Income Inequality

This page intentionally left blank

 Income
Inequality
Why It Matters and Why Most
Economists Didn’t Notice

M A T T H E W P. D R E N N A N

New Haven & London

 Copyright © 2015 by Matthew P. Drennan.
All rights reserved.
This book may not be reproduced, in whole or in part, including illustrations,
in any form (beyond that copying permitted by Sections 107 and 108 of the
U.S. Copyright Law and except by reviewers for the public press),
without written permission from the publishers.

Yale University Press books may be purchased in quantity for educational,

business, or promotional use. For information, please e-mail
[email protected] (U.S. office) or [email protected] (U.K. office).

Set in Minion type by Integrated Publishing Solutions.

Printed in the United States of America.

Library of Congress Control Number: 2015940164

ISBN 978-0-300-20958-7 (cloth : alk. paper)

A catalogue record for this book is available from the British Library.

This paper meets the requirements of ANSI/NISO Z39.48-1992

(Permanence of Paper).

10 9 8 7 6 5 4 3 2 1
 To my wife, Katherine Van Wezel Stone,
My children, Matthew, Maureen, and Erica,
And my grandchildren, Grace and Ava Drennan
This page intentionally left blank
 Contents

Acknowledgments ix

Chapter I Introduction 1
Chapter II Trends in Income Distribution 8
Chapter III Possible Causes of Rising Income Inequality 20
Chapter IV Consumers’ Shift to Debt 35
Appendix Panel Regression Analysis of
State and National Data 63
Chapter V Consumption Theory and Its Critics 73
Chapter VI Has This Happened Before? 111
Chapter VII Conclusion 129

Notes 133
Bibliography 139
Index 149
This page intentionally left blank
 Acknowledgments

This work has been more than four years in the making. It
began as an insight in early 2009, in the wake of the initial
shock of the financial crash, when I realized that the dramati-
cally rising income inequality of the past three decades might
have played a role. The more I explored the data and the de-
bates about causes of the crash and ensuing Great Recession,
the more convinced I became that the conventional economic
explanations were missing a critical piece of the puzzle. I re-
alized then that it was necessary not only to put income in­
equality back into the story but also to explain why that part of
the story had not been told already. That is, I wanted to under-
stand why economists had failed to see the significance of the
most important economic trend of the past three decades—
the dramatic rise in inequality.
Many institutions and people have assisted in this work.
The Russell Sage Foundation generously financed the first
phase of this study. In addition, Cornell University’s Podell
Emeriti Award for Research and Scholarship provided subse-
quent funding for research expenses. From the beginning of
this project, the Luskin School’s Department of Urban Plan-
ning at the University of California, Los Angeles, provided
the library resources, an office, IT assistance, and most im-
x Acknowledgments

portantly, colleagues, which are so necessary for academic re-

search. For the semester that I spent in New York City, I was
generously provided with an office at the New York University
Schack Institute of Real Estate by the dean, Rosemary Scanlon,
and a professor there, Hugh Kelly.
UCLA’s statistical consulting group at the Institute for
Digital Research and Education gave me invaluable assistance
at every step. I sent them countless email queries that they an-
swered in a day or less—right answers, too. I made many trips
to their walk-in consulting sessions, trips that were always
worthwhile.
I had the good fortune of having a number of astute, sup-
portive, yet critical readers. I particularly want to thank Alan
Altshuler, Charles Brecher, Robert Hockett, Raymond Horton,
Morton Horwitz, Christopher Jencks, and David Rigby for
pushing me to sharpen my arguments and sharing with me im-
portant literature. I also want to thank participants in the Brown
Bag Lunch speaker series at the New York Federal Reserve Bank,
where I made an invited presentation on this project in the fall of
2012. Among the participants, Erica Groshen, Andrew Haugh-
wout, and James Orr offered cogent remarks, and they pointed
me to important data sources. Two anonymous referees de-
serve thanks for suggestions that markedly improved this
manuscript.
I have had the great benefit of the services of several tal-
ented graduate students at UCLA. The spare simplicity and
clarity of the tables and figures I attribute to two excellent re-
search assistants at UCLA: Anne Brown and Taner Osman.
They both know that the sole purpose of tables and figures is
to elucidate the argument, not to drown it in obscurity. Of the
nineteen tables, Anne produced sixteen of them, and I did the
other three (Tables 4.4, 4.8, and 4.11). Of the thirteen figures,
Acknowledgments xi

Anne produced seven (Figures 2.1, 2.2, 4.1, 4.4, 6.1, 6.2, and
6.4) and Taner produced five (4.3, 5.1, 5.2, 5.3, and 6.3). One
figure, 4.2, is taken from an International Monetary Fund
(IMF) paper with permission. Anne Brown has been a critical
assistant in the final preparation of the manuscript to Yale’s
exacting standards. Mike Manville, a former Ph.D. student in
urban planning at UCLA (now an assistant professor at Cor-
nell), read every word of various drafts in the early stage of
producing this book. He made both substantive and technical
suggestions that I mostly accepted.
I wish to acknowledge my editors at Yale University
Press, William Frucht and Jaya Chatterjee, who have made
this book production process exciting rather than tedious. My
copy editor, Joyce Ippolito, and my production editor, Ann-
Marie Imbornoni, purged the manuscript and proofs of nu-
merous flaws I had overlooked, and they did that with speed
and grace. Their standards of excellence and respectful treat-
ment would flatter any author.
Most of all I thank my wife, Katherine Stone, who en-
couraged me in pursuing this project from the beginning and
discussed the ideas in depth. She also read every word more
than once, and her suggestions have enhanced the final manu-
script. She has been an invaluable asset for my project, always
eager to help and always giving excellent suggestions.
This page intentionally left blank
 I
Introduction

his book tells two stories. The first tells how rising
income inequality over the past decades led to rising,
indeed surging, household debt to support consump-
tion, a surge that brought on the financial crisis and
Great Recession of 2008–09. The second shows that main-
stream economists have adhered to a theory of consumption
that assigns no role to the distribution of income, and there-
fore is inadequate for fully understanding the Great Recession
or preventing the next one.
Of course rising income inequality is only one of the
causes of the surge in household debt, but it is an important
one that is too often neglected by economists and policy mak-
ers. The period from about 1995 to 2007, especially post 2000,
can be characterized as a perfect firestorm of household in-
debtedness, fueled by four factors: (1) stagnant incomes for
most households related to the long-term rise in income in-
equality; (2) unusually low interest rates after 2000; (3) legal
and institutional changes that relaxed borrowing standards of
lenders, raised the availability of credit, and made housing a
more liquid asset; and (4) the housing price bubble. The burst-
2 Introduction

ing of that bubble in 2006–07 precipitated the financial crisis

and the Great Recession, but it was only the last straw. The
debt-supported expansion of consumption became unsustain-
able after 2007. Because consumers have begun to reduce their
debt—deleveraging—and increase their saving, consumption
will be depressed for some years, producing an anemic recovery.
Most analyses of the financial crash and Great Recession
identify factors (2) through (4) as causes but not (1), income
inequality. Some, such as Till van Treeck, identify (1), rising
income inequality, as well.

There is substantial evidence that the rising inter-

household inequality in the United States has im-
portantly contributed to the fall in the personal
saving rate and the rise in personal debt (and a
higher labour supply). Aided by the easy availabil-
ity of credit, lower and middle income households
attempted to keep up with the higher consumption
levels of top income households. This has contrib-
uted to the emergence of a credit bubble which
eventually burst and triggered the Great Recession.1

The evidence that lower- and middle-income households were

trying to keep up with the consumption of top-income house-
holds is less substantial than the evidence that they sought to
maintain their living standards in the face of stagnant or de-
clining incomes.2 Joseph Stiglitz, Raghuram Rajan, Paul Krug-
man, and Thomas Palley also name rising income inequality
as a cause of the jump of indebtedness and ensuing economic
crash.3 Two writers go further, linking inequality to the stag-
nant economic recovery from the Great Recession.4 They all
make well-reasoned arguments linking growing personal in-
Introduction 3

debtedness in part to rising income inequality, but fail to pro-

vide empirical support for such a link.
This book goes beyond their writings in three ways.
First, it presents econometric evidence supporting such a link.
Second, it uses household budget data to show that house-
holds’ increased indebtedness was not merely for leisure or
competitive conspicuous consumption. Rather, the drivers of
debt were increased spending on what most would agree are
necessities. Spending on shelter, health, and education has in-
creased significantly despite stagnant incomes. In other words,
with stagnant or declining incomes, households maintained
their consumption on essentials through massive borrowing.
And finally, it presents persuasive historical evidence that
the nation has been through this before—this is not the first
time that rising income inequality accompanied by growing
and unsustainable household debt and the bursting of a real
estate bubble ended in a severe economic crash.
Why did most economists fail to see this problem com-
ing? The inclusion of rising income inequality as one of the
four major causes of the financial crash and Great Recession
does not comport with the mainstream economic theory of
consumption. Indeed the econometric evidence, the house-
hold budget evidence, and the historical evidence argue that
the mainstream theory of consumption, which posits no role
for income inequality in the economy, is seriously flawed. The
story here is that increasing consumer indebtedness, which sup-
ported consumption until the crash in 2008–09, was driven by
the pressure for most households to maintain consumption in
the face of stagnant income as income inequality relentlessly
rose for thirty years or so. That debt-supported expansion of
consumption became unsustainable after 2007 once house
prices tumbled.
4 Introduction

Economists have ignored or misunderstood the effects

of rising income inequality on macroeconomic outcomes.
More­over, the mainstream consumption theories cannot ex-
plain recent trends in relative consumption and saving. Nei-
ther Milton Friedman nor Franco Modigliani and Richard
­Brumberg—the leading theorists of consumption in recent
economic thought—posited any role for the distribution of
income in their theories of consumption. Friedman’s perma-
nent income theory of consumption does not explain the ob-
served rise of debt-fueled consumption in the decade before
the crash. Modigliani and Brumberg’s life cycle theory of
consumption contains the seed of an explanation, but not one
that they anticipated.
However, if we look further back in time, Thomas Mal-
thus, writing in the early nineteenth century, had the germ of
an idea that excess saving, brought on by a top-heavy distribu-
tion of income, would curb effective demand and thus crimp
the expansion of total output. But Malthus had no data, and
his prose was less than lucid. A century later, Keynes picked up
on Malthus’s idea, which had lain dormant thanks to the tri-
umph of David Ricardo’s general equilibrium perspective on
the macro economy. Some of Malthus’s thinking on effective
demand is echoed in John Maynard Keynes’s General Theory
of Employment, Interest, and Money (1936).
Keynes’s theory of consumption, fully developed in the
General Theory and translated into algebra by his interpret-
ers, dominated macroeconomics for many years. It attributed
an important role for income distribution in macroeconomic
outcomes—namely, that the share of all households’ after-tax
income spent on consumption, the average propensity to con-
sume (APC), would decline over time as incomes rose, curb-
ing effective demand and perhaps leading to long-term stag-
Introduction 5

nation. But the post–World War II experience contradicted

his inference. The APC did not decline; it was either stable or
rising. So the Keynesian notion that a more equal distribution
of income would curb the fall of the APC disappeared, and
income distribution no longer mattered for the theory of con-
sumption.
Around 1985, however, something strange began. After a
long period of stability, as hypothesized by Friedman as well
as Modigliani and Brumberg, the APC began a long-term rise.
That meant a long-term fall in the saving rate for the same pe-
riod because the saving rate is equal to (1 – APC). That event
was not supposed to happen in Friedman’s theory.
Was the observed rise of the APC, 1984–2005, unprece-
dented? No. Based on Simon Kuznets’s data, there was a forty-
year rise of the APC, 1894–1903 to 1924–33. Kuznets’s data on
income distribution, which begins in 1920 and ends in 1938,
shows rising income inequality in the 1920s. The fact that
Kuznets’s long period of rising APC includes a decade of ris-
ing income inequality, and that the thirty-eight-year rise of
income inequality, 1974–2012, includes a long period of rising
APC, raises the question of whether there is a causal link from
rising income inequality to rising APC.
The mainstream consumption theory of Friedman as
well as Modigliani and Brumberg cannot explain such a link.
Instead, faced with slow or no income growth, households
might resort to increased borrowing to maintain some de-
sired level of consumption. The demand for borrowing can be
curbed by interest rates and a hard income constraint. But in
the period from about 1995 to 2007, especially post 2000, there
was an unusually huge rise in household indebtedness, fueled
by the four factors noted above.
This book does not address the question of how to fix ris-
6 Introduction

ing income inequality through public policy. However, it does

address the possible causes of rising inequality.
As to policies to redress rising income inequality, some
authors have recommendations that would be moves in the
right direction. They include reducing the amount of money
in political campaigns and lobbying, and enforcing the labor
laws and the antitrust laws on the books. But any effort to re-
dress income inequality must begin with a story about why
curbing and reversing income inequality matters for the long-
term health of the economy. That is the central goal of this
book.
The book will begin by briefly laying out the facts about
rising income inequality, a topic that has been exhaustively
covered elsewhere.5 Income inequality has been rising for
almost four decades. Median incomes and wages have stag-
nated, while the share of income going to the top 1 percent has
soared. We will list and evaluate the possible causes of rising
income inequality, and then examine the large rise in con-
sumer indebtedness post 1995. The rise of debt and income in-
equality has been accompanied by a measured increase in eco-
nomic insecurity among consumers. Also, relative spending
on housing, health, and education has risen markedly, squeez-
ing relative spending on other necessities, so we will see some
reasons why income inequality matters. Turning to economic
theory, the book traces the treatment of income distribution,
or lack thereof, in theories of consumption from Malthus, Ri-
cardo, and Keynes to Friedman, Modigliani, and Brumberg,
to modern critics of mainstream neoclassical consumption
theory, including behavioral economists. The dominance of
Friedman’s as well as Modigliani and Brumberg’s theories of
consumption among macroeconomists up to the present ex-
plains why most, but not all, economists have not noticed that
Introduction 7

income inequality matters. Then we present an outline for a

revised theory of consumption that fits the facts. As we will
see, the rising inequality and debt leading up to the Great
Recession matches a similar trend that preceded the Great
­Depression.
 II
Trends in Income Distribution

he changing distribution of income in the United

States has some distinguishing characteristics. The first
is that the share of national income going to labor
has been declining. Splitting labor income between
the share
the share goinggoing
to thetotop
the1 top 1 percent
percent of salary,
of wage, wage, salary, and
and bonus
bonus earners
earners and theand the bottom
bottom 99 percent
99 percent shows
shows that the that the top
top share is
share isTrends
rising. rising. in
Trends in productivity,
productivity, hourly earnings,
hourly earnings, male and male
fe-
and female,
male, as wellasaswell as all household
all household income income show favoring
show shifts shifts favor-
the
ing the highest
highest incomeincome
groups.groups.
It is important to explain what income inequality means,
what it is and what it is not. If the proportion (share) of ag-
gregate income received by the lower end of the income dis-
tribution is falling over time, that is rising income inequality.
But how is “lower” defined—99 percent, 95 percent, 90 per-
cent? An easier rule of thumb is that if the Gini coefficient
or the Theil index is rising over time, that is rising income
inequality.
Rising income inequality does not necessarily mean stag-
nant real incomes for most households. During the Depression
years, the nation had stagnant incomes for most households
Trends in Income Distribution 9

but falling income inequality. From 1947 through 1973 the na-
tion had rapidly growing incomes for most households and
also had falling income inequality. And in the decades leading
up to the financial crash and Great Recession, the nation had
stagnant income growth for most households (“stagnant” here
is defined as real mean income growth of less than 1 percent
per year) and rising income inequality. The point is that the
two factors—rising income inequality and stagnant household
income growth—do not necessarily always occur together.
Indeed, although the data are murky, household income was
likely growing from 1890 to 1914 and income inequality was
rising. In what follows, most of the evidence is about the de­
cades preceding the financial crash, when both rising income
inequality and stagnant income growth for most households
occurred together. That was not a coincidence. The strong
growth of aggregate income, 1947–73, was cut almost in half
in the recent period, 1973–2007. At the same time, the share of
income going to the top 10 percent rose from 32 to 46 percent.
Slower growth plus a declining share for the bottom 90 per-
cent meant that average real income of that group did not rise.
Of course, it rose strongly for the top 10 percent.1

Labor’s Declining Share of National Income

One of the regularities noted in the past about the United
States economy was the long-run stability of the shares of na-
tional income paid out to labor and to owners of capital—two-
thirds to labor and one-third to capital.2 Table 2.1 presents the
share to all labor for the years 1969 to 2013. The share barely
changed in the twenty years from 1969 to 1989: 65.0 percent to
65.2 percent. But then it began to edge downward, falling to
60.7 percent by 2013.
10 Trends in Income Distribution

Table 2.1. Labor Share

of National Income,
1969–2013
Year All Earnings
1969 65.0%
1979 66.6%
1989 65.2%
1999 64.1%
2007 63.4%
2013 60.7%
Source: Bureau of Economic Analysis
(n.d.), Table 1.12.

The decline in labor’s share of national income is not

limited to the United States, but is seen in other rich coun-
tries, as noted by Peter Orzsag.3 He calculates that the drop
of five percentage points of labor’s share of private sector in-
come from 1990 to 2011 is equivalent to a loss of $500 billion.
He attributes the drop in share to technological change and
globalization. Joseph Stiglitz does not agree. About the declin-
ing wage share, Stiglitz noted, “The pattern and magnitude of
changes in labor compensation as a share of national income
are hard to reconcile with any theory that relies solely on con-
ventional economic factors.”4 And further, “If technological
change increases the effective supply of labor, and labor and
capital are not very substitutable, then technological change
drives down the share of labor. But the pattern of increase of
wages—with wages at the very top (e.g., of bankers) increasing
so much relative to that of others—is consistent with the view
that something else besides technological change is causing
the decline in the wage share.”5
Trends in Income Distribution 11

Table 2.2. Shares of Earnings to Top

1 Percent and 99 Percent, 1929–2011
Year Top 1 Percent 99 Percent
1929 8.7% 91.3%
1969 5.2% 94.8%
1979 6.2% 93.8%
1989 8.7% 91.3%
1999 11.7% 88.3%
2007 12.2% 87.8%
2011 11.0% 89.0%
Source: Based on Saez (2013), Table B2.

The Rise of the 1 Percent Among Wage Earners

There is a long-standing popular perception that the rich get
their income from ownership of capital, while workers get
their income from wages and salaries. This was roughly true
during the roaring 1920s, but it is not true any longer. Based on
income tax data, the top 1 percent of tax filers for 1929 received
29 percent of their income from wages. But in more recent
years, their share from wages (which includes bonuses) has
ranged from 54 to 66 percent.6 It is still certainly the case that
most capital income accrues to the 1 percent, but they, mostly
managerial and professional workers, are receiving an increas-
ing share of all earnings.7 Table 2.2 shows the shares of total
wage income for the top 1 percent and the other 99 percent for
selected years, 1929 to 2011. The two shares sum to 100 percent,
of course. The share of the top 1 percent moves down from 9
percent of all earnings in 1929 to 5 percent in 1969. Thereaf-
ter their share rises to 12 percent in 2007 and drops to 11 per-
cent by 2011 with the onset of the Great Recession. The 1 per-
12 Trends in Income Distribution

cent’s gain in share since 1969 mirrors the 99 percent’s drop in

share.

Productivity, Hourly and Annual Earnings, and

Rising Income Inequality
It is a fundamental truth of economics that living standards
can rise in the long run only if productivity rises. Although
that condition is necessary, it is not sufficient. Compensation
must rise in tandem with productivity growth for living stan­
dards to improve. Figure 2.1 shows the trends of real hourly
compensation and productivity (real output per hour worked)
for the United States for the post–World War II period. The
two measures rise together until the 1970s, and then diverge.
Growth of compensation no longer keeps up with growth of
productivity.
But the average growth rates conceal how the productiv-
ity gains have been distributed. In an analysis of Internal Reve-
nue Service (IRS) data examining productivity growth, the au-
thors conclude, “Our most surprising result from the large IRS
data set is that, over the entire period 1966–2001, only the top
10 percent of the income distribution enjoyed a growth rate of
total real income (excluding capital gains) equal to or above the
average rate of economy-wide productivity growth. The bottom
90 percent of the income distribution fell behind or were even
left out of the productivity gains entirely.”8 This pattern of pro-
ductivity growth outstripping wage growth over the past three
decades is repeated for other rich nations.9
The drop in the earnings share of national income, the
drop in the 99 percent’s share of earnings, and the disconnect
of hourly compensation growth from productivity growth all
point to relative slowing in the growth of individuals’ earnings.
Trends in Income Distribution 13

Figure 2.1. Productivity and real hourly compensation, nonfarm

business sector, 1947 through first quarter 2012. Source: Bureau of
Labor Statistics (2015).

