householder with no spouse
present” and “all other.”5 Each of these household types includes family
members living in the household; so, for example, a female householder
with no spouse present need not be a single female living alone.
Chart 3 shows how the mix of household types changed from 1976 to 2006.
The fraction of married-couple households declined from almost
two-thirds of all households in 1976 to one-half of households in 2006,
while the fraction of male and female householders without spouses
Dividing households into these basic types leads to a surprising result:
Each household type has considerably higher median income growth than
the overall household median growth of 26 percent. Chart 4
shows that married-couple households—the largest type—had a median
income gain of 42 percent, while female householders with no spouse
present—the second largest type—had a striking 56 percent gain in
At first blush, this result seems like a mathematical contradiction: How
can all subgroups grow faster than the entire group? But there is no
contradiction. The explanation lies in the changing household mix.
Married-couple households have much higher incomes than other household
types, and there has been a large decline in married-couple households.
This decline depresses overall median income growth. As an extreme but
illustrative example, consider what would happen if one-half of all
married couples were to divorce next year. Median household income would
plummet as each higher-income married-couple household is dissolved into
two lower-income households—the same income is spread across more
households. This would be true even if wages increased substantially for
all workers, so that household types had large income gains. (See the
table below for detailed results used throughout this article.)
A further breakdown of household types
The three major household types—married-couple, female householder
without spouse and male householder without spouse—still encompass a
wide range of households. Slicing these three categories each into four
subtypes—all households with children under 18 years of age, young
householders (ages 15–29) without children, working-age householders
(ages 30–59) without children and retirement-age householders (ages 60
and over) without children—reveals other interesting results. Chart 5
shows median household income gains by these subtypes, as well as for
all households and the “all other” household type.
Median household income increased by at least 36 percent for most
subtypes. Only for male householders with children and for working-age
male householders without children did income grow by substantially
less. The low growth rate for these two subtypes, which comprise 10
percent of all 2006 households, is consistent with the well-established
finding that average male wages increased little over this period. Young
male and female householders without children had median income growth
of 30 percent and 33 percent, respectively. The relatively narrow bars
for these household types in Chart 5 show that they represent a small
slice of all households. In fact, household types with at least 36
percent median household income gains comprise over 85 percent of all
Retirement-age households had the largest median income gains, ranging
from 47 percent for married couples to 74 percent for male householders
with no spouse present. All subtypes of female householders had sizable
increases, ranging from 33 percent to 54 percent. Female householders
and retirement-age subtypes started with relatively low household
incomes, so these high growth rates somewhat diminish the large income
differences across subtypes.
Outside the median
The results discussed for household types up to this point have been for
the median households only. Here I look at how households above and
below the median fared. Chart 6 shows income growth for married-couple
households at the 25th, 50th (median) and 75th percentiles of the income
distribution. Within each household type, the 75th percentile of
households (higher-income) had larger income gains than the 25th
percentile (lower-income). This reflects an increase in income
inequality that has been widely documented. The results are
qualitatively similar for the other household types.
But even with the increase in inequality, income gains for a broad set
of middle-income households of most types were substantial. Incomes of
the middle 50 percent of households—between the 25th and 75th
percentiles—increased by at least 22 percent and as much as 59 percent
for most household types, with gains exceeding 30 percent for most
households. Retirement-age male householders had much larger gains,
while working-age male householders and male householders with children
had much smaller increases.
Key points: The change in household mix had a major impact on reported
median household growth. While overall median household income grew by
only 26 percent, the median gains for most household types ranged from
36 percent to 54 percent. Inequality increased notably within household
types. Still, gains for most middle-income households ranged from 30
percent to 60 percent.
Adding missing income
The final step in calculating the income gains made by middle American
households requires a closer look at differing definitions of “income.”
The Census Bureau uses a narrow definition of income in its report on
median household income that focuses on money income and excludes
nonmonetary sources of income. The BEA, in contrast, defines personal
income as income received from all sources. Examples of income excluded
by the Census Bureau, but included by the BEA, are employer
contributions to employee pension and insurance funds and in-kind
transfer payments such as Medicaid, food stamps and energy assistance.