Table 2.3 documents the slowdown of earnings growth. Panel

A of Table 2.3 shows trends in median real earnings, male and
female, of full-time year-round workers. The median real wage
for males increased about $7,000 in the seven years from 1967
(the first year of the Current Population Survey income data)
to 1974 (the year after which income inequality began to rise).
However, in the thirty-three years from 1974 to 2007, that wage
barely changed. The median for females grew more than the
male median in both periods, but it also slowed markedly in
the long period, 1974–2007.
It is clear that growth of real wages was either stagnant
(for females, 0.9 percent per year) or almost nonexistent (for
males, 0.1 percent per year). Some observers have argued that
the massive rise of women in the labor force post 1970 was in
 Table 2.3. Summary of Real Earnings and Income Trends, 1967–2013
Average annual percent change
1967 1974 2007 2013 1967–74 1974–2007 2007–13
Panel A. Median real
earnings of full-time
year-round workers,
2013 $
  Male 43,796 50,540 50,684 50,033 2.1% 0.1% –0.2%
  Female 25,307 29,694 39,436 39,157 2.3% 0.9% –0.1%
Panel B. Median 43,558 47,702 56,436 51,939 1.3% 0.5% –1.4%
household real
income, 2013 $
Panel C. Mean 48,717 55,784 75,957 72,641 2.0% 0.9% –0.7%
household real
income, 2013 $
Source: Median real earnings of full-time year-round workers from DeNavas-Walt and Proctor (2014), Table A-4. Median and
mean household real income from Table A-1.
Trends in Income Distribution 15

part a coping measure by households to deal with flat wages

for men.10 In Panel C of Table 2.3, median household income is
compared with mean household income. Growth of both the
median and mean slowed markedly during 1974–2007 com-
pared with 1967–74.
None of the statistics presented so far measures income
inequality directly. Rising income inequality means that the rel-
ative distribution of income from all sources—earnings, divi-
dends, interest, rent, and transfer payments—becomes smaller
for those at the lower end of the distribution and larger for
those at the higher end of the distribution. There had been a
long decline in income inequality, a rise in income equality,
that was evident in the years after World War II. That decline
began during the Depression of the 1930s and was accelerated
by World War II.11 The turnaround to rising income inequality
occurred in the mid- to late 1970s. In Table 2.4, 1974 is the
turnaround year, because after that date, the two measures of
income inequality, the Gini coefficient and the Theil index, are
always above their 1974 levels (for both measures, increases
mean greater inequality and decreases mean less inequality).
Panel A of Table 2.4 presents percentage shares of ag-
gregate income (before taxes) going to each quintile of house-
holds. Note that in the turnaround year of 1974 the relative
distribution is quite similar to the 1967 distribution. Income
inequality rose after 1974. Every quintile, except the highest in-
come quintile, showed drops of share from 1974 to 2007. Loss
of share for the bottom four quintiles continued through 2013.
Panel B of Table 2.4 shows mean household income by
quintile in 2013 dollars for selected years, 1967–2013. The last
column presents the average annual percentage changes in
the means from 1974, the turnaround year, to 2007, the last
year before the Great Recession began. All of the changes are
 Table 2.4. Summary of Income Distribution Trends, 1967–2013
Average annual percent
1967 1974 2007 2013 change, 1974–2007
Panel A. Shares of household income of quintiles
Lowest quintile 4.0% 4.3% 3.4% 3.2%
Second quintile 10.8% 10.6% 8.7% 8.4%
Third quintile 17.3% 17.0% 14.8% 14.4%
Fourth quintile 24.2% 24.6% 23.4% 23.0%
Highest quintile 43.6% 43.5% 49.7% 51.0%

Panel B. Mean household income of quintiles, 2013 $

Lowest quintile 9,615 11,685 12,792 11,651 0.3%
Second quintile 26,643 29,044 32,603 30,509 0.4%
Third quintile 42,540 46,811 55,344 52,322 0.5%
Fourth quintile 59,519 67,603 87,606 83,519 0.8%
Highest quintile 107,112 119,777 186,009 185,206 1.3%

Panel C. Measures of income ­inequality

Gini Index 0.397 0.395 0.463 0.476
Theil 0.287 0.267 0.391 0.415

Panel D. Household income ratios

90th percentile/10th percentile 9.23 8.58 11.18 12.10
90th percentile/50th percentile 2.11 2.23 2.71 2.89
Source: DeNavas-Walt and Proctor (2014), Tables A-2 and A-3.
Trends in Income Distribution 17

positive, but they are quite dissimilar. Quintiles one through

four grew less than 1 percent per year, defined here as stagnant
growth. The top quintile grew more than 1 percent per year.
Panel C lists the two measures of income inequality, the Gini
index and the Theil, for the same years. In 1974 both the Gini
index and the Theil were slightly lower than in 1967, indicating
reduced income inequality. However, both are markedly
higher in 2007 and in the last year, 2013.
Panel D shows the household income ratios of the 90th
percentile (households in the top 90 percent of the income dis-
tribution) to the 10th and the 50th percentiles (households in
the bottom 10 percent of the income distribution and house-
holds in the 50th percent). A rising ratio over time means that
income growth in the 10th percentile is falling behind income
growth in the 90th percentile. A falling ratio means the oppo-
site. Note that the 90:10 ratio falls from 9.23 in 1967 to 8.58 in
1974, meaning that income growth was faster in the 10th per-
centile (the bottom group) than in the 90th percentile (the top
group). That trend was reversed after 1974. In 2007 the ratio
was 11.18, and it was still higher in 2013. The pattern is similar
for the 90:50 ratio, although the rise after 1974 is less extreme.
The national trend of rising income inequality begin-
ning in the mid-1970s is matched by a state trend of diverging
wage growth beginning in the 1980s. The neoclassical model
infers that wages in sub-parts of the nation will converge over
time, as lower-wage labor moves to higher-wage areas, putting
downward pressure on wages, and reduced labor supply in the
low-wage areas puts upward pressure on wages. Historical data
for states back to 1870 mostly support long-run convergence,
except for the decades of the 1920s and the 1980s, but the di-
vergence in the 1980s and the 1920s was dismissed by Robert
Barro and Xavier Sala-i-Martin as aberrations at the time they
18 Trends in Income Distribution

wrote.12 A more recent study using metropolitan areas instead

of states found that the income divergence of the 1980s con-
tinued into the decade of the 1990s.13 Thus divergence of in-
come among cities and states may be replacing convergence of
income argued by theory and mostly supported by the earlier
data. What does this have to do with income inequality? If ris-
ing national income inequality is accompanied by divergence
of income among parts of the nation, then regional divergence
may be one source of the increase in national income ine-
quality. James Galbraith provides recent striking examples of
metropolitan areas surging way ahead of the pack in income
levels.14
The survey-based Census data in Table 2.4 indicate that
strong gains of income are concentrated in the top quintile or
the top decile. The careful analysis of the long-term chang-
ing distribution of income, by Thomas Piketty, Emmanuel
Saez, and their collaborators, is focused upon a breakdown of
the top decile.15 Their data, based on tax files rather than sur-
veys, produces more accurate estimates of top income shares.
Their data show that the gains in income share of the highest
quintile illustrated in Table 2.4 are concentrated in the top 10
percent and especially the top 1 percent. Figure 2.2 shows the
share of income received by the top 10 percent and top 1 per-
cent from 1917 to 2012. The top 10 percent received over 32
percent of household income in 1974, a share hardly changed
since 1947. But by 2012, the top 10 percent share was 47 per-
cent. The top 1 percent fared even better, receiving 22 percent
of household income in 2012, up from 8 percent in 1974.16
The sharp rise in income shares going to the top 10 per-
cent and 1 percent post 1974 is not accounted for by renters.
“The large shocks that capital owners experienced during the
Great Depression and World War II seem to have had a per-
Trends in Income Distribution 19

Figure 2.2. Shares of pre-tax income to top 10 percent and top 1

percent, excluding capital gains, 1917–2012. Source: Saez (2013).

manent effect: top capital incomes are still lower in the late
1990s than before World War I.” On the other hand, they show
that wage shares “were flat before World War II and dropped
precipitously during the war. Top wage shares have started
recovering from this shock only since the 1970s but are now
higher than before World War II.”17
 III
Possible Causes of Rising
Income Inequality

W
hy has income inequality been rising for al-
most forty years, as documented in Chapter II?
There have been some excellent books and
articles on causes of income inequality. They
have one characteristic in common—they treat the major
causes as political and institutional, not economic.1
This chapter is no exception. It weighs the evidence and
passes tentative judgment. Of four broad categories of possible
causes—economic, demographic, institutional, and political
—the first two seem to be the least important.

Economic Causes
The view among most economists is that the pre-tax distri-
bution of income is the result of market forces. The govern-
ment amends the market outcome through taxes, transfers,
and expenditures. Therefore, in the search for causes of rising
income inequality among those who hold that belief, political
Possible Causes of Income Inequality 21

causes are off the table. Among economists, the three most cited
causes of rising income inequality are globalization, skill-biased
technological change (SBTC), and job polarization. By globali-
zation, they mean a number of factors that have become more
important over the decades in the U.S. economy, including
reduced trade barriers, increased immigration, lower inter-
national transport costs, off-shoring of production, foreign
competition, and increased capital flows. In other words, U.S.
labor is faced with more competition from foreign labor than
in the past, because tariff barriers and transport cost barriers
have diminished, making labor costs relatively more impor-
tant. The increased off-shoring of production reflects the rise
in importance of relative labor costs. Transport technology
(container ships, super tankers, jet freight) and political agree-
ments (the World Trade Organization, multi-nation trading
blocs) have reduced transport costs and tariff barriers, making
relative labor costs loom larger. It is not only goods production
that has been off-shoring to nations with lower labor cost. Ser-
vices such as call centers, routine legal and medical services,
and software production have also shifted abroad, usually to
English-speaking nations. The shift of services would not have
been possible without the massive decline in cost and time of
telecommunication services over the past fifty years. What
does globalization have to do with rising income inequality?
All of the factors noted put U.S. labor at a cost disadvantage
with Asian, Latin American, and eastern European labor. U.S.
wages will tend to rise more slowly than in the past before
globalization. Most economics texts and media analysts treat
globalization as the end of the story, but there is a problem
in that analysis: Canada, the United Kingdom, France, Ger-
many, and Japan are subject to the same forces of globaliza-
tion as the United States. Have they had the same increases
22 Possible Causes of Income Inequality

of income inequality? The United Kingdom and Canada have

had increases, though less extreme than in the United States.
Germany and Japan have had very little increases of inequality,
and France has had none at all.2 The comparison suggests that
there is far more than globalization underlying the U.S. rise of
income inequality.
Along with globalization, SBTC is another cause most
cited by economists in explaining the rise of inequality. SBTC
is defined as:

a shift in the production technology that favours

skilled over unskilled labour by increasing its rela-
tive productivity and, therefore, its relative demand.
Traditionally, technical change is viewed as factor-
neutral. However, recent technological change has
been skill-biased. Theories and data suggest that
new information technologies are complementary
with skilled labour, at least in their adoption phase.
Whether new capital complements skilled or un-
skilled labour may be determined endogenously by
innovators’ economic incentives shaped by relative
prices, the size of the market, and institutions. The
“factor bias” attribute puts technological change at
the center of the income-distribution debate.3

Some examples can help here. In the past, the same workers
who dug ditches with shovels could learn to operate a back-
hoe. The same workers who moved and stacked boxes in a
warehouse could learn to operate a forklift. Productivity rose,
but the labor skills required were not of a higher order. That
has changed with the ubiquitous use of computers in factories,
offices, and retail stores. There is a premium on information
Possible Causes of Income Inequality 23

technology (IT) skills, often associated with college education.

Even though the supply of college graduates has been expand-
ing while the supply of high-school-only graduates has been
shrinking, there has been a growing premium for the better
educated. From 1979 to 2007 the median hourly wage of those
with college degrees rose four times faster than the median
hourly wage of those with only high school degrees.4 This sug-
gests that demand for highly educated workers has been out-
running the increasing supply.
One prominent economist who names SBTC as a chief
cause of rising income inequality, even among the top 1 per-
cent, is Gregory Mankiw. He claims that rising income ine-
quality at the top is not because of politics or rent-seeking
but rather supply and demand.5 In other words, SBTC makes
employers search out the best and brightest and reward them
handsomely. Although one would think that SBTC does not
affect incomes at the very top, some claim that SBTC affects
pay of chief executive officers (CEOs), financial executives,
attorneys, and athletes.6 To place CEOs in the same cate-
gory as athletes ignores the distinction between market and
non-market forces. As Ian Dew-Becker and Robert Gordon
have argued, “The core distinction is that superstars and other
market-driven occupations have their incomes chosen by the
market, whereas CEO compensation is chosen by their peers
in a system that gives CEOs and their hand-picked boards of
directors, rather than the market, control over top incomes.”7
Piketty and Saez are critical of the SBTC explanation of
rising income inequality. Wage shares in the United States, they
argue, cannot be fully accounted for by skill-biased technologi-
cal change, the favored explanation among economists. But for
one of the same reasons globalization cannot fully explain rising
income inequality in the United States, neither can SBTC. All
24 Possible Causes of Income Inequality

industrialized nations experience SBTC, yet only the United

States has had extreme increases of income inequality.8
Piketty and Saez are not alone in questioning the SBTC
explanation for rising income inequality. Joseph Stiglitz notes,
“Skill biased technological change has little to do with the enor-
mous increases in wealth at the very top.” Political scientists
Jacob Hacker and Paul Pierson note that the rising inequality
story is in the top 1 percent of households. They argue that
education differences among workers and skill-biased techno-
logical change cannot fully explain the hyperconcentra­tion of
income at the top.9
There is no doubt that some of the rising inequality below
the very top can be attributed to globalization and SBTC. But
that cannot be the full story, because other rich, developed
nations subject to the same forces have had more modest in-
creases of inequality than the United States, and neither glo-
balization nor SBTC can explain the huge gains of income
share of the top 1 percent and 0.01 percent.
The third possible cause of rising income inequality de-
veloped by prominent labor economists is job polarization,
“usually defined as stronger employment growth in jobs at the
top and bottom of the wage distribution than in the middle.”10
Job polarization is commonly described as a “hollowing out”
of middle-skill jobs. However, the job polarization model does
not well describe changes in the labor market and link them
to the rise of wage inequality. As Lawrence Mishel and his col-
leagues note, upgrading of occupations in the United States
has been a long-term trend that can be traced back to 1950
with available data. Thus, it was occurring in the period of fall-
ing wage inequality, 1947–74, and in the period of rising wage
inequality, in the late 1970s.11
In his recent book Capital in the Twenty-first Century,
Possible Causes of Income Inequality 25

Piketty identifies an economic cause of rising income inequal-

ity at the top. As the capital-income ratio slowly rises over time
and the annual return to capital grows faster than gross do-
mestic product (GDP), the share of national income going to
the owners of capital rises, and so the share going to labor de-
clines. The ownership of capital is highly concentrated among
the top 1 percent of the income distribution, and so their share
inexorably rises.12 But that is a very long-term story and can-
not fully account for the rising share of the top 1 percent docu-
mented in the previous chapter. It reflects the fact noted there
that one-half to two-thirds of the income of the top 1 percent
comes from earnings rather than capital. So Piketty’s hypothe-
sis, if true, would be a minor cause for the United States. Thus,
to account more fully for the rise of income inequality requires
looking beyond economic explanations.

Demographic Causes
The demographic portrait of the United States has undergone
marked changes over the past thirty years or so. Some of those
could raise income inequality. Married couples with young
children have diminished as a share of all households, while
single-person households have risen. The elderly population
is growing rapidly as the baby boom generation moves into
retirement. One of the most noted changes has been the in-
creased labor force participation by women, which rose from
51 percent in 1979 to 58 percent in 2012.13 One of the demo-
graphic changes that has raised income inequality is the ten-
dency of highly educated employed females to marry males
of the same status. In the 1950s, college-educated males who
married were far more likely to have stay-at-home wives than
today. A one-earner family back then with a college education
26 Possible Causes of Income Inequality

had higher earnings on average than a one-earner family with

high school only. Today, two-earner families, where both have
college degrees, have far higher earnings than a one- or even
two-earner family with high school degrees only. Thus they
are further apart on the income distribution.
How much of the rise of income inequality could be ex-
plained by demographic shifts like the ones above? Rebecca
Blank undertook a thorough examination of that issue in her
book Changing Inequality (2011). Blank performs a number of
careful simulations of effects on income distribution of various
hypothesized demographic changes, such as: What if family type
and size remained unchanged from 1979 to 2007? She finds, “In
general, the results suggest that none of these changes, by them-
selves, would have major effects on income distribution. . . .
Even large changes, however, leave income inequality closer to
its 2007 level than its 1979 level, suggesting that a major rever-
sal in inequality is unlikely in the absence of substantial and
currently unforeseen changes.”14 Her rigorous analysis of de-
mographic factors concludes that only 15 percent of the rise of
income inequality since 1979 can be attributed to them. As she
summarizes her findings, “The results of this detailed analysis
indicate that changes in family composition and family size
account for about 15 percent of the rise in U.S. income inequality,
while changes in income account for the remaining rise in ine-
quality. Most of this rise is due to increases in wage inequality.”15
Thus, neither economic causes nor demographic causes can fully
explain rising income inequality in the United States.

Institutional Causes
The most convincing explanation for rising income inequal-
ity lies in an examination of institutional and political factors.
Possible Causes of Income Inequality 27

One is the decline in labor unions. The peak of unionization in

the United States was 30 percent in 1960. In 2012, union mem-
bership was down to 11 percent.16 One could note forces in a
modern economy pushing union membership downward. For
example, rising productivity in manufacturing has led to ab-
solute reductions in the number of production workers even
as output increases. Furthermore, employment has shifted out
of goods production and distribution industries (manufactur-
ing, wholesale trade, transportation, and warehousing), where
unions were traditionally strong, and into service-type indus-
tries, such as retail trade and health services, where unions
had not been prominent. But those forces are at play in other
modern rich nations without a similar effect on unionization.
When the U.S. unionization rate was 30 percent in 1960, the
rate in Canada was 32 percent. It is still around 30 percent
there.17 Given that both economies are subject to the same
market forces, how can we explain the precipitous drop in U.S.
unionization while in Canada the rate is where it was forty
years ago? Jacob Hacker and Paul Pierson argue that the differ-
ence in labor law in the two countries account for union cov-
erage shrinking in the United States and not in Canada. Some
Canadian provinces have laws that allow for card check certi-
fication and first contract arbitration. Provinces ban the hiring
of permanent strike replacements and employer interference
into unionization campaigns. In contrast, anti-union action by
employers in the United States has met little resistance by gov-
ernment authority. As they point out, inaction as well as action
can undercut the power of unions. “The absence of an updat-
ing of industrial relations policy has had brutal effects on the
long-term prospects of organized labor.”18 A major labor law
reform bill promoted by organized labor in 1978 that would
have accomplished an updating of labor policies by banning
28 Possible Causes of Income Inequality

the use of strike replacements was supported by a majority of

the House and Senate and President Jimmy Carter, all Dem-
ocrats. However, it was derailed by a filibuster in the Senate,
supported by some Democrats, and was never enacted.19
Of course the waning power of labor unions is not the
only factor to explain the tremendous rise of income shares
at the top of the distribution. Rather, a large part of the ex-
planation lies in the increasing political power and effective
organization of business interests, including businesses whose
clients are at the top 1 percent of the income distribution, such
as mutual funds and other financial firms. As income going to
the top end of the distribution has been rising, they raise fund-
ing to influence political outcomes in Washington and state
capitals by lobbying and political contributions. The failure of
progressive labor law legislation is only one example of their
success in influencing social policy. More examples follow in
the next section.

Political Causes
Economists stress market forces and technology as causes of
rising income inequality. Political scientists stress the median
voter theorem in their analysis of why income inequality has
been rising.20 But neither of those views takes into account
organized interests. In an article called “Winner-Take-All
Politics” (2010), which they later developed into a book with
the same title, Jacob Hacker and Paul Pierson present a co-
gent empirical story about the sharp rise of income shares at
the top developed around three claims. First, government in-
volvement in the modern economy is broad and deep. Second,
policy transformation occurs through both enactment and
Possible Causes of Income Inequality 29

non-enactment. Third, shifts in organized interests are a major

force in policy change.
The first, government involvement in the modern econ-
omy is broad and deep, flies against the conventional view
among most economists that the distribution of pre-tax income
is the result of market forces. The role of government, they
argue, is limited to the fiscal side: taxation and transfers that
can alter the market distribution of income. This is a naive
view. A number of government policies tilt the distribution of
pre-tax income in favor of the very top of the income distribu-
tion, including:

1. Tort reform and arbitration law trends that cur-

tail power of consumers and stockholders to hold
corporation management legally accountable for
purported wrongdoing.
2. Special treatment of corporate stock option
awards.
3. Restricting access to bankruptcy protection for
consumers and business.
4. Extending time of copyright protection for some
large firms.
5. Extending time of patent protection for non-­
generic drugs.
6. Forbidding Medicare to bargain for lower phar-
maceutical prices.

Note that all six of these policies favor corporations and their
owners. The current broad and deep involvement of the fed-
eral government in the economy is similar to that of the Gilded
Age. Long ago, in 1892, John R. Commons, an economics pro-
30 Possible Causes of Income Inequality

fessor at the University of Syracuse, argued that a substantial

share of U.S. corporations owed their quasi-monopoly mar-
ket power to privileges and protections, such as patents and
copyrights, bestowed by the federal government.21 The ensuing
storm of protest from business and from economists led to his
dismissal by the University of Syracuse. He was right, and he
touched a nerve.
The second claim is that policy transformation occurs
through both enactment and non-enactment, or what they
call “policy drift.” Non-enactment occurs through filibusters
in the Senate, a tactic increasingly pursued in the polarized
body. According to Senate rules, ending a filibuster requires a
supermajority of sixty votes. Thus, a determined minority can
use the filibuster to block legislation. In the fifty years from
1919 to 1969, fifty-six motions were filed to stop a filibuster, but
from 1969 through 2009 there were 1,100 filed, most of them
after 1991.22
Finally, shifts in organized interests are a major force in
policy change. Hacker and Pierson document a huge rise of
special-interest organizations in Washington beginning in the
1970s. Corporations with a public affairs office in Washing-
ton went from one hundred in 1968 to five hundred by 1978.
Further, the three giants of promoting and protecting corpo-
rate interests, the National Association of Manufacturers, the
Business Roundtable, and the Chamber of Commerce, greatly
expanded their membership and budgets after 1970. Needless
to say, the headquarters of all three are in Washington, D.C.
The National Association of Manufacturers was formerly in
New York, but it moved to Washington around 1974.23
Labor and consumers, the main countervailing powers
to businesses in a capitalist democracy, have no similar weight
in Washington. Unions are the only organizations pushing for
Possible Causes of Income Inequality 31

Table 3.1. Lobbying Presence

in Washington
Institution 1981 2006
Business 7,059 12,785
Union 369 403
Public interest 237 405
Source: Based on Drutman (2010).

bread-and-butter issues for workers in Washington, yet they

have a small fraction of the lobbyists employed by business,
and they are falling behind. Other liberal organizations push-
ing environmental issues, civil rights, and women’s issues are
also falling behind, as shown in Table 3.1.
An example of the overwhelming numbers of lobbyists
representing business interests: there are about one thousand
registered Washington lobbyists who list taxes as one of their
areas. Yet in the estate tax fight, an issue of great importance
for income distribution, only one union lobbyist was available
to represent worker and consumer interests.24
Although most attention of the media is on election
campaign funding, that apparently is not where corporations
spend more to influence government outcomes. “Companies
generally spend about twelve times more on lobbying than
they spend on campaign contributions [political action com-
mittees, or PACs].”25 Lobbying expenditures in Washington,
adjusted for inflation, have risen 77 percent since 1998. That
is far more than other measures of legislative activity, such as
bills introduced (+43 percent), federal budget (+38 percent),
and Federal Register pages (+18 percent).26
Shifts in organized interests favoring the issues of corpo-
rations and the wealthy are also reflected in the rise of think
32 Possible Causes of Income Inequality

tanks funded by conservative interests—to name the most

prominent, the American Enterprise Institute (AEI), the Her-
itage Foundation, the Olin Foundation, the Hoover Institu-
tion, and the Cato Institute. They are heavily engaged in lobby-
ing and political suasion on behalf of conservative viewpoints.
Older think tanks such as the Brookings Institution and the
Twentieth Century Fund (now the Century Fund) could be
described as centrist or liberal, and engage in much less ad-
vocacy than the new conservative organizations. The Heritage
Foundation allocates 20 percent of its budget on public rela-
tions and outreach, whereas Brookings allocates 5 percent.27
Many if not all of the objectives of business lobbying,
election campaigning, and advocacy can be described as
rent-seeking. The economic definition of rent-seeking is

Spending time and money not on the production of

real goods and services, but rather on trying to get
the government to change the rules so as to make
one’s business more profitable. This can take vari-
ous forms, including seeking subsidies on the out-
puts or the inputs of a business, or persuading the
government to change the rules so as to keep out
competitors, tolerate or promote collusion between
those already engaged in an activity, or make le-
gally compulsory the use of professional services.28

In this definition, rent-seeking is the expenditure of resources

to make one’s slice of the pie—GDP—larger at the expense of
someone else’s share. Resources so spent are wasted because
they do not add to society’s total output; they simply change
the shares received by each of the parties. Rent-seeking is not
always facilitated by government action, as implied by the
Possible Causes of Income Inequality 33

quoted definition. It can result from government inaction as

well as from actions by private parties. Successful rent-seeking
that shifts more of national output (income) to the top 1 per-
cent or 0.01 percent may well be the most important cause of
rising income inequality at the top of the income distribution
over the past forty years in the United States. It is covered here
under “political causes” because it is most often facilitated by
government action or government failure to act.
In his recent book The Price of Inequality, Joseph Stiglitz
places rent-seeking front and center in Chapter 2 (“Rent Seek-
ing and the Making of an Unequal Society”). He claims that
“some of the most important innovations in business in the
last three decades have centered not on making the economy
more efficient but on how better to ensure monopoly power or
how better to circumvent government regulations intended to
align social returns and private rewards.”29 Mankiw argues to
the contrary that “there is no good reason to believe that rent
seeking by the rich is more pervasive today than it was in the
late 1970s.”30 But there is a good reason. The top marginal tax
rate was around 70 percent in the late 1970s. It has since been
lowered a few times as well as raised and is now 39.6 percent.
That means any successful rent-seeking effort by those in the
top tax bracket today has an after-tax payoff almost double the
size of a similar one in the 1970s.
For example, it is assumed by economists that perfect
competition requires parties to transactions to be equally en-
dowed with information. Yet bankers, the sellers of derivatives,
have been fighting to keep derivatives in the opaque “over-the-
counter” market where the bankers know far more about the
derivatives they trade daily than the sometime buyers.
Echoing Commons in 1892, Stiglitz argues that patent
law can protect monopoly power. “The details of patent law
34 Possible Causes of Income Inequality

can extend the life of the patent, reduce entry of new firms,
and enhance monopoly power. America’s patent laws have
been doing exactly that. They are designed not to maximize
the pace of innovation but rather to maximize rents.”31
The failure of government to act in corporate govern-
ance provides what might be the largest single cause of rising
income inequality at the very top. As is well known, average
CEO pay has been growing rapidly since about 1980, with
some cyclical ups and downs. But the base of corporate reve-
nues, value added, or stock prices has been growing slower. In
other words, CEOs are taking a bigger slice from a moderately
growing pie. That is rent-seeking. The losers are lower-level
employees and stockholders. But there is not agreement on
that issue. In a paper examining why CEO pay has increased so
much, the authors develop a model that “can explain the recent
rise in CEO pay as an equilibrium outcome of the substantial
growth in firm size.”32 Gordon and Dew-Becker are skeptical:
“We endorse their idea [principal-agent control of stockholders
should be reversed] that managerial power lies behind some
of the outsized gains in CEO pay, while also recognizing that
stock options created an automatic spillover from the stock
market gains of the 1990s directly into executive pay.”33
It is difficult to show rent-taking with available data.
However, Josh Bivins and Lawrence Mishel present evidence
showing growth of CEO pay (including options exercised) of
the top 350 Standard & Poor (S&P) 500 firms based on sales.
By looking at the S&P 500 stock index over many years, they
show that when the S&P 500 went up in 1978–2000 (+513 per-
cent), CEO compensation went up much more (+1,279 percent).
When the S&P went down in 2000–2012 (–28 percent), CEO
compensation went down about the same (–29 percent).34
 IV
Consumers’ Shift to Debt

R
ecent years have seen a huge rise of household debt,
and rising income inequality has likely been a major
cause of this increase in debt.1 Here we will take a
look at the rise of debt and examine the economet-
ric evidence that supports the argued link from rising income
inequality to the rise of household debt. Debt-to-income ratios
rose sharply as growth of household debt far exceeded growth
of income. Households’ stagnant incomes and rapidly rising
house values induced them to take on far more debt, a move
facilitated by relaxed credit standards and low interest rates.

Household Debt: National Macro Data

The System of National Accounts financial data indicate that
households, combined with nonprofit institutions, were net
lenders for most of the long period since the 1960s. That is,
their net savings exceeded their net capital formation (primar-
ily residential investment).
Figure 4.1 illustrates that pattern of households shifting
from long-term net lenders to net borrowers in relative terms.
36 Consumers’ Shift to Debt

Figure 4.1. Net saving and net capital formation as percentage of

disposable income, 1974–2012. Source: Bureau of Economic Analysis
(n.d.), “National Economic Accounts,” Table S.3.a, December 2013.

The top line in the left part of Figure 4.1 is net saving (lend-
ing) as a percentage of disposable income. The lower line is net
capital formation (primarily residential investment) as a per-
centage of disposable income. Both are for the household and
nonprofit institutions sector. From the mid-1970s to the early
1980s, the share of savings fluctuated around 10 percent of dis-
posable income; thereafter it mostly declined through 2006.
The net capital formation (borrowing) share of disposable in-
come fluctuated below 5 percent in the early years, and then
in the mid-1990s it began rising to its peak in 2005. By the
late 1990s, the net capital formation share moved above the
declining net savings share. After 2005, both lines abruptly
change direction, so that by 2010 the net savings share is well
above the net capital formation share. The fact that households
Consumers’ Shift to Debt 37

went from a long-term net lending position to a massive net

borrowing position beginning in 1999 suggests that the System
of National Account data had pointed to an imminent finan-
cial crisis.2
To get a detailed picture of the rising indebtedness among
families alone, excluding nonprofit institutions, requires data
from the Survey of Consumer Finances (SCF), which is pro-
duced by the Federal Reserve Board every three years. The
earliest SCF data is from 1980, and the latest from 2010. Broad
variables covered by the SCF are income, assets, and debt of
families. Because the SCF collects data on assets that are heav-
ily concentrated among the richest families, in order to be rep-
resentative and meet validity standards, the sample is designed
to capture sufficient numbers of upper-income families.
The dollar value of debt holdings, in real terms, has risen
sharply. Table 4.1 presents the median value of debt holdings by
debt categories for three years: 1995, 2007, and 2010. (The SCF
data are only collected every three years). Note that families
with no debt in a given category are excluded from the calcula-
tion of the medians. Residential mortgage debt dwarfs the other
three categories in size. It includes not just first mortgages on a
family’s primary residence but also second mortgages, refinanc-
ing, home equity loans, and vacation homes. The median family
mortgage debt in constant dollars rose from $126,000 in 1995 to
$217,000 in 2007. Most of that rise occurred from 2001 to 2004,
the period when aggregate mortgage liabilities increased by
$2.5 trillion (see Table 4.7). The 1995 to 2007 increase in median
real mortgage debt, 73 percent, is many times larger than the
increase in median household real income over that period—
only 8 percent.3 It has diminished somewhat since its peak of
$225,000 in 2004, reflecting the bursting of the housing price
bubble and the resulting reduction of mortgage lending.
 38 Consumers’ Shift to Debt

Table 4.1. Median Value of Family Debt Holdings

(2010 $ thousands)
Primary
residence Credit card
and other and lines
residential of credit
mortgage other than Installment Any
Year debt residential loansa Otherb debt
1995 125.7 6.7 16.4 2.7 29.2
2007 216.9 7.1 13.6 5.2 70.6
2010 207.6 8.6 12.6 4.5 70.7
Percent change
1995–2007 72.6% 6.0% –17.1% 92.6% 141.8%
a
Includes education, vehicle, and other.
b
Includes cash value of life insurance loans, pension account loans, margin account loans,
and other miscellaneous loans.
Source: Federal Reserve Board (2012).

The largest part of consumer debt is mortgages, as noted

above. The Federal Reserve publishes two measures of house-
hold debt burden: the Debt Service Ratio (DSR) and the Finan-
cial Obligation Ratio (FOR). The DSR measures debt payments
as a share of disposable income for all households. The FOR
measures mortgage debt, home insurance, property tax, and
consumer debt as well as automobile leases as a percentage of
disposable income for homeowners only. Both are shown for
selected years in Table 4.2. Although both indices are higher
in 1995 than in 1980, the individual years’ data reveal no trend.
In some years before 1995 the indices are higher, and in some
years lower. That changes post 1995 when most year-to-year
changes are positive. The 2007 values shown are record highs
Consumers’ Shift to Debt 39

Table 4.2. Household Debt Burden

First Quarter of Year DSR FOR
1980 10.62 15.43
1995 10.90 16.22
2007 12.88 17.70
2014 9.96 15.34
Source: Federal Reserve Board (2014b).

for each index. But following the financial crash and Great Re-
cession, both measures were lower than their 1995 values.

Household Debt by Income Group

Growth in median family residential mortgage debt among
different quintiles and deciles of the income distribution over
the 1995 to 2007 period was substantial and broadly similar, as
reported by the Survey of Consumer Finances and presented
in Table 4.3. Shown is the largest debt category, mortgages on
primary residences, which increased substantially—from 41
percent in quintile two to 90 percent in the second-from-the-
top decile (80.0–89.9).
Based on data from the SCF, Figure 4.2 shows debt-to-
income ratios for the top 5 percent of the income distribution
and the bottom 95 percent. The authors note about the figure,

In 1983 the top income group is somewhat more

indebted than the bottom group, with a gap of
around 20 percentage points. In 2007, the situation
was dramatically reversed. The debt-to-income
ratio of the bottom group, at 147.3% compared to
an initial value of 62.3%, was now more than twice
 40 Consumers’ Shift to Debt

Table 4.3. Median Family Residential Mortgage Debt by

Quintiles, 1995, 2007, and 2010 (2010 $ thousands)
Percent change
Quintile 1995 2007 2010 1995–2007 2007–10
Quintile 1 25.0 41.9 54.6 67.6% 30.3%
Quintile 2 37.9 53.4 65.5 40.9% 22.7%
Quintile 3 50.0 92.9 90.0 85.8% –3.1%
Quintile 4 77.1 120.5 116.6 56.3% –3.2%
2nd top decile, 80–89.9 90.6 171.8 158.0 89.6% –8.0%
Top decile, 90–100 121.7 210.6 241.0 73.0% 14.4%
Source: Federal Reserve Board (2012), Tables 13 95 through 13 07.

as high as that of the top group. Between 1983 and

2007, the debt to income ratio of the bottom group
therefore more than doubled while the ratio of the
top group remained fluctuating around 60%.4

But the figure also shows that the huge run-up of the debt-to-
income ratio for the bottom 95 percent occurred in the period
after 2001. The authors infer from Figure 4.2 that it is part of
the explanation for why consumption inequality has not in-
creased nearly as much as income inequality. That is, the bot-
tom 95 percent of the wealth distribution has taken on much
more debt in order to maintain their consumption.

Subprime Mortgages
One of the direct causes of the financial crash was the increased
volume of subprime mortgages that were bundled into securi-
ties and sold to investors. The collapse of prices for those secu-
ritized debt obligations touched off the financial crisis. There
Consumers’ Shift to Debt 41

Figure 4.2. Debt-to-income ratios, 1983–2007. Source: Reprinted

with permission from Michael Kumhof, Romain Ranciere, and
Pablo Winant (2013), “Inequality, Leverage and Crises: The Case
of Endogenous Default,” International Monetary Fund Working
Paper, WP/13/249, November, p. 38.

was a stunning rise of subprime mortgage originations from

slightly over 400,000 in 1999 to over two million in 2005, the
peak year.5 The total of originations is split between refinanc-
ings and purchases. In every year from 1999 through 2006, the
refinancing with subprime mortgages is 60 to 75 percent of total
originations. A major purpose of mortgage refinancing is to take
out cash. As shown in Figure 4.2, the ratio of debt to income for
the bottom 95 percent of the wealth distribution shot up sharply,
from 90 percent in 2001 to near 150 percent in 2007. Some part
of that rise reflects the fivefold increase in subprime mortgages.
That rise has not been geographically concentrated so
much as credit score concentrated. In a paper that splits a
huge sample of zip codes into quartiles based on credit scores,
the bottom quartile is labeled subprime—that is, it has the
highest share of households with credit scores of 660 or less.
The authors found that the mortgage default rate in 2006 of
subprime zip codes was three times higher than the rate in
42 Consumers’ Shift to Debt

prime zip codes. Additionally, mortgage credit growth in sub-

prime zip codes (the quartile with the highest share of credit
scores below 660) was two times greater than mortgage credit
growth in prime zip codes (the quartile with the lowest share
of credit scores below 660). Subprime zip codes are not region-
ally concentrated but rather are present in most metropolitan
areas. Correlation between mortgage credit growth and in-
come growth was negative in the 2002–05 period, whereas it
was positive in the prior fifteen years. Before the expansion in
subprime mortgage lending, applications for mortgage credit
from subprime zip codes were more likely to be denied than
those from quartiles with higher credit scores. However, from
2002 to 2005, denial rates for subprime zip codes fall dispro-
portionately. An examination of house price indices by zip
code shows that house price gains for subprime zip codes in a
county are greater than gains for non-subprime zip codes.6
Christopher Mayer and Karen Pence established that
the use of subprime mortgages is not ubiquitous over states
or metropolitan areas, or in demographic characteristics of
borrowers. Looking at the data by state, they showed that sub-
prime originations as a percentage of all originations in 2005,
the peak year for subprime originations, were 19 percent for the
nation. The four states with the highest concentration of sub-
prime and Alt-A mortgages as reported by the New York Fed-
eral Reserve Bank in 2011 had subprime originations in 2005 at
or above the national average: Nevada, 25 percent; Florida, 24
percent; Arizona, 21 percent; and California, 19 percent. The
state shares of subprime originations range from the high of
25 percent in Nevada to the low of 8 percent in West Virginia.
The same data for metropolitan areas in 2005 show much
greater variation. The 2005 average for 107 large metro areas is
20 percent. Memphis and Bakersfield, California, are tied for
Consumers’ Shift to Debt 43

first place with 34 percent. Madison, Wisconsin, is in last place

with only 9 percent. Anecdotal comparisons that spring to
mind in thinking of those three places are that both Memphis
and Bakersfield have high shares of minority population, low
median incomes, and low levels of educational attainment.
Madison is the opposite on all three measures. The top ten
places in share of subprime mortgage originations include the
most likely suspects because of large house price increases and
construction booms: Las Vegas, Miami, and Houston, as well
as Detroit, which had well under average price appreciation
and certainly no construction boom. Four of the top ten are
mid-size California metros far from the coast, some of which
had construction booms, and all of which have large Hispanic
population shares and low educational attainment.7
Mayer and Pence found that subprime lending is not
elevated only in metros with strong housing price surges.
They cite New York and Boston as places with relatively high
house price appreciation, but not much in the way of subprime
mortgages. But subprime lending surged in depressed housing
markets of the Midwest because conventional lending had di-
minished. Looking at neighborhoods with zip code data, they
found that “subprime mortgages are concentrated in locations
with high proportions of black and Hispanic residents, even
controlling for the income and credit scores of these zip codes.”8