These sources of income contribute to economic well-being and should be
included in the definition of income.7
Unfortunately, BEA data on personal income are not available for
individual households, and I am left to use Census data with their
narrow definition of income. Does the difference in definitions matter
for measuring the growth in median household income? Most likely, yes.
Chart 7 presents data on income per person (average, not median) using
BEA data on personal income and Census data on money income.8 As
expected, Census income is noticeably smaller than personal income—30
percent smaller in 2007. More important, Census income grew by 15
percentage points less over the 30-year period.9 This reflects the fact
that Census income excludes some rapidly growing nonmonetary income,
such as health insurance benefits paid by employers. As a result, the
income gains for middle Americans reported thus far are likely understated.
Providing an accurate estimate of how much larger household income
growth would be using the broader BEA definition of income is beyond the
scope of this article. However, a back-of-the-envelope calculation is
The median household income gains of 36 percent to 54 percent tend to be
well over half of the 65 percent increase in Census income per person. A
conservative estimate is that median household income growth is one-half
of income per person growth. Applying that ratio to the additional 15
percentage point growth in personal income would add about 8 percentage
points to median household income growth.10
Using this estimate for the missing income, median household income for
most household types rose 44 percent to 62 percent. Gains for most of
the middle 50 percent of households in each type ranged from 35 percent
to 65 percent.
Key points: A rough estimate is that 8 percentage points would be added
to median household income growth if the nonmonetary income excluded by
the Census but included by the BEA could be assigned to households. This
raises the range of median gains for most household types to 44 percent
to 62 percent.
Reconciling with BEA personal income gains
No further adjustments to median household income remain to be made. Yet
the 44 percent to 62 percent range for median household income growth is
substantially below the 80 percent increase in personal income per
person. Why? While the accounting is not nailed down precisely, here are
Some of the difference is due to a decline in the average number of
children per household in households with children. A striking example
is female householders with children, which had a 43 percent gain in
median household income but a 65 percent median gain on a per person
basis.11 Married couples with children and male householders with
children also had a decline in the average number of children and a
larger (5–10 percentage points) gain in income per person.12
But much of the remaining difference between the reported median
household income and the larger gain in personal income appears to be
attributable to the increase in income inequality. When income rises
faster for the richest households, median income grows by less than
average income. Bill Gates’ rising income over this period certainly
increased the average income in Medina, Wash., but it had no effect on
Chart 7 shows that Census income per person rose by 65 percent, and I
calculate that median Census income per person rose by 50 percent. This
suggests that increasing inequality may account for roughly 15–20
percentage points of the difference between the growth in median
household income and the growth in BEA personal income per person. While
this is a notable amount, the analysis in this article does not support
the claim that only the rich have benefited from the economic growth of
the past 30 years.
The claim that the standard of living of middle Americans has stagnated
over the past generation is common. An accompanying assertion is that
virtually all income growth over the past three decades bypassed middle
America and accrued almost entirely to the rich.
The findings reported here—and summarized in Chart 8—refute those
claims. Careful analysis shows that the incomes of most types of middle
American households have increased substantially over the past three
decades. These results are consistent with recent research showing that
the largest income increases occurred at the top end of the income
distribution. But the outsized gains of the rich do not mean that middle
Why does the debate about middle America matter? Because an accurate
assessment of the economic progress of middle America is a crucial input
in formulating good public policy. Claims of long-term middle America
stagnation—such as those quoted at the beginning of this article—are
often part of a broader argument about the adverse impact of
globalization, outsourcing and free trade. And middle class stagnation
is used as motivation for a specific set of policies. But if middle
America has not stagnated—as this analysis has shown—then this
motivation for those policies is without merit.