Household Debt, Bankruptcies, and

House Prices: State Data
The strong rise of household indebtedness documented here
was not underpinned by a strong rise of household income. It
was underpinned by the housing price bubble and supply-side
factors that increased the availability of credit—subprime mort-
44 Consumers’ Shift to Debt

gage lending, low interest rates, relaxed credit standards, and

financial deregulation that made residential property more liq-
uid. Both the explosion of debt and fallout from the collapse of
the housing price bubble were mostly ubiquitous across states,
although four states stand out for larger gains and more severe
declines: Arizona, California, Nevada, and Florida. Those are
the states that had the highest concentration of toxic real es-
tate assets: subprime and Alt-A mortgages. Although the four
states’ share of all housing units in the nation is 20 percent,
they account for 28 percent of all subprime mortgages and
45 percent of all Alt-A mortgages.9 The top panel of Table 4.4
summarizes debt, mortgage debt, house prices, and median
household income for all states and the District of Columbia
from 1999 to 2007 and then to 2012. The bottom panel repeats
those measures for the four more volatile states named above.
For all states in the period 1999–2007, per capita total debt,
mortgage debt, and the house price index rose strongly, while
median household income hardly changed, gaining less than
1 percent. But then in the following years, 2007–12, they all
moved in the opposite direction, including median household
income, which fell almost 6 percent. Large as those swings are,
they are more extreme in the four states with high concen-
trations of subprime and Alt-A mortgages. Note that the data
in Table 4.4 are simple averages for states. Excluding the four
states from the calculations for the top panel, the picture does
not change much. That is, debt and house prices rise some-
what less from 1999 to 2007, and decline somewhat less from
2007 to 2012. The point is that the huge run-up in debt and the
housing price bubble cannot be attributed only to toxic mort-
gages in the four states with high concentrations of such loans.
The sluggish growth of household income by state com-
pared with soaring per capita household debt and house prices
 Consumers’ Shift to Debt 45

Table 4.4. Household Debt Compared with House Prices

and Median Income, 1999–2012
Median
Total debt Mortgage debt household
per capita per capita House income
(2008 $ (2008 $ price index (2008 $
Year thousands) thousands) (1991 Q1 = 100) thousands)
Simple averages for 50 states and District of Columbia
1999 28.7 19.4 138.4 52.7
2007 48.6 35.5 228.5 53.0
2011 46.3 34.2 197.3 48.3
2012 44.3 33.2 189.5 48.2
Simple average for four statesa
1999 34.8 25.9 126.9 51.7
2007 73.1 57.2 257.4 53.3
2011 55.8 42.1 169.0 48.1
2012 51.1 39.3 159.8 45.1
a
Arizona, California, Florida, and Nevada.
Source: Household debt per capita from Federal Reserve Bank of New York (2013).
Median household income in 2010 dollars from U.S. Census Bureau (n.d.), “Current
Population Survey.” House price index from Federal Housing Finance Agency (2014).

from 1999 to 2007, presented in Table 4.4, points to a precari-

ous financial condition for many households. A surge of per-
sonal bankruptcies post 2000 prompted a strong pro-­creditor
reaction by Congress so that upward trends were abruptly
halted in 2006. The Bankruptcy Abuse Prevention and Con-
sumer Protection Act enacted in April 2005 and effective in
October of that year briefly slowed the rise. The purpose of the
law was to curb the rise of bankruptcies, both business and
non-business, by raising the barriers to filings. The declines
46 Consumers’ Shift to Debt

from 2005 to 2006 are around 75 percent. However, the earlier

upward surge was resumed in 2007, so that by 2010 per capita
bankruptcies were well above their 1999 levels.10 The intention
of Congress was apparently overtaken by overwhelming fi-
nancial hardship post 2006. The act of Congress was inspired
more by the perception of bankruptcy abuse than by a desire
for consumer protection. The characterization of personal
bankrupts as deadbeats was probably important for passage
of the bill. But that characterization was false. A careful study
of non-business bankruptcies found that “Bankrupts’ incomes
are low at the time of filing, the consequence of about two-
thirds of the families reporting a job loss, failure of a small
business, a cutback in hours worked, or some other income
interruption. But when they are measured by the enduring
criteria of education, occupation, and home ownership, about
90% of the debtors qualify as solidly within the middle class.”11

R E G R E S SIO N R E SU LT S L I N K I N G R I SI N G HO U SE HO L D
D E BT T O R I SI N G I N C OM E I N E QUA L I T Y: S TAT E DATA

A central argument of this book is that the huge run-up in

household debt that was one of the major causes of the finan-
cial crisis, and the Great Recession was itself in part a mani-
festation of the long rise of income inequality. That possible
link of rising income inequality to rising household debt has
been explored econometrically using the New York Federal
Reserve Bank data on household debt by state, 1999 to 2010.
The key advantage of that data is that it provides time-series
data on household debt for all states, enabling the estimation
of panel data regression equations. Panel data equations are
less fraught with the estimation problems of time-series data,
and they have much larger sample sizes. Instead of having a
Consumers’ Shift to Debt 47

sample limited to the number of years, a panel regression sam-

ple size is equal to the number of time periods, twelve years
in this case, multiplied by the number of entities, fifty states
and the District of Columbia, or 612. Panel regression equa-
tions have been estimated to measure the connection, if any,
between rising income inequality and rising household debt.
The details about those estimated equations are presented in
an appendix to this chapter. Here the substantive results are
presented in a non-technical manner.
Table 4.5 shows the estimated elasticities from the panel
regression equations presented and explained in the appendix.
These elasticities are the percent change in per capita house-
hold debt with respect to a 1 percent change in the variable
named, holding all else equal. The key variable is the income
share of the bottom 80 percent by state and year, the first elas-
ticity shown. Its estimated value is –0.2. That means that a 1
percent fall of the state income share (rising income inequal-
ity) for the bottom 80 percent of households is expected to
produce a rise of 0.2 percent in state per capita household debt
three years later, other things being equal. In other words, ris-
ing inequality is accompanied by rising household debt with a
lag of three years. The estimated elasticity is highly significant,
as noted in the table.
The other elasticities have the expected signs—negative
for the national unemployment rate and positive for the state
house price index and state median household income. All
are highly significant except for the state median household
income variable. Note that the largest effect on per capita
household debt is the state house price index, with an elastic-
ity of +0.3. The interpretation of that result is that rising house
prices induced a perception of enhanced wealth, which moti-
vated households to borrow more.
48 Consumers’ Shift to Debt

Table 4.5. Elasticities of Per Capita

Household Debt with Respect to Income
Inequality and Other Measures by State
and Year
Variable Elasticity
State income share bottom 80%, –0.2b
lagged three years
U.S. unemployment rate –0.1a
State house price index +0.3a
State median household income +0.1c
a
significant at .001 level
b
significant at .01 level
c
significant at .05 level
Source: Author’s calculations.

These results, elaborated in the appendix, of course do

not prove a causal link from a diminishing state share of in-
come received by the four lower income quintiles (rising in-
come inequality) to rising household debt, but they do sup-
port the argument of a causal link. If the equations showed a
positive effect or no significant effect of rising income inequal-
ity on debt, then there would be little inducement to further
work supporting that claim. But these results beg for stronger
support or refutation.
Stronger support is provided in an article by Robert
Hockett and Daniel Dillon. Their careful econometric analysis
shows that a rise of the very top income share, the share of the
0.1 percent, is followed two years later by a rise in all house-
hold debt per capita, ceteris paribus. The same holds true on
the downside. A decline of the income share of the top 0.1
percent is associated with a drop in household debt two years
Consumers’ Shift to Debt 49

later. The authors argue: “This positive feedback loop presents

evidence of the hypothesized relationship between inequality
and debt, namely that as the wealthy amass more of the aggre-
gate income, the average household ramps up its borrowing to
maintain accustomed living standards.”12
In another paper that covers some of the same material
on inequality and household debt as this book, the authors es-
timate time-series regression equations relating measures of
debt to measures of inequality and other variables. Estimated
coefficients on their inequality variables are not statistically
significant in about half of the cases. That may reflect a flaw
in the choice of some inequality variables—namely, the labor
share of national income. As noted in Chapter II, the top 1
percent’s share of labor income has doubled since 1969. Thus,
labor share does not fully capture the rise of inequality (see
Table 2.2). Also, the authors use quarterly data with a sample
size of only 93, whereas the sample size here is 407.13

R E G R E S SIO N R E SU LT S L I N K I N G R I SI N G
HOU SE HO L D D E B T T O R I SI N G I N C OM E I N E Q UA L I T Y:
NAT IO NA L DATA

This second use of regression to estimate association between

income inequality and household debt shifts from a state to a
national focus using the Consumer Expenditure Survey (CEX)
data by income quintiles for the years 1984 through 2007. The
interview survey data from the CEX of the Bureau of Labor
Statistics (BLS) is the only annual survey that provides after-
tax income by income class as well as change in liabilities by
income class. The observations in that analysis are average
dollar amounts by quintile and year, not individual consumer
units.
50 Consumers’ Shift to Debt

In addition to covering numerous out-of-pocket consumer

expenditure categories, the CEX also collects before and after-tax
income data, household characteristics, as well as some finan-
cial data. The consumer units are subdivided into five income
quintiles. For most of the years of the continuous annual sur-
vey, 1984–2003, the quintile data excluded consumer units that
did not report their income. The numbers of such consumer
units were shown in the annual reports. They amounted to
about 16 percent of all consumer units. Beginning in 2004,
the BLS imputed incomes for non-reporting households, and
so the distribution by quintile included all consumer units.
Before then, CEX reported average income was well below
the average income reported on the Census Bureau’s Current
Population Survey (CPS). The effect of imputation was to raise
substantially the ratio of CEX pre-tax income to CPS pre-tax
income, making the two series closer. In 2003, before income
imputation, CEX income was 75 percent of CPS income. In
2004, after income imputation, that ratio rose to 91 percent.14
CEX data for previous years were not revised, and so the CEX
for 2004 forward is not quite comparable to earlier years.
As with the state regression analysis, the dependent var-
iable is a measure of household debt by years and by income
quintile. The national panel regression equation by quintile
and year, presented in the appendix, uses the Theil index of
inequality (a rise means increased income inequality). That is
the variable of central interest.

Economic Insecurity
All of the evidence presented in this chapter on the increasing
amounts of debt taken on by households suggests that they
might have been feeling more anxious about their economic
Consumers’ Shift to Debt 51

situation than in the past. There is a statistical series that at-

tempts to capture not the feeling of anxiety but rather the ob-
jective fact of financial insecurity. The 2010 Economic Security
Index (ESI) is designed to capture that. The measure is based
on panel data and at present covers most of the years from
1986 through 2010. The major data source used in construct-
ing the index is the Survey of Income and Program Participa-
tion (SIPP) of the Census Bureau. Also used are the BLS Con-
sumer Expenditure Survey and the University of Michigan’s
Panel Study of Income Dynamics.

The ESI is designed to fill a gap in existing theo-

retical and empirical analyses of economic secu-
rity grounded in panel data on economic status.
Prior research has focused primarily on individual
sources of insecurity, such as earnings volatility
and the incidence of large medical expenditures.
The ESI by contrast represents the first attempt
to incorporate several key influences—income
declines, medical spending shocks, and financial
wealth buffers—into a single unified measure.15

For each household i in the large sample, a dichotomous var-

iable ri is calculated. If the year-to-year loss of discretionary
income is 25 percent or more, ri = 1. Discretionary income is
total household income less out-of-pocket medical expenses
and debt service. Sufficient liquid assets for a household can
offset some or all the decline in discretionary income. The ESI
in any year is the weighted sum of the r scores over the sample
size.
Figure 4.3 presents the ESI for selected years (a high ESI
indicates greater economic insecurity) for all households. Al-
52 Consumers’ Shift to Debt

Figure 4.3. Economic Security Index (ESI), all households, 1986–

2010. Source: Hacker (2015).

though there are ups and downs associated with cyclical in-
fluences, the ESI has a decidedly upward trend, rising from
around 14 percent in 1986 to over 20 percent in the most re-
cent years, 2008–10.
The ESI has been calculated for individual income quin-
tiles as well as all households. The quintile data are not pub-
lished, but they were provided by one of the authors, and these
data are presented in Table 4.6.16 The ESI tends to be smaller
for higher income quintiles, except that the fifth quintile is
above (higher economic insecurity) the fourth in some years.
It appears that the ESI captures something not reflected in the
common objective measures of hardship: the unemployment
and poverty rates. Note that although unemployment and
poverty were lower in 2004 than in 1991, the ESI was much
higher in 2004 for the middle three quintiles.
Consumers’ Shift to Debt 53

Table 4.6. Economic Security Index, All Households

and Quintiles, Selected Years
1991 2004 2009
All households 16.90% 17.50% 20.50%
Quintile 1 26.30% 20.80% 29.50%
Quintile 2 17.60% 21.20% 22.60%
Quintile 3 14.30% 17.30% 18.90%
Quintile 4 12.50% 15.40% 16.20%
Quintile 5 13.50% 12.90% 15.50%
Unemployment rate 6.80% 5.50% 9.30%
Poverty rate 14.20% 12.70% 14.30%
Source: Author’s compilation of ESI by quintile from Hacker (2015). Unem-
ployment rate from Council of Economic Advisers (2013). Poverty rate from
DeNavas-Walt, Proctor, and Smith (2011).

Why Household Debt Soared

The argument that the rise of consumer indebtedness is linked
to income inequality is well stated by Stiglitz.

The negative impact of stagnant real incomes and

rising income inequality on aggregate demand was
largely offset by financial innovation in risk man-
agement and lax monetary policy that increased
the ability of households to finance consumption
by borrowing, especially in the United States. . . .
But increasing household indebtedness was not
sustainable. Or rather what was perceived to be
sustainable was dependent on artificially inflated
asset prices that created the illusion that household
wealth was increasing at a faster pace than their
debt. The support for the bubble thus depended on
54 Consumers’ Shift to Debt

expansionary monetary policy together with finan-

cial sector innovation leading to ever-­increasing
asset prices that allowed households virtually un-
limited access to credit.17

Christopher Carroll, Misuzu Otsuka, and Jiri Slacalek argue

that the use of enhanced housing values to support consump-
tion is a long-standing pattern, and they find a substantial ef-
fect that may or may not be as large as the effect of enhanced
financial wealth. They do not consider any role for rising in-
come inequality. Other writers point to extreme leveraging by
consumers as contributing to the financial crash but ascribe
no role to stagnant incomes and increasing income inequal-
ity. Their listed causes for that leveraging are all on the supply
side: financial innovation, rising house prices, subprime mort-
gage securitization, low interest rates, and massive inflows of
foreign credit.18
The absolute and relative rise in household debt doc­
umented in this chapter was facilitated by a number of fac-
tors. One was the advent of subprime and Alt-A mortgages
aggressively pushed by mortgage lenders who understood that
investment banks were eager to buy them and bundle them into
bonds—securitized debt obligations. Another was the develop-
ment of mortgage refinancing, home equity loans, and home eq-
uity lines of credit, financial instruments not widely available to
households until the 1990s. “Two pieces of legislation, the Mon-
etary Control Act of 1980 and the Garn-St Germain Act of 1982,
unlocked this wealth [residential property worth 106 percent
of GDP]. The new laws made it easier for households to refi-
nance their mortgages and borrow against the value of their
homes.” Also, Congress was pushing g­ overnment-sponsored
Fannie Mae and Freddie Mac to boost mortgage lending to
Consumers’ Shift to Debt 55

underserved households.19 Finally, unusually low interest rates

following the onset of recession in 2000–2001 and after the
World Trade Center attack in 2001 were continued until 2005.
As Stiglitz noted about that period leading up to the financial
crash and Great Recession, “Greenspan lowered interest rates
flooding the market with liquidity. With so much excess ca-
pacity in the economy, not surprisingly, the lower interest rates
did not lead to more investment in plant and equipment. They
worked—but only by replacing the tech bubble with a hous-
ing bubble, which supported a consumption and real estate
boom.”20 All of those supply-side factors raised the availabil-
ity of credit. But increased supply does not ensure increased
demand for credit by households. The rise of demand was for
two purposes: to buy homes and to take out cash for maintain-
ing consumption.
Borrowing to buy an asset with an expected growth in
value well above the cost of borrowing is not irrational. As the
data above show, most of their borrowing and most of their
outstanding debt has been for residential property, the dom-
inant part (60 percent in 2007) of their non-financial assets.
Consumers knew house prices were rising rapidly from 1995
to 2006 while mortgage interest rates were trending down. The
rate of house price appreciation, 10 percent per year during
1997–2006, was well above the average interest rate on a con-
ventional mortgage for that period, 6.8 percent.21 Thus it was
wealth enhancing to borrow to buy a house as long as house
prices continued to rise.
The ten-year rapid rise of house prices lured consum-
ers into thinking that the future would be similar to the past.
A Case-Shiller survey of home buyers in the spring of 2005
revealed that the median expectation of house price appreci-
ation for the next ten years was 7 percent annually. In fact,
56 Consumers’ Shift to Debt

house prices nationally declined 25 percent from the spring of

2005, when the Case-Shiller survey was taken, to the spring of
2009.22
Not all the debt that shows up as mortgage debt, by far
the largest debt category for households (see Table 4.1), repre-
sents borrowing to purchase a home. Many households tapped
that increased wealth with second or junior mortgages, cash-
out refinancing, and home equity loans or lines of credit. In-
deed, an analysis by two authors of what existing homeowners
did in reaction to their increased home values argues that ris-
ing home values lured consumers into taking on more debt.
The effect they calculate is large. They find “an elasticity of
borrowing with respect to increased home equity of 0.60. Al-
ternatively we find that households borrow 25 to 30 cents on
each additional dollar of home equity from 2002 to 2006.”23
Recall that, as mentioned above, rising house prices have the
largest positive impact on per capita state debt.
Table 4.7 shows that the rise in such cash extraction from
residential property accounted for about one-half of the spec-
tacular rise in mortgage debt from 2004 to 2007. But with the
bursting of the housing price bubble in 2005–06, those credit
sources were shut down for most households. The burden of
mortgage debt as a percentage of the market value of house-
holds’ real estate jumped from just over 40 percent in 2004 to
50 percent in 2007 and 60 percent in 2010. In 2007 home eq-
uity loans outstanding were $1.1 trillion, compared with $0.4
trillion in 2000. Note that the changes in mortgage liabilities
and cash-out from mortgages became large negative amounts
in 2010. That sharp turnaround does not simply reflect fore-
closures and write-offs of household liabilities by financial in-
stitutions after the financial crash in 2008. A Federal Reserve
 Consumers’ Shift to Debt 57

Table 4.7. Change in Mortgage Liabilities and Change in

Cash-Out from Mortgaged Properties, 1992–2010
Change in Change in Change in Residential
residential cash-out from cash- mortgage
mortgage residential out as % of liabilities as %
liabilities mortgages change in of market
Year ($ billions) ($ billions) mortgages value
1992 580 130 22.3% 39.1%
1995 479 236 49.4% 41.2%
1998 735 478 65.0% 41.8%
2001 1,267 324 25.5% 39.2%
2004 2,535 1,286 50.7% 41.5%
2007 2,711 1,351 49.8% 50.7%
2010 –616 –1,077 174.8% 60.0%
Source: Change in residential mortgage liabilities, three-year intervals, calculated from
Federal Reserve Board (2014a), Table L.100; cash-out calculated from applying SCF per-
cent shares, Table 14, to aggregate mortgage liabilities. Liabilities as a percent of market
value calculated from Federal Reserve Board (2014a), Tables B.100 and L.100.

Board analysis shows that households have voluntarily paid down

their debt.24
What did households do with all that borrowed money?
The shifting relative distribution of households’ consumption
provides some insight into what they have been spending on,
as shown in Table 4.8. That table shows relative expenditure
shares for selected categories of consumption for all house-
holds and by income quintile, 1984 to 2007. The first column
presents the relative shares for all households for selected years.
The first category includes food, apparel, and transportation.
It fell almost seven percentage points for all households from
1984 to 2007. Each quintile showed similar declines. The three
 Table 4.8. Expenditure Shares by Selected Categories and Quintile,
1984 and 2007
1984
Quintile Quintile Quintile Quintile Quintile
All 1 2 3 4 5
Food, apparel, and 40.6% 39.6% 42.6% 42.7% 41.1% 38.8%
transportation
Shelter 15.5% 18.9% 15.8% 15.5% 14.8% 15.0%
Health care 4.8% 5.9% 6.8% 5.7% 4.3% 3.5%
Education 1.4% 3.1% 1.1% 1.0% 1.0% 1.3%
Sum of three 21.7% 27.9% 23.7% 22.2% 20.1% 19.8%
2007
Food, apparel, and 33.8% 34.4% 35.4% 36.0% 35.6% 31.1%
transportation
Shelter 20.2% 24.2% 21.5% 19.9% 19.5% 19.5%
Health care 5.7% 7.2% 7.9% 6.7% 5.7% 4.4%
Education 1.9% 3.0% 1.1% 1.3% 1.2% 2.6%
Sum of three 27.8% 34.4% 30.5% 27.9% 26.4% 26.5%
Source: Bureau of Labor Statistics (n.d.), “Consumer Expenditure Survey, 1984–2010.”
Consumers’ Shift to Debt 59

categories that have had increased relative shares—­shelter,

health care, and education—are shown separately. Shelter in-
cludes the major costs of owning (mortgage interest, property
taxes, insurance, and maintenance) as well as renting. Not in-
cluded are utilities, appliances, and furniture. Health care in-
cludes only out-of-pocket spending, not expenses covered by
private insurance, Medicare, or Medicaid.
Note that in the first year, 1984, the share spent on food,
apparel, and transportation, 40.6 percent, was almost two
times greater than the share spent on the three other catego-
ries, 21.7 percent. But gradually the share spent on food, ap-
parel, and transportation dropped, while the share spent on
shelter, health care, and education rose. That same pattern
of increased shares for shelter and health care is repeated in
all five quintiles. The share spent on shelter, the largest of the
three, was over 15 percent for all households in 1984. By 2007
the share spent on shelter by all households was above 20 per-
cent, ranging from a high of 24 percent in the first quintile to
about 20 percent in the other four quintiles. The share spent
on education in the first two quintiles did not increase from
1984 to 2007, but it doubled in the highest income quintile.
Aside from the much-noted increased demand for shel-
ter, health care, and education over the past decades, their
prices have risen much faster than inflation, while the prices
for food, apparel, and transportation have risen at or below
inflation. Table 4.9 presents the percentage growth of the
Consumer Price Index (CPI) for those categories as well as all
items from 1982–84 to 2007. The all items index rose 107 per-
cent, a doubling of the cost of living in about twenty-five years.
The shelter index rose more—140 percent—and the health care
index jumped over 250 percent. Largest of all are the increases
in the major components of the education price index—­
60 Consumers’ Shift to Debt

Table 4.9. Percentage Rise of Consumer Price Index, All

Items and Selected Categories, 1982–84 to 2007
Percentage change, 1982–84
Category to 2007
All items 107.3%

Food and beverages 103.3%

Apparel 19.0%
Transportation 84.7%

Shelter 140.6%
Medical care 251.1%
Educationa
Tuition, school fees, and childcare 394.1%
Education books and supplies 320.4%
a
Education CPI now published on a December 1997 base = 100, and so it is not
comparable with these percent changes.
Source: Bureau of Labor Statistics (n.d.), “Consumer Price Index,” 2013.

tuition, fees, and child care, up almost 400 percent, and books
and supplies, up 320 percent.
Why did the prices for shelter, education, and medical
care rise so much more than the overall price level? One rea-
son is that they are not in competition with cheap imports,
unlike electronics and apparel. But the more salient reason for
this analysis is that the demand of the top 10 percent for those
categories must have expanded strongly with their expanded
share of income. Income elasticity of demand for them most
certainly is plus one or higher. In Figure 4.4 the trends in those
CPI components are compared with the trend of income share
for the top 10 percent, 1993 through 2012 or 2013. During that
period the income share of the top 10 percent rose from just
under 40 percent to just over 48 percent. That gain must have
Consumers’ Shift to Debt 61

Figure 4.4. Top 10 percent income share compared with selected

CPI categories, 1993–2013. Sources: Top 10 percent income share
from Alvaredo et al. (n.d.). CPI from Bureau of Labor Statistics
(n.d.), “Consumer Price Index.”

strongly boosted demand for shelter, education, and medical

care among the top 10 percent, thus putting upward pressure
on prices for those items. Figure 4.4 illustrates how similar the
income share trend is to the price trends. Of course that does
not prove that rising income inequality at the top pulled up
those prices, but the visual correlation is suggestive of such an
effect.
Those categories—shelter, health care, and education—
are taking a much bigger bite out of households’ spending
than in the past, and they are not expenditures that can be
postponed, such as replacing the car or taking a vacation trip.
The immediacy of such demands, combined with decades of
stagnant household incomes for most, must have made the
62 Consumers’ Shift to Debt

easy availability of credit to be an almost irresistible solution

to the problem of households’ squeezed budgets. This analy-
sis of what happened to households’ relative consumption is
another support for the argument that stagnant incomes and
rising income inequality led to an explosion of debt.
Atif Mian and Amir Sufi note, “We find little evidence
that borrowing in response to increased house prices is used
to purchase new homes or investment properties. We also
find no evidence that home equity-based borrowing is used
to pay down credit card balances. . . . We find that a total of
$1.45 trillion of the rise in household debt from 2002 to 2006
is attributable to existing homeowners borrowing against the
increased value of their homes. That translates to 2.8% of GDP
per year.”25
The data presented in this chapter show that the money
taken out from appreciating housing was not used to pay
down debt because indebtedness rose. Rather it was used to
support consumption in the face of stagnant income. “Money
extracted from increased home equity is not used to purchase
new real estate or pay down high credit card balances, which
suggests that borrowed funds may be used for real outlays (i.e.,
consumption or home improvement).”26 That point has been
made by others as well. “Where did the borrowings go? Some
have asserted that it went to investments in stocks. However, if
this were the case, then stocks as a share of total assets would
have increased over this period, which it did not (it fell from 13
to 7 percent between 2001 and 2007). . . . Instead middle class
households experiencing stagnating incomes, expanded their
debt almost exclusively in order to finance consumption ex-
penditures.”27 Edward Wolff asks if debt was increased in order
to support normal consumption or to expand consumption.
Analyzing the Consumer Expenditure Survey data over that
Consumers’ Shift to Debt 63

period, Wolff concludes, “Thus the CEX data, like the NIPA
data, show no acceleration in consumer spending during the
debt splurge of the 2000s. As a result it can be concluded that
the debt build-up of the 2000s went for normal consumption,
not enhanced consumption.”28
This chapter on consumers’ debt and Chapter II on in-
come distribution lay out the facts of stagnant incomes and
rising income inequality and the unusually large increase of
consumer debt beginning in the mid-1990s that culminated in
the financial crash and the Great Recession. Econometric evi-
dence presented in this chapter links rising income inequality
to the expansion of household debt, an expansion that was un-
sustainable. If the Great Recession was in part caused by the
big rise in household debt, and that rise in debt was in part
the result of stagnant incomes and increased income inequal-
ity for three decades, then surely increasing income inequality
matters. It matters for understanding how to prevent another
big recession. It matters for understanding why the economic
recovery has been so sluggish. What is required is a theory to
link long-term rising income inequality to a huge rise of un-
sustainable household debt. The next chapter, on consumption
theory, addresses that issue.

Appendix
PA N E L R E G R E S SIO N A NA LYSI S O F STAT E DATA

The left-hand or dependent variable in every estimated equa-

tion is a measure of per capita household debt by state and
year. The key right-hand variable in every equation is the share
of aggregate income of quintiles one through four by state and
year. The central importance for this study of the estimated
panel regression equations is that they support the hypothesis
64 Consumers’ Shift to Debt

of a positive link between rising income inequality and rising

household debt.
Descriptive statistics, means and standard deviations, for
all of the variables used are shown in Table 4.10 for three years:
1999, 2007, and 2010. The choice of years is not arbitrary. The
first year of the household debt series by state is 1999. The last
year before the financial crisis and Great Recession is 2007, and
2010 is the last available year. Most of the variables presented are
covered for the same period in Table 4.4. One new variable is the
income share of quintiles one through four. Another new varia-
ble is the U.S. unemployment rate. That is a national variable that
does not vary by state, and hence its standard deviation for each
year shown is zero. Note that the income share shows a slight
decline of 0.9 of a percentage point from 1999 to 2010—in other
words rising income inequality. But in the previous eleven-year
period, 1988–99, that share of income to the bottom four quin-
tiles dropped more than three percentage points. Nonetheless,
the moderate drop of income share (rising inequality) since 1999
appears to have a positive effect on per capita debt.
Three debt panel regression equations are estimated. The
left-hand variables are total debt, mortgage debt, and other
debt per capita by state and year. One of the assumptions of
regression using time-series data is that the left-hand varia-
ble Yt is stationary. That is, it does not have a stochastic trend.
That assumption is rarely satisfied with economic data. The
common way to avoid that problem is to express Yt in first
differences—that is, (Yt − Yt−1 ). There may be important deter-
minants of the Y variables that are not included among the
right-hand variables. Such omissions result in biased estimates
of regression coefficients. In the case of panel regression, the
most common omitted variables are place fixed effects that
vary by entity (states in this case) but not by time. The solution
 Consumers’ Shift to Debt 65

Table 4.10. Descriptive Statistics, 1999, 2007, and 2010

1999 2007 2010
Variables Mean SD Mean SD Mean SD
Real total household 28.7 7.4 48.6 15.3 44.7 13.0
 debt per capita (2008 $
thousands)
Real household mortgage 19.4 7.1 35.5 15.8 32.7 11.6
 debt per capita (2008 $
thousands)
Real household other 9.3 0.9 13.1 1.7 12.0 1.6
 debt per capita (2008 $
thousands)
Real household median 52.7 7.6 53.0 7.7 50.1 7.5
 income (2008 $ thou-
sands)
House price index 138.4 21.0 228.5 42.6 200.9 37.8
(1991 Q1 = 100)
Income shares quintiles 1 51.9 2.5 51.5 2.6 51.0 2.4
through 4
U.S. unemployment rate 4.2 0.0 4.6 0.0 9.6 0.0
Source: Household debt per capita from Federal Reserve Bank of New York (2011). Author’s
compilation of bankruptcy filings from U.S. Courts (n.d.). Median household income in 2010
dollars from DeNavas-Walt, Proctor, and Smith (2011). House price index from Federal Housing
Finance Agency (2014). Unemployment rate from Council of Economic Advisers (2013).

for such omitted variables is to express the included variables

as first differences over time because then possibly omitted
variables that do not vary over time would drop out. That may
deal with the problem of non-stationary variables, but one
cannot be sure. The panel regression equations are estimated
in log differences, which is equivalent to percent changes.
There may be time fixed effects that vary over time but
66 Consumers’ Shift to Debt

are constant across states in any year. One solution for that
problem is to include a variable that varies by year but is con-
stant across states in any one year. The U.S. unemployment
rate variable presumably deals with the problem of time fixed
effects and so is included in all three debt equations.
The panel regression equations for estimating changes in
household debt are shown in Table 4.11. All three equations are
estimated as fixed effects with robust standard errors. All var-
iables are in log difference form. That presumably eliminates
state fixed effects that do not vary by time. The share of income
for the first through fourth income quintiles by state and year
is the key right-hand variable. A plausible assumption is that
changes in debt respond with a lag to changes of income dis-
tribution among the bottom 80 percent of households. Having
no theory as to how long such a lag may be, trial and error es-
timation with no lag up to a four-year lag revealed a three-year
lag as consistently negative and significant. In each of the three
debt equations, the coefficient on the income share variable is
negative, as expected (a fall in income share is associated with
a rise in per capita debt), and significant.
The other three right-hand variables in the debt equations
are likely determinants of household debt: unemployment,
house prices, and median household income. The unemploy-
ment rate is negative as expected (higher unemployment lowers
debt) and significant. Both the house price index and median
household income variables have positive coefficients and are
statistically significant, with the exception that the median in-
come coefficient in the equation for other debt is not significant.
These results tentatively support the argument that ris-
ing income inequality results in rising household debt, and the
estimated effect is not trivial. The equations in Table 4.11 are
estimated in first differences of logs, and so the coefficients
 Consumers’ Shift to Debt 67

Table 4.11. Estimated Panel Regression Equations, Log Difference

in Household Debt Per Capita by Year (1999–2010) and State
Left-hand variables (all are log differences)
Right-hand Total Household Other
variables (all are log household debt mortgage debt household debt
differences) per capita per capita per capita
Coefficients and t-statistics
(robust standard errors)
State income share, –0.186 –0.229 –0.118
quintiles
1–4 lagged three years (–2.7) (–2.6) (–2.0)
U.S. unemployment –0.083 –0.095 –0.055
rate (–8.1) (–8.0) (–0.4)
State house price +0.350 +0.383 +0.269
index (+9.9) (+9.4) (+8.5)
State median house- +0.106 +0.133 +0.050
hold income (+2.1) (+2.0) (+1.0)
Constant +0.035 +0.047 +0.010
(+27.0) (+32.0) (+9.6)
n 407 407 407
2
R within 0.38 0.32 0.34
R2 between 0.26 0.41 0.04
R2 overall 0.37 0.32 0.32
Source: Author’s calculations.

indicate percent changes of the left-hand variables, the per

capita debt measures, resulting from a 1 percent change in a
particular right-hand variable. The coefficients are estimated
elasticities. Thus their values can be compared. However, the
differences between coefficients of any variable are not statisti-
68 Consumers’ Shift to Debt

cally significant, and so one cannot infer, for example, that the
change of income shares has a larger effect on mortgage debt
than on all debt or on other debt. In every equation the largest
coefficient is on the state house price index, indicating that a 1
percent rise in house prices is associated with anywhere from
a 0.4 percent rise in per capita mortgage debt (equation 2) to
a 0.3 percent rise in other debt (equation 3), ceteris paribus.
That suggests that the rise of house prices was important for
the increase of household debt. Note that the coefficients on
the per capita income variable are much smaller, ranging from
+0.05 to 0.13, indicating that a 1 percent rise of per capita in-
come is associated with a rise in debt of only about 0.1 percent.
This result linking rising income inequality to rising
household debt is illustrative and not definitive for a number
of reasons. First, the inferred one-way causality—rising house
prices leading to rising debt—could be two-way; that is, rising
debt may contribute to rising house prices. Second, using aver-
age state data, there is a risk of an ecological fallacy—namely,
drawing inferences about individual household behavior from
averages for all households in a state. For example, the house-
holds in a state going into debt may be different from the
households losing ground in the income distribution. With
these caveats in mind, the claim that these results support the
causal hypothesis that rising income inequality leads to ris-
ing household debt is tentative and begs for future thorough
econometric investigation.
The most novel feature of the estimation procedure used
is to let one variable, the national unemployment rate, stand
for the time fixed effect. That is, it varies by year but not by
state. The conventional way to deal with time fixed effects in
panel regression is to introduce dummy variables, one less than
the number of time periods. The equations were estimated
Consumers’ Shift to Debt 69

again, dropping the national unemployment rate and adding

ten dummy variables because the time period is eleven years.
Rather than show those alternative estimated panel regres-
sions (they are available upon request), the results are summa-
rized here. First, the key right-hand variable, the share of state
income in quintiles one through four, is always negative as ex-
pected but is never significant. Second, the time dummy var-
iables as a group are highly significant in all three equations,
with F statistics ranging from 53 to 254. Third, the overall R2
values are much higher, ranging from 0.79 to 0.65. One could
argue that the conventional way to deal with time fixed effects
is more appropriate than the use of a single national variable.
Nonetheless, the key right-hand variable measuring income
share of the bottom four quintiles is negative as expected in
both sets of equations.
The equations shown in Table 4.11 were estimated yet again
with a shorter time frame: 1999–2007, eliminating the years of
the financial crash and onset of the Great Recession—the reason
being that the effect of rising income inequality on the change in
debt would be much less or absent with those shocks that shut
down access to credit. But the time frame of the panel, eleven
years, was already questionably short. Not surprisingly, those
estimated alternative equations are poor, with two of the four
right-hand variables insignificant in each equation. Even so, the
key right-hand variable, the income share of the bottom four
quintiles, is negative as expected, but not significant.

PA N E L R E G R E S SIO N A NA LYSI S O F NAT IO NA L DATA

A panel regression equation is estimated relating consumer lia-

bilities to income, an income inequality measure, and two other
right-hand variables. The equation is estimated with fixed ef-
70 Consumers’ Shift to Debt

fects over twenty-four years and five quintiles. For all the rea-
sons noted above about estimation of the state panel regres-
sions, the equation is estimated in first differences. Also, there
is a time fixed effect variable—that is, it varies by year but not
by quintile. It is the Theil index of income inequality.
The dependent or left-hand variable in the equation is net
change in average annual liabilities in each quintile (1–5 quin-
tiles) and each year (1984–2007). Net change in average annual
liabilities is defined by the CEX as the sum of change in mort-
gage principal (primary residence as well as vacation home,
other properties, and home equity loans and home equity lines
of credit), change in principal balance on vehicle loans, and
change in amount owed to other creditors, primarily credit card
balances. The first independent or right-hand variable is change
in average annual after-tax income in a quintile and a year. The
inequality variable, the Theil index, varies by year and not by
quintile, and so it is a time fixed effect in the equation.
The Theil index of income inequality is as reported in
the CPS. A rising Theil index indicates increasing income ine-
quality. The expectation is that there is a positive relationship
between the change in the Theil index and the change in aver-
age household liabilities. That is, with rising income inequality,
measured by a positive change in the Theil index, households
increase their liabilities to maintain their consumption, as ar-
gued by some of the authors cited and as inferred from the
state panel regression analysis. Access to credit can cushion
the effect of increased income inequality upon consumption.
The CEX data are subject to an important break be-
ginning in 2004 when income of non-reporting households
(formerly 16 percent or so of the CEX sample) is imputed as
explained above. The 1984–2003 data are not comparable with
the 2004–07 data. Rather than end the estimation period in
Consumers’ Shift to Debt 71

Table 4.12. Estimated Panel Regression Equation. Change

in Household Liabilities by Year (1984–2007) and Quintile
Coefficients and
Right-hand variables t-statistics
Constant 7,476
(11.7)
Change in after-tax income, lagged 0.977
(2.1)
Year dummy 1,931
(1.3)
Interaction term (Year dummy  income) 0.127
(3.4)
Change in Theil index 42,277
(1.6)
R2
Within 0.41
Between 0.99
Overall 0.42
n 110
Source: Author’s calculations.