Furthermore, if it is understood that middle America has indeed
experienced substantial gains, policy priorities may change. For
example, more emphasis might be placed on policies that promote
continued economic growth or that target deeply rooted poverty rather
than middle class stagnation. But regardless of the specific policy,
policymakers and the public should base their decisions on an accurate
assessment of how the economy has impacted and continues to impact
1 Median household income is calculated using the U.S. Census Bureau’s
recently revised CPI-based price index series. This revision lowered
inflation-adjusted median household income growth during the 1976–2006
period from 21 percent to 18 percent.
2 For a discussion of the biases in the consumer price index, see
Michael J. Boskin, 2005, “Causes and Consequences of Bias in the
Consumer Price Index as a Measure of the Cost of Living,” Atlantic
Economic Journal 33:1–13.
3 Average household size fell from 2.86 persons in 1976 to 2.56 persons
in 2006. See Historical Income Tables—Households, U.S. Census Bureau.
4 Current Population Survey definitions: A household consists of all the
people who occupy a housing unit. The count of households excludes group
quarters. The householder refers to the person (or one of the people) in
whose name the housing unit is owned or rented (maintained) or, if there
is no such person, any adult member, excluding roomers, boarders or paid
employees. If the house is owned or rented jointly by a married couple,
the householder may be either the husband or the wife.
5 To maintain additional uniformity across households of each type, all
household members are required to be from one family (related by birth,
marriage or adoption), and at most one married couple may be present.
About 5 percent of households do not satisfy these restrictions and are
assigned to a generic “all other” category.
6 Data for the Annual Social and Economic Supplement of the Current
Population Survey are used throughout this study and were obtained
through IPUMS databases maintained by the Minnesota Population Center:
cps.ipums.org/cps. Miriam King, Steven Ruggles, Trent Alexander, Donna
Leicach and Matthew Sobek, Integrated Public Use Microdata Series,
Current Population Survey: Version 2.0 (machine-readable database).
Minneapolis, Minn.: Minnesota Population Center (producer and
7 The differences between Census income and personal income are more
subtle than indicated here. For an excellent conceptual and quantitative
discussion of these differences, see John Ruser, Adrienne Pilot and
Charles Nelson, November 2004, manuscript, Alternative Measures of
Household Income: BEA Personal Income, CPS Money Income, and Beyond.
8Census income includes data for people who live in group quarters, who
contributed 1.3 percent of total Census income in 1976 and 2006. The
growth rate of income from 1976 to 2006 based only on household income
is the same.
9 The BEA’s measure of national income per person deflated using the PCE
deflator increased 74 percent over this period. That is 6 percentage
points less than personal income growth, and 9 percentage points greater
than Census income growth.
10 I also did a rough calculation of the impact of adding one excluded
source of income: health benefits provided by employers. In this
exercise I used data from the national income and product accounts to
estimate the ratio of employer-paid health benefits to wages, and
multiplied household wages in the CPS ASEC survey by this ratio. The
result of this accounting exercise is that the median income gains of
young and working-age households and households with children across all
types increased by an additional 5 to 10 percentage points.
Retirement-age household types gained less—1 to 4 percentage
points—since fewer are currently working. Census income plus NIPA health
benefits still grew by 7 percentage points less than personal income;
adding other sources of excluded income would also impact growth rates.
These results roughly support the 8 percentage point addition used in
11 Per person income is defined as follows: Income for each individual
is total household income for his or her household divided by the number
of household members.
12 The average number of children per household for female householders
with no spouse fell by 0.36 (2.15 in 1976 versus 1.79 in 2006). For the
middle 20 percent of these households, the decline was a more dramatic
0.72 children (2.51 versus 1.78). The drop in the average number of
children was notable but less striking for married couples and male
householders. The average number of children in the third quintile for
married couples fell by 0.24 (2.10 in 1976 versus 1.86 in 2006), and for
male householders with children by 0.32 (1.82 in 1976 versus 1.51 in 2006).
Are Income Gains Due to More Wives Working Longer Hours?
Has Middle America Stagnated? The Region, Sept. 2007)