2003, a dummy variable is introduced equal to one in the years

of income imputation, 2004–07, and zero in the prior years.
The dummy variable is interacted with the average after-tax
income by quintile and year. Of course both the dummy and
the interaction term must be expressed as levels, not changes.
As changes they would both equal zero. Although the CEX data
are available through 2010, the estimation period for the equa-
tion is limited to 1984–2007 in order to exclude the recession
and financial crisis years. The liabilities and income data are
from the CEX. All dollar amounts are expressed in 2008 dollars.
Table 4.12 presents the estimated equation. As before, all
72 Consumers’ Shift to Debt

are panel regressions estimated in first differences (changes

in variables rather than levels, except for the dummy variable
and the interaction term). The dependent variable is change
in average household liabilities by quintile and year. The first
three right-hand variables are the change in average income
lagged one year, the year dummy variable, and the interaction
term of income and the year dummy. The fourth right-hand
variable is the change in the Theil index of income inequality.
That is the variable of central interest because it is expected to
capture the effect, if any, of income inequality upon changes in
household liabilities.
The coefficient on the change in the Theil index is positive
as expected (higher income inequality is associated with more
household borrowing), but it is not statistically significant.
Turning to the other right-hand or independent varia-
bles, the change in income variable is positive as expected and
highly significant. The year dummy is positive and insignificant.
The interaction term is positive as expected and highly signifi-
cant. The interaction term appears in the years 2004–07, when
the change in liabilities was much higher than in earlier years.
This result by quintile and year provides weak support
of the hypothesis that rising income inequality leads to ris-
ing household debt. The fact that the coefficient on the Theil
index is not statistically significant is the main drawback. That
may in part result from the small sample size, 110, compared
with the state regression equations with sample sizes of 400
or more. Again, however, there is risk of the ecological fallacy
because the observations are for averages of quintiles, not indi-
vidual households. Future work on this issue would be mark-
edly improved by using the individual household data from
the Consumer Expenditure Survey.
 V
Consumption Theory
and Its Critics

T
he neoclassical theory of consumption is not ger-
mane for understanding the financial crisis and the
Great Recession. Jettisoning that theory in favor of one
that gives central place to the distribution of income,
relative income, and consumption, as well as household debt,
is necessary for devising public policies to shorten the Great
Recession by dealing with the huge overhang of household
debt.
This chapter traces the development of the economic
theory of consumption from Malthus and Ricardo to Milton
Friedman. The arguments of critics of the prevailing Friedman–
Modigliani–Brumberg paradigm are presented, leading to the
outline of a theory to replace it. In the existing paradigm, there
is no place for income distribution. Consequently, public policy
analysis of the Great Recession leaves out of consideration what
is a major cause of the Great Recession and the slow recovery.
The central concept of Pareto efficiency for economic the-
ory renders income distribution unimportant. Also, Kuznets’s
74 Consumption Theory and Its Critics

hypothesis of eventually falling income inequality in rich

democratic states, which was long under way in the United
States when he wrote, was morally appealing in the Cold War,
and the “no free lunch” message of Okun was sobering for
liberals. Malthus and Ricardo disputed over effective demand
and saving. Ricardo accepted Say’s law (supply creates its own
demand) while Malthus did not. Malthus believed that the dis-
tribution of income affected aggregate demand, and that ex-
cess saving by capitalists and landlords could lower demand.
Under Say’s law, that cannot happen in the aggregate economy
because the act of production generates sufficient demand. Ri-
cardo won that dispute in that classical economics followed
his analysis based on Say’s law. A century later, Keynes de-
murred. He argued that the attempt to save too much could
lower output, income, and employment, thus reducing saving,
too. Keynes was sympathetic to Malthus’s view on effective de-
mand and the importance of the distribution of income. He
inferred from his analysis of the average and marginal pro-
pensity to consume that as aggregate income rises, the average
propensity to consume (APC) falls. Thus in the long term, the
economy would face slower growth or stagnation. To post-
pone such a development, Keynes believed that a more equal
distribution of income would make effective demand stronger
than a less equal distribution of income. Empirical evidence
and new theories of consumption eclipsed Keynes’s view on
consumption. The post–World War II boom in the United
States included a stable, not a falling, APC as incomes rose
rapidly. Modigliani and Brumberg’s theory of consumption
argued that the APC was stable, and the empirical evidence
appeared to confirm that. Income distribution played no role
in their theory. Also excluded were psychological factors driv-
Consumption Theory and Its Critics 75

ing consumption that Keynes considered important. However,

the theory of Modigliani and Brumberg does provide an ex-
planation for the positive coupling of income inequality with
a rising APC through their emphasis on wealth, although they
did not anticipate that link.
The permanent income theory of consumption devel-
oped by Milton Friedman has become, along with the views of
Modigliani and Brumberg, the mainstream view. Two of the
three basic assumptions of his consumption theory no longer
hold true—namely, a stable income distribution and a con-
stant saving rate. The APC has not been stable, as Friedman
assumed, but rather has risen for twenty-five years. Recent
empirical work calls into question his view that income fluc-
tuations are mostly confined to transitory income and thus do
not affect consumption.
There are some modern critiques of the mainstream the-
ory of consumption. They posit roles for income inequality
and household debt in explaining consumption equality in the
face of rising income inequality and the falling rate of saving.
We consider both Veblen’s and Frank’s theories of consumption,
in which emulation of the consumption of others looms large.
Duesenberry’s theory of consumption gives central place to
relative income and consumption. Some behavioral economists
have attacked the all-important rational consumer premise of
mainstream consumption theory, resurrecting Keynes’s psycho-
logical motives in consumption. We will consider the technical
arguments as to whether the observed rise of the APC is real
or not. Finally, we will look at a revised theory of consumption
in which income distribution and consumer debt are central,
and the rational consumer is replaced by a distribution of het-
erogeneous consumers whose choices are not always rational.
76 Consumption Theory and Its Critics

Pareto Efficiency, Kuznets’s Inverted-U

Hypothesis, and Okun’s Tradeoff
“An economic situation is Pareto efficient if there is no way to
make some group of people better off without making some
other group of people worse off.”1 Pareto efficiency means that
any change that leaves at least one person better off, and no
person worse off, is an improvement. Applying the Pareto effi-
ciency principle to the distribution of income, any change that
raises the income of some while leaving the income of others
unchanged is an improvement—that is, a rise in total utility.
Thus an income increase that all goes to the top 1 percent of
the income distribution (or the bottom 1 percent) is Pareto
efficient. This principle of Pareto efficiency is a fundamental
tenet of neoclassical economics. The historical distribution of
income in industrialized nations mostly shows all groups with
absolute gains. Such changes are Pareto efficient, and the dis-
tribution of gains is not an issue for positive economics. Hence
economic theory has nothing to say about the distribution of
income. “But efficiency is not the only goal of economic pol-
icy. For example, efficiency has almost nothing to say about
income distribution or economic justice.”2
But it was not just economic theory that turned econ-
omists’ attention away from income distribution; it was also
empirical evidence provided by Simon Kuznets.3 Using his-
torical data for a number of countries, Kuznets argued that
in the shift from agricultural societies to industrialized soci-
eties, income inequality would rise rapidly because of grow-
ing concentration of industrial wealth. Eventually, with the
expansion of democracy and institutions favoring economic
protections for labor and a strong social safety net, income in-
equality would decrease from its peak in the period of rapid
Consumption Theory and Its Critics 77

i­ndustrialization—thus the name, “Kuznets’s inverted-U hy-

pothesis.” With time on the horizontal or x-axis and a meas-
ure of income inequality, such as the Gini coefficient, on the
vertical or y-axis, the curve would rise (more inequality) in
the rapid industrialization phase, then reach a peak, and begin
to decline. His hypothesis gained wide attention and support
in the post–World War II era of the United States because it
seemed to trace the actual path of United States income in-
equality: rising in the Gilded Age, peaking in 1929, and de-
clining during the Depression, World War II, and the postwar
boom proceeding when Kuznets’s hypothesis was promul-
gated during the 1950s and 1960s. It was a reassuring story fa-
voring capitalism in the midst of the Cold War competition
with communism. It is not, however, a story that has held up
well, as income inequality has reversed course, rising for al-
most four decades.
Another reassuring story that gained much attention
and respect was told by Arthur Okun, a liberal economist
from Yale who had served in Democratic administrations,
in his book Equality and Efficiency: The Big Tradeoff (1975).4
The message was that gains in social welfare through progres-
sive taxation could come only at the loss of efficiency, which
would reduce economic growth and future living standards.
Although Okun provided no strong evidence, his notion of
a strict tradeoff took root in the policy world. It was already
congenial to mainstream economists because of their belief
in full employment, whereby the shift of resources to social
goods necessarily meant fewer resources for private goods.
The centrality of Pareto efficiency in economics sug-
gested that we need not worry about income inequality in
theory because any redistribution of income that made some
better off and others worse off was not Pareto efficient and so
78 Consumption Theory and Its Critics

was beyond the domain of positive economics. Kuznets’s in-

verted-U hypothesis, supported empirically, suggested that we
need not worry about income inequality in fact because it was
destined to fall. Okun’s tradeoff was a sober announcement to
liberals that there are “no free lunches.” All that questionable
past baggage has been dragged into the present, resulting in
little interest in income distribution or inequality by main-
stream economists. Counting articles published in the most
prestigious economic journals from 2009 through 2013, very
few are about income inequality or income distribution gen-
erally, based on their titles. The highest count, nine in those
five years, goes to the annual American Economic Review Pa-
pers and Proceedings. Seven of the nine, however, were in the
most recent annual for 2013. The regular American Economic
Review and the Journal of Political Economy are tied, with six
articles each. The Quarterly Journal of Economics comes in
last, with five articles over five years. That total of twenty-six
articles about income distribution or income inequality over
five years amounts to less than 2 percent of the 1,561 articles
published in those journals. That certainly indicates lack of
interest and perhaps some hostility. As Krugman noted in his
review of Thomas Piketty’s book Capital in the Twenty-first
Century, “Some economists (not to mention politicians) tried
to shout down any mention of inequality at all. ‘Of the ten-
dencies that are harmful to sound economics, the most se-
ductive, and in my opinion the most poisonous, is to focus
on questions of distribution,’ declared Robert Lucas Jr. of the
University of Chicago, the most influential macroeconomist
of his generation, in 2004.”5 Although economic theory may
be silent about income distribution, macroeconomics, in its
incipient early phase before it was carved out as a subfield of
economics, did consider income distribution in its musings
Consumption Theory and Its Critics 79

about what we now call “effective demand.” What those early

writers had in mind was not the distribution of income among
households as we mean by it today, but rather the distribution
of factor payments between three groups: wages to laborers
and profits to capitalists and landlords. In that simpler world
of early nineteenth-century Britain, the groups demarked the
“haves” were capitalists and landlords, and the “have-nots”
were laborers. Thus it is not mistaken to read their remarks on
the “distribution of product” as pertaining to the distribution
of income.

The Malthus–Ricardo Dispute

over Effective Demand
Say’s law has been commonly stated as “supply creates its own
demand.” What Say actually wrote on a section dealing with
demand in his Treatise on Political Economy (translated from
French in 1824) is, first, “Which leads us to a conclusion, that
may at first sight appear paradoxical; viz. that it is production
which opens a demand for products.” And second, “Thus the
mere circumstance of the creation of one product immediately
opens a vent for other products.” And finally, “Thus, it is the
aim of good government to stimulate production, of bad gov-
ernment to encourage consumption.”6
Say’s law was accepted by Ricardo.7 As Cottrell points
out, “the central thrust of the ideas that have been thus la­
beled [Say’s law], in the writings of Say and Ricardo, is simple
enough: there can be no such thing as a general deficiency of
demand, although particular goods can obviously be overpro-
duced in the sense of being produced in such quantities that
they cannot be sold at their natural price.”8
Malthus did not accept Say’s law, as noted by Schum-
80 Consumption Theory and Its Critics

peter: “A much more important reason for Malthus’ dissent

from Say and much more basic to his principle of effectual or
effective demand was, however, his opinion that saving, even
if promptly invested, may lead to a deadlock if carried beyond
a certain optimal point.”9 In his biographical essay on Malthus,
Keynes notes, “In economic discussions Ricardo was the ab-
stract and a priori theorist, Malthus the inductive and intuitive
investigator who hated to stray too far from what he could test
by reference to the facts and his own intuition.” Contrasting
the two, largely based on their correspondence over many
years, Keynes writes, “Ricardo is investigating the theory of
the distribution of the product in conditions of equilibrium,
and Malthus is concerned with what determines the volume
of output day by day in the real world. Malthus is dealing with
the monetary economy in which we happen to live; Ricardo
with the abstraction of a neutral money economy.”10
Although Malthus was not as clear in his writing as one
would wish, in a letter to Ricardo he emphasizes the central
importance of what we would call the distribution of income
as well as discretionary or non-necessity consumption and the
danger of excess saving: “But the grand question is whether
it [total production] is distributed in such a manner between
the different parties as to occasion the most effective demand
for future produce: . . . an attempt to accumulate very rap-
idly which implies a considerable diminution of unproduc-
tive consumption by greatly impairing the usual motives to
production must prematurely check the progress of wealth.”11
Keynes argues that the point about saving was made clearer in
Malthus’s Principles of Political Economy (1820) than in the cor-
respondence with Ricardo, where Malthus addresses the slump
in Britain following the end of the Napoleonic war: “He [Mal-
thus] points out that the trouble was due to the diversion of
Consumption Theory and Its Critics 81

resources, previously devoted to war, to the accumulation of

savings; . . . and that public works and expenditure by land-
lords and persons of property was the appropriate remedy.”12
Keynes argues that Malthus saw that saving is not necessarily
a public virtue. Attempts to save could shrink total output.
Keynes notes that in Malthus’s preface to his Principles of Po-
litical Economy, he recognizes the importance of saving for the
increase of production, but that too much saving could curb
the growth of wealth. Ricardo disagreed. Keynes notes that
“Surely it was a great fault in Ricardo to fail entirely to see any
significance in this line of thought.”13
The careful analysis of the Malthus–Ricardo debate by Allin
Cottrell faults Malthus and argues that “Ricardo deserved to get
the better of the debate.”14 Maybe so, but Ricardo did not struggle
with the important issues of effective demand, and excess saving,
issues that lay dormant in economics until Keynes’s General
Theory of Employment, Interest and Money (1936).

Keynes’s Theory of Consumption

Keynes developed his theory of consumption in that book.
Mostly expressed in words, it has been translated into algebra
in generations of macroeconomics texts, a clarifying improve-
ment made possible by the development of systems of national
income accounts in developed nations after 1936. However,
something was lost in the translation into algebra—namely,
the importance of psychological factors in determining con-
sumption. His key idea was that the marginal propensity
to consume is positive and less than one, while the average
propensity to consume falls as income rises. His controver-
sial idea, in light of his classical economics predecessors and
contemporaries, is that the rate of interest is relatively unim-
82 Consumption Theory and Its Critics

portant in determining consumption, whereas income is of

central importance. His consumption function is:

C = a + bY, (1)

where C is aggregate consumption, Y is aggregate income, a > 0,

and 0 < b < 1 is the marginal propensity to consume. Dividing
both sides by Y gives the average propensity to consume, APC.

APC = C/Y = a/Y + b (2)

This satisfies Keynes’s condition that as income, Y, increases,

the average propensity to consume falls. But the post–World
War II economy of the United States did not satisfy Keynes’s
condition. That is, higher incomes after the war, induced by
strong economic growth, did not result in a declining APC
and thus a rise in the rate of saving. Thus the postwar expe-
rience did not support Keynes’s contention that the APC falls
as income rises: “Since I regard the propensity to consume as
being (normally) as such to have a wider gap between income
and consumption as income increases, it naturally follows that
the collective propensity for the community as a whole may
depend . . . on the distribution of incomes within it.”15
Although Keynes doesn’t explicitly state “average pro-
pensity to consume,” his definition of “propensity to consume”
in the General Theory clearly means the average, not the mar-
ginal.16 What is unclear is if he meant that the long-term APC
of the nation falls as income rises, which Kuznets showed not
to be the case, or that the cross section of APC in any year
across income groups falls with rising income. He may have
meant both, in which case he was only half wrong.17
Recent time-series data that contradict Keynes expec-
Consumption Theory and Its Critics 83

Figure 5.1. Average propensity to consume (APC), 1967–2013. Sources:

Council of Economic Advisers (2011), Table B-30, and Bureau of Eco-
nomic Analysis (n.d.), “National Economic Accounts,” July 2014.

tation are shown in Figure 5.1. There the line is the average
propensity to consume calculated from aggregate personal
consumption expenditures and disposable income, 1967–2013.
Rather than decline over those forty-six years, it oscillates be-
tween a ceiling of 0.90 (1969) and a floor of 0.86 (1982) from
1967 to 1986. Thereafter it rises for twenty years, except in
recessions, hitting a peak of 0.95 in 2005. That pattern of a
long-term rise of the APC was not anticipated by Keynes, and
certainly not by Friedman or Modigliani and Brumberg.

Modigliani and Brumberg’s Theory of Consumption

Modigliani and Brumberg were aware of the fact that the long-
term data of stable APC did not agree with the short-term data
84 Consumption Theory and Its Critics

of falling APC.18 Their life-cycle hypothesis solved that anom-

aly by replacing current income with lifetime income and by
introducing wealth, accumulated through saving, in the con-
sumption function. The rational consumer saves while work-
ing and dissaves when retired with the object of maintaining
stable consumption over the consumer’s expected lifetime.
Their consumption function for the economy, as represented
in the currently popular macroeconomics text by Mankiw, is

C = αW + βY, (3)

where W is accumulated wealth and Y is income. For simplicity

of exposition, the growth path of W and Y as well as the effect of
the interest rate on W are omitted. The graph of this consump-
tion function for individuals in a given year is a straight line
where the intercept is αW. The average propensity to consume
is derived by dividing the consumption function by Y.

APC = C/Y = α(W/Y) + β (4)

This formulation solves the anomaly of the short-run

and the long-run APC, as Mankiw explains: “Because wealth
does not vary proportionately with income from person to
person or year to year, we should find that high income corre-
sponds to a low average propensity to consume when looking
at data across individuals or over short periods of time. But,
over long periods of time, wealth and income grow together,
resulting in a constant ratio W/Y and thus a constant average
propensity to consume.”19 Contrary to the views of Modigliani
and Brumberg, the APC might rise over a long period, as it did
from 1984 to 2005. A theory that can explain those long-term
increases in the APC lies in Modigliani and Brumberg’s analy-
Consumption Theory and Its Critics 85

Figure 5.2. Real net worth of households and real disposable

income. Dotted line: Disposable income. Solid line: Net worth.
Source: Bureau of Economic Analysis (n.d.), “National Economic
Accounts,” Table S.3.a, December 2013.

sis of consumption in which the importance of wealth for con-

sumption is asserted. Examining equation (4), it is clear that
APC will rise if wealth, W, grows faster than income, Y, for a
long period, and if αʹ > 0. As shown in Figure 5.2, real net worth
of households has been expanding faster than real disposable
income from the mid-1990s until 2007, except for the sharp
drop when the high-tech stock market bubble burst in 2000. It
would be a stretch to call that twelve-year pattern “temporary.”
Figure 5.3 plots the W/Y term of the Modigliani APC,
using net worth as W and disposable income as Y. As shown,
W/Y edged up from 4.4 in 1974 to 5.0 in 1994, a gain of 0.6
points in two decades. In the next five years, 1994–99, it shot
up 1.5 points, to 6.5. Then it tumbled in the tech bubble and
86 Consumption Theory and Its Critics

Figure 5.3. Ratio of net worth to disposable income, 1960–2012.

Source: Bureau of Economic Analysis (n.d.), “National Economic
Accounts,” Table S.3.a, December 2013.

rebounded in the housing bubble to a high of 6.7 in 2006 be-

fore dropping to 5.0 in 2009. In the thirty-two years from 1974
to 2006, W/Y rose 2.3 points, contrary to Mankiw’s assertion,
cited above, that “over long periods of time, wealth and in-
come grow together, resulting in a constant ratio W/Y and
thus a constant average propensity to consume.”20
The expression for APC based on Modigliani and Brum-
berg’s consumption function, equation (4), can be used to explain
the long-term rise of the APC (see Figure 5.1). Clearly, wealth
grew faster than income (shown in Figure 5.2), and so by equa-
tion (4) APC would rise, as it did. That is not an empirical result
expected by Modigliani and Brumberg. Likewise, they posited no
role for income distribution in their theory of consumption.
As shown, increasing income inequality and a rising
Consumption Theory and Its Critics 87

APC occurred together in the period 1984–2005. But what is

the theory that would support such a causal link? Faced with ris-
ing income inequality, households may attempt to maintain con-
sumption through increased borrowing. In the face of a “hard in-
come constraint,” consumption limited by income, such attempts
would not be successful. But a “soft income constraint” enables
households to maintain or increase consumption through in-
creased borrowing. As indicated in Chapter IV, household in-
debtedness across all income quintiles expanded much more
than income from 1995 to 2007. Because the largest part of
their increased indebtedness was for residential property, that
borrowing enhanced their wealth as house prices rose during
the bubble. Hence, income inequality indirectly contributed to
the faster rise of wealth than income.

Friedman’s Theory of Consumption

and Some Critics
Friedman, like Modigliani and Brumberg, rejected Keynes’s
formulation that consumption depends on current income.21
Friedman hypothesized that annual income of a person or
household consisted of two parts: permanent income and tran-
sitory income. Rationality would require that the consumer
would base consumption only upon permanent income. Thus
Friedman’s consumption function states that consumption is
some fraction of permanent income, Yp, noted by Mankiw as

C = αYp (5)

Then Mankiw shows the APC in Friedman’s model as

APC = C/Y = αYp/Y (6)

88 Consumption Theory and Its Critics

And he notes, “According to the permanent-income hypothe-

sis, the average propensity to consume depends on the ratio of
permanent income [Yp] to current income [Y]. When current
income temporarily rises above permanent income, the aver-
age propensity to consume temporarily falls; when current in-
come temporarily falls below permanent income, the average
propensity to consume temporarily rises.”22
In Friedman’s theory of consumption, the rational con-
sumer plans permanent consumption based on permanent in-
come. For one consumer, the propensity to consume out of
permanent income is a function of the rate of interest, ratio of
wealth to permanent income, and household characteristics,
such as age and family size. When Friedman generalizes in-
dividual consumption to aggregate consumption, he assumes
that the “distribution of consumer units by income is indepen­
dent of their distribution by (i), w, and u [rate of interest, ratio of
wealth to permanent income, and household characteristics].”23
So his aggregate consumption function has permanent con-
sumption a linear function of permanent income,

C p = k (Yp), (7)

where the coefficient k is determined by the myriad factors

noted above for millions of households. But he recognizes that
“The assumption . . . that the distribution of consumer units by
income is independent of their distribution by (i), w, and u is
obviously false in a descriptive sense.” Nonetheless, he asserts
in the same paragraph,

At the same time, although the interdependence

between these variables and the distribution of in-
come may be important for some problems, it may
Consumption Theory and Its Critics 89

not be for this aggregation. The interdependence

enters in a rather complex way and equation (2.10)
[equation (7) above] is a good approximation even
when interdependence exists. . . . If . . . equation
(2.10) [7] is a good approximation of the relation
among observed magnitudes, this must be inter-
preted to mean that the interdependence is of only
secondary importance.24

That is a critical assumption because it removes the “distri-

bution of consumer units by income” as a determinant of the
APC and thus the saving rate. Thus Friedman’s k, the APC, is
constant in the long run, and therefore so is the saving rate.
The development of Friedman’s theory of consumption
by mainstream economists uses a “representative agent” to
stand for the millions of heterogeneous consumer units. The
intellectual trend to establish micro foundations for macro-
economics, championed by Lucas, has failed to displace the
representative agent from macro models of consumption and
saving. Although one writer, Kevin Hoover, states of main-
stream macroeconomists that “most have accepted micro
foundations in the form of the representative agent model or
some near variant, despite the fact that a plausible case has
never been offered for how any such agent could legitimately
represent millions of economic decision makers.”25
In a paper comparing representative agent models, the
mainstream approach, with consumer heterogeneity models,
the authors conclude:

In representative agent models with time additive

preferences, it seems difficult to obtain the departure
from permanent income behavior that we observe
90 Consumption Theory and Its Critics

in the incomplete markets models with heteroge-

neous preferences. In other words, the interaction
of consumers with heterogeneous preferences in an
incomplete-market setting leads to new insights.
Among the models we study, those that come clos-
est to matching real-world wealth distributions are
precisely models with heterogeneous preferences
and incomplete markets. . . . As we have shown in
this paper, introducing preference heterogeneity
into the standard model allows a closer match be-
tween the model and data.26

In the concluding section of his book A Theory of the Con-

sumption Function (1957), Friedman takes issue with the inter-
pretations attached to the Keynesian theory of consumption
—namely, that the APC will decline over time as income rises.
He asserts instead, “Yet there is ample evidence that (a) in-
equality of income has, if anything, decreased over time in
the United States, (b) savings have been a roughly constant
fraction of income over time in the United States, (c) com-
puted regressions have steadily been higher, the later the date
of the budget study. All three observations are entirely consis­
tent with the permanent income hypothesis presented in this
monograph.”27 His first two observations, declining income
inequality and stable saving rate (which infers a stable APC),
are not true for the period 1984–2007 (see Table 2.2 and Figure
5.1), and also not true for the early twentieth century, as shown
in Chapter VI. His third observation is based on regression
equations relating consumption to income in real dollars that
he estimated using data from eight budget studies ranging
over the long period from 1888–90 to 1950. He claims that
those regression equations show that “computed regressions
Consumption Theory and Its Critics 91

have steadily been higher, the later the date of the budget
study.” That is, calculated consumption rises with later budget
studies, contrary to the view that “regressions computed from
budget studies made at widely spaced dates will not differ sys-
tematically.”28 Friedman does not show the estimated regres-
sion equations that he used to calculate the real consumption
levels.
Friedman’s period of analysis, pre-1957, was mostly one of
falling income inequality, except for the 1920s (which he over-
looked), so he had no reason to ask the question of whether
income inequality leads to consumption inequality. In light of
the long rise of income inequality, the question is now ger-
mane, and it is posed by Dirk Krueger and Fabrizio Perri in an
empirical and theoretical article. In a careful analysis of Con-
sumer Expenditure Survey data, they find that rising income
inequality has not been accompanied by a similar rise of con-
sumption inequality. In their theoretical model, maintaining
stable consumption in the face of increasing income inequality
may be facilitated by credit and insurance—in other words, by
a soft budget constraint. “If, however, the structure of private
financial markets and informal insurance arrangements does
not respond to changes in the underlying stochastic income
process of individuals, then no further hedging against the in-
creasing risk is possible, and the increase in income inequality
leads to a more pronounced rise in consumptions inequality.”29
In other words, if consumers face a hard budget constraint.
They present a figure in which the Gini coefficient is plotted
against non-secured consumer credit as a share of disposable
income. It shows the two moving upward in tandem from the
late 1960s to the early 2000s. They conclude, “Combining this
figure with our consumption inequality observations may
suggest that consumers could, and in fact did, make stronger
92 Consumption Theory and Its Critics

use of credit markets exactly when they needed to (starting in

the mid-1970s), in order to insulate consumption from bigger
income fluctuations.”30
Those authors in the Friedman mainstream tradition
(Krueger and Perri) thus assert that rising income inequality
leads to increased consumer borrowing, a major proposition
of this book. But in their case, such borrowing is only tempo-
rary, in order to smooth consumption. As shown in Chapter
IV, the massive rise of household debt cannot be described as
temporary.
The empirical part of their paper, covering 1980–2003,
convincingly shows that even with the large rise of income
inequality in the United States, consumption inequality has
risen very little. The theoretical part of their paper hypoth-
esizes that “the volatility of idiosyncratic labour income has
been an important factor in the increase in income inequal-
ity.”31 What they mean by “idiosyncratic labour income” is
transitory income. So the implication is that the rise of in-
come inequality is all in the transitory component, not in the
permanent component. That is simply an assertion with no
empirical support. Without that assertion, income inequality
would have to be recognized as affecting permanent income,
contradicting Friedman’s view. Permanent consumption is
protected or maintained through borrowing, and Krueger
and Perri protect Friedman’s permanent income hypothesis
by assuming the rise of income inequality is almost all in the
transitory component. That assumption is called into question
by van Treeck, who cites a number of works that provide sta-
tistical evidence of increased variance of permanent income,
not transitory income. “These results seem to conflict with the
view that the rise in inequality was driven by insurable tempo-
rary shocks over the 1990s.”32
Consumption Theory and Its Critics 93

A paper that is strongly empirical, utilizing a vast So-

cial Security database going back to 1937, reaches the oppo-
site conclusion from Krueger and Perri. “In particular we find
that increases in annual earnings inequality are driven almost
entirely by increases in permanent earnings inequality, with
much more modest changes in the variability of transitory
earnings.”33
Another author also links income inequality to rising
household debt and stable consumption inequality. Matteo
­Iacoviello shows that income inequality and the debt of house-
holds move together. Measuring inequality with the cross
­section standard deviation of log earnings, he finds that it was
stable from 1963 to 1980 and rose sharply thereafter. Household
debt as a percentage of disposable income follows a quite sim-
ilar path, rising from 66 percent of disposable income in 1981
to 113 percent in 2003, the last year for his data. He develops a
dynamic general equilibrium model to explain the trend and
cycle in household debt. He concludes, “The rise of income
inequality of the 1980s and 1990s can, at the same time, ac-
count for the increase in household debt, the large widening of
wealth inequality, and the relative stability of consumption in-
equality.” Similar to Krueger and Perri, he finds that expanded
use of credit prevents a rise of consumption inequality when
income inequality increases. “In the model presented here, the
mechanism through which consumption inequality rises less
than income inequality in [sic] an expansion of credit from the
rich to the poor.”34
For this study, income inequality and consumption in-
equality measures have been calculated—namely, standard
deviations from the relative distribution of aggregate after-tax
income and aggregate consumption for the CEX data by quin-
tile for selected years from 1984 to 2007. The income standard
94 Consumption Theory and Its Critics

deviations are much larger, 46 percent to 58 percent larger,

than the consumption standard deviations, indicating that in-
come inequality is much larger than consumption inequality.
This agrees with the findings of both Krueger and Perri and
Iacoviello.

More Doubts About Friedman’s Consumption

Theory: Rajan, Van Treeck, and Others
Recent writers challenge Friedman on a number of issues in
his theory of consumption. One, that relative income, not ab-
solute income, is of central importance because households
care about their relative standing.35 Two, that saving rates
across income classes are not stable but rather they increase
with income.36 Three, that rising earnings inequality is not
driven by fluctuations in transitory earnings but rather by
rises in permanent earnings inequality.37 Four, the assumed
“hard intertemporal budget constraint, does not adequately
describe household spending and borrowing decisions in the
past quarter century . . . with financial innovation and greater
access to debt, the year-by-year budget constraint has become
soft.”38 Five (and not recent), that the APC is independent of
the level of permanent income.39
In his book Fault Lines, Raghuram Rajan also ties ris-
ing income inequality to rising consumer borrowing that be-
came unsustainable and brought on the financial crash. But in
making the link from income inequality to rising household
borrowing, he indirectly undermined the permanent income
hypothesis of Friedman by arguing that the massive rise of
household debt was unsustainable. In Friedman’s model, debt
was a temporary recourse to smooth consumption over peri-
ods of lower transitory income. “I have argued that an impor-
Consumption Theory and Its Critics 95

tant political response to inequality was populist credit expan-

sion, which allowed people the consumption possibilities that
their stagnant incomes otherwise could not support.”40
Here is a paraphrase of what van Treeck calls the “Rajan
hypothesis.” The benefits of rising aggregate income over past
decades have been confined to a small set of households at the
top of the income distribution, so that consumption of lower
groups has been financed through rising use of credit. The
process was facilitated by government directly through credit
promotion and indirectly through deregulation of the finan-
cial sector. But with the drop of housing prices and the sub-
prime mortgage crisis beginning in 2007, overindebtedness of
the household sector became apparent and debt-financed pri-
vate demand expansion ended. What van Treeck draws from
Rajan’s argument was not stated by Rajan but may be implied:

The more important implication of his analysis is

the rejection of the conventional theories of con-
sumption, which see no link between the inequal-
ity of (permanent) income and aggregate personal
consumption, and hence no need for government
action to stimulate consumption and jobs in re-
sponse to higher inequality. . . . The Rajan hy-
pothesis posits that given the rise in inequality the
credit expansion in the personal sector was both
necessary for supporting aggregate demand and
employment, and it was unsustainable. . . . This
lack of attention to inequality seemed justified by
the permanent income hypothesis . . . which pos-
its that household consumption is unrelated to the
inequality of permanent income. . . . The Rajan hy-
pothesis, which relies on the assumption of a higher
96 Consumption Theory and Its Critics

inequality in the permanent component of income,

is thus of great theoretical importance, and it bears
resemblance to the relative income hypothesis.41

This following quote from van Treeck may be reading too

much into Rajan, but it underscores the importance of rela-
tive income: “Rather, the Rajan hypothesis can be seen as an
application of the relative income hypothesis, which predicts
that households will react to a decline in (permanent) relative
income by lower saving and higher debt.”42
Two recent authors, Michael Bordo and Christopher
Meissner, take issue with both Rajan and with Michael Kum-
hof, Romain Ranciere, and Pablo Winant because they “pro-
pose that rising inequality led to a credit boom and eventually
to a financial crisis in the United States in the first decade of
the 21st century as it did in the 1920s.” In their econometric
analysis, they examine data from 1920 through 2000 for four-
teen advanced nations and conclude, “Credit booms heighten
the probability of a banking crisis, but we find no evidence that
a rise in top income shares leads to credit booms.”43
Tuomas Malinen criticizes the econometrics of the
Bordo and Meissner article: “As real income keeps on stag-
nating, credit acquired by lower income households keeps on
growing and this trend eventually leads to defaults and stress
among financial institutions. First differencing removes this
trend and focuses the analysis on the short-term effects of
inequality on credit. If the relationship between inequality
and credit is long-run, i.e., trending in nature, using first dif-
ferenced variables may give biased information.”44 His panel
econometric results lead him to conclude: “In this study we
have tested the existence of such a long-run relationship [be-
tween income inequality and the share of credit to income].
Consumption Theory and Its Critics 97

According to the results, there is a long-run steady-state rela-

tionship between income inequality and leverage in developed
economies. The long-run elasticity of leverage with respect to
income inequality was found to be positive. This indicates that
income inequality increases leverage in the economy.”45 In an
article by Christopher Brown that asks whether income distri-
bution matters for effective demand, simulations are presented
by income deciles for the United States in 2001. His first-order
autoregressive equation for aggregate consumption includes
the Theil index of income inequality on the right-hand side.
When he does a simulation at Theil = 0.14 and then Theil = 0,
perfect equality of income, the resulting rise in consump-
tion is only 1 percent. Then he does simulations where Theil
= 0.14 in both cases, but in one he imposes a hard income
constraint—that is, consumption by quintile cannot be greater
than income, and not in the other. With the hard income con-
straint, aggregate consumption drops almost 16 percent. “The
results indicate that, if income imposes a hard constraint on
spending, income distribution can have very significant im-
plications for effective demand.”46 He notes the importance of
widened credit availability in the past few decades. “Thus wid-
ened credit availability is comparable to a decrease in income
inequality in terms of its effects upon the propensity to con-
sume. It follows that the aggregate propensity to consume can
remain stable or even increase, amidst a sharp rise in income
inequality—given a sufficient surge in borrowing.”47 As shown
in Figure 5.1, the APC has increased in a period of rising in-
come inequality made possible by a surge in borrowing.
One of the key assumptions of Friedman’s permanent
income hypothesis is that the APC is stable over time. By defi-
nition, if the APC is stable over time, then so too is the saving
rate, which is equal to 1 – APC, a stability that Friedman notes
98 Consumption Theory and Its Critics

in his conclusion. But the personal saving rate in the United

States has been trending down. Prior to 1984 it was fluctuating
around 10 percent. Thereafter it mostly declined, reaching a
low of 3 percent in 2005 (see Figure 4.1). Neither the perma-
nent income hypothesis nor the life-cycle hypothesis can ex-
plain the long decline in the saving rate (the inverse of the long
rise in the APC), as one Friedman critic, David Bunting, notes.
He seeks the explanation for the drop in the U.S. saving rate in
the distribution of income, which is usually ignored. Making
use of income distribution series of both CEX and CPS, he
combines quintiles one and two into the low-income group,
quintiles three and four into the middle-income group, and
quintile five into the high-income group, and then analyzes
the saving behavior by each group from the mid-1980s until
2005. Bunting concludes:

As the distribution data clearly shows, since 1980

saving by high income households accounts for not
the bulk, but all of aggregate saving. . . . Representing
20 percent of households, the high income group
saving rates have remained virtually unchanged
since 1985–87. . . . On the other hand, undetected
in the aggregate data, both the saving rates and the
saving of the lowest 80 percent of households
has deteriorated since 1985–87 as the low income
group rates fell 30 percent or more while dissaving
increased by $360 billion and the middle income
group saving collapsed with both saving rates and
saving turning from positive to negative figures.48

He argues that the saving rate should be calculated as a

weighted average rather than assuming that the aggregate rate
Consumption Theory and Its Critics 99

represents all behavior. The use of the aggregate rate to repre-

sent all consumers implies that saving is unaffected by poverty
or affluence and that the distribution of income does not matter.
In an earlier article, Bunting is more explicit in fault-
ing Friedman as well as Modigliani and Brumberg for using
a representative agent as reflecting aggregate consumption
behavior: “Unfortunately, by reducing aggregate spenders to
one, it [assumption of a representative agent] had the great
disadvantage of precluding consideration of any distributional
and demographic influences on aggregate behavior because
with only one agent, income as well as age, race, or sex differ-
ences are meaningless.”49 The critics noted above are of course
not the only critics of the neoclassical theory of consumption.
That puts them in the company of Malthus and Keynes and in
opposition to Ricardo and the neoclassical synthesis, which
has as one of its major tenets the theory of consumption de-
veloped by Friedman and also by Modigliani and Brumberg,
who hold the view, by omission, that income distribution does
not matter for a theory of consumption and saving. Their crit-
icisms, as well as those of Veblen, Duesenberry, and the behav-
ioral economists, are central to the theme of this chapter.

Veblen and Frank

Up to this point the parts of this chapter are mostly chrono-
logical: before and after Friedman. Here that pattern is bro-
ken by introducing Thorstein Veblen, who wrote long before
Friedman. The reason for this departure is that Veblen can be
considered a forerunner of behavioral economics.
Veblen, an economist classified in the Institutionalist
School, wrote in the late nineteenth and early twentieth cen-
tury. At that time there were no good data available on income
100 Consumption Theory and Its Critics

distribution or aggregate consumption in the United States.

His major works, however, present a theory of consumption
and have implications for income distribution. Conspicuous
consumption of the rich, he argued, would stimulate emulation
by the non-rich. In The Theory of Business Enterprise, he pos-
tulated a clash between engineers and professionals with busi-
nessmen who seek to maximize return. He thought the former
were interested in technical improvements for raising output
and lowering price, while the latter, businessmen, sought to
limit output and raise price. The first, he argued, would lead to
full employment and high wages, while the second to unem-
ployment and low wages. The implication for income distribu-
tion is clear.50 His Theory of the Leisure Class was enormously
popular, especially during the Great Depression, when he was
lionized: “The collapse of orthodox economic doctrine during
the years of the world depression has vindicated the keenly
analytic and prophetic writings of Thorstein Veblen.”51 But he
never made much impression on mainstream (neoclassical)
economics. In Joseph Schumpeter’s magisterial History of Eco-
nomic Analysis, first published in 1954, the index notes four
page references to Veblen. All of them are incidental asides.
None notes his challenge to the pre-Friedman neoclassical
theory of consumption.52 To be fair to Schumpeter, his title is
History of Economic Analysis, not history of economic thought.
Because Veblen produced no analytical models or empirical
support for his theories, we can presume that Schumpeter did
not include him among his parade of economists. Obscurity
was no barrier to inclusion, just as fame was no assurance of
inclusion.
Veblen’s influence apparently lives on. In an empirical
paper, “Emulation, Inequality, and Work Hours: Was Thor-
stein Veblen Right?” the authors, Bowles and Park, define “Veb­-
Consumption Theory and Its Critics 101

len effects” as “the manner in which a desire to emulate the

consumption standards of the rich may influence an individu-
al’s allocation of time between labor and leisure.” The authors
develop a model of the “Veblen effects” that they estimate
econometrically for many OECD nations. They conclude: “We
have shown that increased inequality induces people to work
longer hours and have also provided evidence that the un-
derlying cause is the Veblen effect of the consumption of the
rich on the behavior of those less well off.”53 The mainstream
consumption theory cannot countenance a “Veblen effect”
because preferences are assumed independent—that is, no one
cares what those better off are consuming.
One of Veblen’s challenges to mainstream economics
appeared in one of the profession’s preeminent journals, then
and now, the Quarterly Journal of Economics, entitled “Why
Is Economics Not an Evolutionary Science?” In the first par-
agraph he complained, “It may be taken as the consensus of
those men who are doing the serious work of modern anthro-
pology, ethnology, and psychology, as well as those in the bi-
ological sciences proper, that economics is hopelessly behind
the times, and unable to handle its subject matter in a way
to entitle it to standing as a modern science.”54 More than a
hundred years after Veblen wrote that, Robert Frank argued in
his 2011 book, The Darwin Economy, that in a hundred years
almost all economists would claim Darwin as the father of
economics rather than Adam Smith. “One century hence, if a
roster of professional economists is asked to identify the intel-
lectual father of their discipline, a majority will name Charles
Darwin.”55 In other words, Frank, like Veblen, thought that
economics would become an evolutionary science. However,
in Veblen’s case it was a hope, while in Frank’s case it is his
optimistic perception of how economics is evolving. His re-
102 Consumption Theory and Its Critics

viewer in the Journal of Economic Literature (2012) was kind

but dismissive, wondering if reading it was “an efficient use” of
economists’ time.56
Frank’s case for Darwin replacing Smith as the intellec-
tual parent of economics is that Darwin’s definition of com-
petition includes cases where it is good for the individual but
bad for society and where it is good for both, whereas Smith’s
definition of competition (dumbed-down by libertarians)
is good for the individual and good for society in all cases.
Why does it matter? If the libertarians are right, there is no
need for government regulation. If the Darwinians are right,
then regulation can improve social outcomes. If the libertar-
ians (read mainstream economists) are right, then absolute
quantities (as in absolute income and absolute consumption)
matter, not relative. If the Darwinians are right, then relative
ranking and ­relative strength matter, not absolute. Without
going into detail, prisoner’s dilemmas and arms races are cases
where individual rational choices doom the individual or the
society. Only collective control can improve the outcome. The
mainstream theory of consumption developed by Friedman
as well as by Modigliani and Brumberg is based on rational
actors making choices about absolute income and absolute
consumption. But evidence indicates that real actors respond
to relative income and relative consumption.

Duesenberry and the Behavioral Economists

James Duesenberry’s book Income, Saving and the Theory of
Consumer Behavior was published in 1949, eight years before
Friedman’s Theory of the Consumption Function. The key point
he argues there is that consumer preferences are not indepen­
dent, as assumed in the neoclassical model of consumer behav-
Consumption Theory and Its Critics 103

ior ascribed to by Friedman as well as Modigliani and Brum-

berg. That tenet of independence is held in spite of evidence
from psychology and sociology to the contrary. Recall that Veb-
len, too, had criticized economics for ignoring the evidence of
other social sciences.57 Duesenberry cites facts to support his
argument that preferences are interdependent. For example,
“Thirty years ago [1911] the average urban family with a $1,500
income in 1940 prices saved 8 percent of its income. In 1941 a
similarly placed family saved nothing. One can hardly argue
that the desire for saving, whatever its source, had diminished
in that period. For some reason, the forces leading to higher
consumption increased during that period.”58 He stressed that
the “fact that the attainment of a higher standard of living as
an end in itself is a major social goal has great significance
for the theory of consumption.” The utility index of a person
“depends not on the absolute level of his consumption, but
on the ratio of his expenditures to those of other people. . . .
Current consumption standards or desires are influenced by
other peoples’ consumption behavior.” If preferences are inter-
dependent, as he argues, then “low income groups are affected
by the consumption of high income groups but not vice versa.”
That point was made by Veblen in 1899.
Relying on data for household saving, Duesenberry
argued that the rational saver of the neoclassical model was
not to be found among most consumers. “About 75 percent of
spending units save virtually nothing. Most of these families
will not be influenced in their saving by changes in interest
rates, income expectations, or even by changes in their own
preference parameters. . . . At (relatively) low levels of income,
desires for current consumption are so strong that they over-
come all considerations of the future.” And further on saving,
“The propensity to save of an individual can be regarded as a
104 Consumption Theory and Its Critics

rising function of his percentile position in the income distri-

bution.”
In the conclusion, Duesenberry states, “The theory of
saving just summarized is based on the assumption that con-
sumers’ preferences are interdependent and irreversible. Our
theory of the relation between income and saving really de-
pends on the validity of a single hypothesis, viz: that the utility
index is a function of relative rather than absolute consump-
tion expenditure.”59 His views (1) that preferences are interde-
pendent and (2) that one’s utility or satisfaction or happiness
depends on relative consumption fly in the face of mainstream
neoclassical consumption theory. It is an early expression of
the behavioral economics view on consumption. That is why
noted behavioral economist Robert Frank has labeled Duesen-
berry the first behavioral economist. About the reception of
economists to Duesenberry’s views, Frank has written: “Still, it
is perhaps an understatement to say that the economics profes-
sion as a whole has shown little interest in the idea that p
­ eople
are deeply concerned about their relative standing in hierar-
chies. Duesenberry’s theory of consumption behavior, for ex-
ample, was quickly relegated to history-­footnote status as soon
as alternative theories appeared.”60 A recent econometric anal-
ysis provides empirical support to Duesenberry’s claim that
consumer preferences are interdependent. “I conclude that the
negative effect of neighbors’ earnings on well-being is real and
that it is most likely caused by a psychological externality, that
is, people having utility functions that depend on relative con-
sumption in addition to absolute consumption.”61
The optimizing rational consumer of the Friedman as
well as the Modigliani and Brumberg models, models that have
become the mainstream for consumption theory, are devoid of
motives and proclivities we recognize as broadly shared, such
Consumption Theory and Its Critics 105

as maintaining status in the perceived peer group, determining

to save more, to lose weight, to stop smoking, next year but not
now. A model that purports to explain behavior of consumers
in facing choices about consumption and saving ignores what
we all know about human behavior. Behavioral economist
Robert Frank and his co-author Seth Levine do not ignore
such behavior in their analysis of consumption. “Changes in
one group’s spending shifts the frame of reference that defines
consumption standards for others just below them on the in-
come scale, giving rise to expenditure cascades.” They develop
a model based on Duesenberry’s relative income hypothesis.
Using data for fifty states and the one hundred most popu-
lous counties, they find “evidence that rapid income growth
concentrated among top earners in recent decades has stimu-
lated a cascade of additional expenditure by those with lower
earnings.”62 They argue that such cascades caused the drop in
savings shown in Figure 4.1.
In another work, Frank criticizes Friedman’s emphasis
on absolute rather than relative income. He argues that empir-
ical work shows that people care about their relative economic
standing, not their absolute income.

Strangest of all, Friedman’s theory assumes that con-

text has absolutely no effect on judgements about liv-
ing standards. . . . In light of abundant evidence that
context matters, it seems fair to say that Duesen-
berry’s theory rests on a more realistic model of
human nature than Friedman’s. . . . Under the rel-
ative income hypothesis, for example, it is easy to
understand why a majority of families experience
significant retrenchments in living standards when
they retire. Under the permanent income hypoth-
106 Consumption Theory and Its Critics

esis, this observation is a jarring anomaly. And yet

Duesenberry’s relative income hypothesis is no
longer even mentioned in leading textbooks.63

Another notable behavioral economist who has criticized

mainstream consumption theory is George Akerlof. Incorpo-
rating true behavior, as George Akerlof asserted in his Nobel
acceptance speech, “Macroeconomics would then no longer
suffer from the ‘ad hockery’ of the neoclassical synthesis,
which had overridden the emphasis in The General Theory
on the role of psychological and sociological factors, such as
cognitive bias, reciprocity, fairness, herding, and social status.
My dream was to strengthen macroeconomic theory by in-
corporating assumptions honed to the observations of such
behavior.”64
Akerlof points out that that one widely known phenom-
ena the New Classical economics (rigorous successor to the
neoclassical synthesis), dominant now in macroeconomics,
cannot explain is chronic undersaving for retirement. “In the
New Classical model, individuals decide how much to con-
sume and to save to maximize an intertemporal utility func-
tion. The consequence is that privately determined saving
should be just about optimal. But individuals commonly re-
port disappointment with their saving behavior and, absent
social insurance programs, it is widely believed that most peo-
ple would undersave. ‘Forced saving’ programs are extremely
popular.” In his conclusion, Akerlof anoints Keynes as a be-
havioral economist.

Keynes’ General Theory was the greatest contri-

bution to behavioral economics before the present
era. Almost everywhere Keynes blamed market
Consumption Theory and Its Critics 107

failures on psychological propensities (as in con-

sumption) and irrationalities (as in stock market
speculation). Immediately after its publication, the
economics profession tamed Keynesian econom-
ics. They domesticated it as they translated it into
the “smooth” mathematics of classical economics.
But economies, like lions, are wild and dangerous.65

Is the APC Really Rising?

If the permanent income hypothesis is accepted, how would
we interpret Figure 5.1, the actual APC for the United States
over the past four decades? The APC rose from 0.88 in 1985
to a peak of 0.95 in 2005. That conforms to the theoretical
case where current income falls below permanent income, but
the actual case can hardly be described as “temporary” with
the upward trend persisting for two decades. It could be ob-
jected that the APC of Figure 5.1 is calculated from annual real
disposable income and consumption rather than Friedman’s
notion of permanent income. An approximation of the APC
from permanent income frequently used is a four- or five-year
moving average of income and consumption. Comparing a
four-year moving average of the APC with the annual APC
of Figure 5.1 does not show much difference. Instead of rising
+0.068 from 1985 to 2005, the moving average rises +0.064.
The long-term rise of the APC does not disappear when one
uses an approximation of permanent income.
If that rise of APC is claimed to reflect increased income
inequality, Friedman would not have agreed. “Empirical data
show no tendency for inequality of income to increase. If any-
thing, inequality seems to have been decreasing in recent de­-
cades.”66 Of course the recent long-term rise of income in-
108 Consumption Theory and Its Critics

equality was not under way when he wrote. Also, the drop
in the APC from 0.95 in 2005 to just under 0.91 in 2009 im-
plies that current income temporarily rose above permanent
income, an unlikely scenario in this recent severe recession.
Chapter VI returns to this issue in reviewing Kuznets’s long-
term data on the APC.
Mainstream macroeconomists have certainly noticed
the observed rise of the APC, and the falling saving rate, since
the mid-1980s. Most of the attention in the literature has been
focused upon the falling rate of saving. For example, in an
International Monetary Fund working paper addressing the
decline of the saving rate, the authors identify causal factors
that are all on the supply side: decreasing interest rates, finan-
cial liberalization, and increased housing liquidity.67 No de-
mand-side factors are considered. But the interesting question
is not why the saving rate declined, but why the APC rose. The
arguments of Duesenberry and Bunting noted above suggest
that for most consumers, saving is a residual, the tail wagged
by the consumption dog.
The question of whether the APC (now referred to in the
literature as the consumption-income ratio) has been r­ eally ris-
ing turns on the technical issue of whether the ratio is stationary
(no trend) or nonstationary (a trend). According to Friedman,
the ratio is stationary—that is, the APC is stable over long pe-
riods. The only rigorous test for stationarity of a time-series
variable is what is called a unit root test in econometrics. In
one article, Steven Cook calculates different unit root tests
than others have used for twenty OECD nations. He finds
support for a stationary consumption-income ratio (APC) in
all twenty nations. In another article, the authors, Eftymios
Tsionas and Dmitris Christopoulos, calculate unit root tests
for fourteen European nations’ consumption-­income ratios.
Consumption Theory and Its Critics 109

Eight of the fourteen turn out to be nonstationary (rejecting

the Friedman claim), while six turn out to be stationary (not
rejecting the Friedman claim).68 The United States was not in-
cluded in either study.
We have performed a unit root test for the APC data for
the United States, 1967–2013, represented in Figure 5.1. The test
we used is the common augmented Dickey-Fuller test. The test
results indicate that the APC is nonstationary (rejecting the
Friedman claim). That is, the test results support the view that
the APC does have a trend. Our eyes did not fool us in this
case.

Outline of a Revised Theory of Consumption

The analysis of aggregate consumption should begin with
recognition of the centrality of income distribution for long-
term saving and consumption. As shown above, the APC and,
therefore, the saving rate can have long periods in which they
are not stable: the APC moves up and the saving rate moves
down. The reason that can happen and persist is that consum-
ers choose to maintain consumption when confronted with
rising income inequality.69 They can succeed in that effort by
reducing saving and increasing borrowing. On borrowing,
Rajan argues that the government eases restrictions on bor-
rowing as an offset to rising income inequality.70 In a sensible
world, consumer borrowing would be directed to buying as-
sets expected to appreciate and thus aid in maintaining their
consumption. That is exactly what they did, as explained in
Chapter IV on debt (see Tables 4.5 and 4.6). The Modigliani
and Brumberg APC equation, (4), shows that APC can rise if
wealth grows faster than income, which it did for most of the
period 1995–2007 (Figure 5.2). But maintaining consumption
110 Consumption Theory and Its Critics

through reduced saving and increased borrowing to acquire

appreciating assets is not sustainable in the long run. The rapid
rise in wealth can be the result of a speculative bubble, as in
the high-tech boom and in the housing boom. A revised the-
ory of consumption must take such factors into consideration.
It must also replace the representative rational consumer of
Friedman, Modigliani, and Brumberg facing complete mar-
kets (insurance for every risk), with a realistic distribution of
consumers with heterogeneous preferences by income, age,
race, and so on subject to rational choices (buying appreciat-
ing assets early) as well as irrational choices (buying appreci-
ating assets late) and facing incomplete markets (some risks
not insurable). Looking back to Keynes and to behavioral eco-
nomics can inform the effort to develop a revised theory of
consumption. A revised theory may be messy, inelegant, and
ugly. The current theory is tidy, elegant, and beautiful, but it is
only true some of the time.
 VI
Has This Happened Before?

The Great Depression and the Great Recession were both caused
by policies derived from nostalgia for the world of the Enlighten-
ment. Drawing on theories from the eighteenth century, hard-
headed policy-makers either assumed or tried to re-create the
idealized conditions described by Hume and Smith.

—Peter Temin (2010)

T
he argument up to now is that the rise of income
inequality beginning around the mid-1970s brought
on a decline in saving (and thus a rising APC) and
eventually a surge in borrowing by households. The
surge in borrowing was prompted by demand-side forces
(stagnant household income and perceived wealth enhance-
ment because of the housing bubble) and supply-side forces
(relaxation of credit standards and low interest rates). With
the bursting of the housing price bubble, the financial sys-
tem almost collapsed, foreclosures ensued, many mortgage
112 Has This Happened Before?

borrowers owed more than the value of their properties, and

credit conditions were tightened by lenders even though in-
terest rates remained low. The outcome to date is that house-
holds are deleveraging through saving more, borrowing less,
and thus curbing their consumption. The rate of growth of real
personal consumption expenditures since the official end of
the Great Recession in the second quarter of 2009 through the
fourth quarter of 2013 is a little over 2 percent per year. That
compares with consumption growth averaging near 4 percent
annually during the 1980s and 1990s. If deleveraging continues
for some years, economic growth will be curtailed.
Has rising income inequality or rising household debt
or a housing price bubble played a role in serious economic
decline in the past? Did a rising average propensity to con-
sume, and thus a falling rate of saving, warn of unsustainable
consumption leading to a serious economic decline before the
present recession? Those are the questions considered in this
chapter, and they all receive the same answer: yes.

Rising Income Inequality in the 1920s

According to Marriner Eccles, U.S. Treasury Secretary during
the Roosevelt administration, one cause of the Great Depres-
sion was income inequality because it led to an unsustainable
rise of debt. “The stimulation to spending by debt-creation of
this sort was short-lived and could not be counted on to sustain
high levels of employment for long periods of time. Had there
been a better distribution of the current income from the na-
tional product . . . we should have had far greater stability in our
economy.”1 He wrote those words around 1951, the year his auto-
biography (Beckoning Frontiers) was published. That is the the-
sis about the Great Recession laid out in the preceding chapters.
Has This Happened Before? 113

Eccles was not alone in naming income inequality as

a cause of the Great Depression. In his 1954 book The Great
Crash, John Kenneth Galbraith lists “five weaknesses [that]
seem to have had an especially intimate bearing on the en-
suing disaster.” First on his list is the “bad distribution of in-
come.”2 He recognized that high-income people were most
directly affected by the financial crash of 1929 because they
owned almost all of the financial wealth. From the peak in Oc-
tober 1929, the Dow-Jones industrial average plummeted more
than 80 percent to its low in late 1932. With so many investors’
financial assets wiped out, their major cutbacks in spending
spilled down to adversely affect lower-income households.
Table 6.1 presents three estimates of income shares from
1917 through 1939. The first two (Piketty and Saez, Kuznets)
show the share of the bottom 95 percent, while the third (Na-
tional Industrial Conference Board [NICB]) shows the share
of the bottom 90 percent. Both the Piketty and Saez series
and the Kuznets series show shares for every year from 1917
through 1939. In both series, 1920 is the peak year of income
share for the bottom 95 percent for all years shown. Thereafter
the share of the bottom 95 percent mostly declines, hitting a
low of 65.2 percent in 1928 (Piketty and Saez) and 73.2 percent
also in 1928 (Kuznets). The drop from the peak in 1920 is 7.3
percentage points in the Piketty and Saez series and a smaller
4.7 percentage points drop in the Kuznets series. Thereafter
both series rise, but not steadily, through 1939. The point em-
phasized here is that there was a significant rise of income in-
equality in the 1920s—that is, the share of income going to
the bottom 95 percent dropped from seven to five percentage
points from 1920 to 1928.
The NICB series for selected years shows a different pat-
tern. The peak share for the bottom 90 percent is 65.5 percent
 Table 6.1. Incomea Distribution, 1917–1939, Three Sources
Piketty & Saez Kuznets NICB
Selected years Bottom 95% Bottom 95% Bottom 90%
1917 69.7 75.4
1918 70.7 77.3 65.5
1919 70.7 77.1
1920 72.5 77.9
1921 69.5 74.5 61.8
1922 69.0 75.2
1923 71.1 77.1
1924 69.1 75.7
1925 67.5 74.8
1926 67.3 74.8
1927 66.6 74.0
1928 65.2 73.2
1929 67.0 73.9 61.0
1930 68.8 73.8
1931 67.0 73.7
1932 67.4 73.3
1933 67.5 74.7
1934 67.0 75.1 66.4
1935 69.0 76.3
1936 67.4 75.7
1937 68.6 76.2 65.6
1938 69.8 77.0
1939 68.7 77.2
a
Excludes capital gains.
Sources: Alvaredo, Atkinson, Piketty, and Saez (2011); Kuznets (1946); U.S. Census
Bureau (1975), series G337–352, and series A176–194, p. 302; National Industrial
Conference Board (1949), Series A176–194, p. 15.
Has This Happened Before? 115

in 1918. Thereafter it is lower through 1929. Thus the NICB se-

ries agrees in direction of change with the other two series,
falling almost five percentage points from 1918 to 1929. But the
NICB drop in share, 1921–29, is less than one percentage point.
By 1934 the NICB share is higher than its value in 1918: 65.5
percent. Yet all three series show increases of inequality in the
1920s, although slight in the NICB data.
But the rise of income inequality from 1920 to 1928,
documented by Kuznets, was not noted by Friedman.3 The
Kuznets data on income distribution matches pretty well the
data on income distribution assembled by Piketty, Saez, and
colleagues from income tax returns for the period 1917–2002.4
In a paper on income inequality and poverty in the twen-
tieth century, Eugene Smolensky and Robert Plotnick cite Jef-
frey Williamson and Peter Lindert, who utilized the Kuznets
data, income tax data, and skilled–unskilled wage ratios.5 “The
chronology of income inequality suggested by this assortment
of time series is as follows. From the turn of the century until
World War I, inequality was higher than in the latter half of
the century. The war had a brief equalizing effect. Starting
about 1920 inequality began to rise, reaching its pre–World
War I level by 1929. From 1929 through 1951, inequality fell
substantially.”6 The data of Table 6.1 and the observations by
Smolensky and Plotkin about the analysis by Williamson and
Lindert are in part at odds with the view of the labor econ-
omists Claudia Goldin and Lawrence Katz, who find “that
the wage structure and the returns to education and skill all
moved in the direction of greater equality considerably before
the better known ‘Great Compression’ of the 1940s. The wage
structure narrowed, skill differentials were reduced, and the re-
turn to education decreased sometime between 1890 and 1940,
most likely in the late 1910s.”7 The part that is at odds with the
116 Has This Happened Before?

data of Table 6.1 and the earlier studies cited above is that the
“Great Compression” began “most likely in the late 1910s.”
But Table 6.1 shows that income inequality rose for parts of
that period, 1920–28, according to the Piketty and Saez or the
Kuznets series, and 1910–29 for the NICB series. Further, the
description by Smolensky and Plotnick of Williamson’s and
Lindert’s analysis claims that “starting about 1920 inequality
began to rise, reaching its pre–World War I level by 1929.”8 The
evidence shows that the turnaround from rising inequality to
falling inequality did not begin until the 1930s.
Most of the analysis by Goldin and Katz relates to wage
ratios for skilled and unskilled labor or professionals com-
pared with manufacturing workers. It is true that their data
on male workers’ wage ratios in manufacturing for two years,
1890 and 1940, by various industries, show wage compression.
However, their annual earnings ratios of college professors to
average workers in manufacturing and average low-skilled
workers in manufacturing for each year from 1908 to 1960
present a picture consistent with the data on income shares in
Table 6.1. That is, although the ratios drop dramatically from
1908 to 1920, they then rise from 1920 through 1932. In some
cases, the ratios of professors’ earnings to those of low-skilled
manufacturing workers were higher in 1932 than in 1908. The
earnings ratios for all clerical workers, females, and produc-
tion workers in manufacturing do not show as sharp a reversal
of trend in the 1920s as with the professors.
Turning to income shares rather than earnings ratios,
Goldin and Katz note, based on the Kuznets data and Gold-
smith’s 1967 and 1954 extensions, that “the share of income
received by the top 5 percent of families (consumer units) de-
clined from 30.0 percent in 1929 to 25.8 percent in 1939 to 20.9
percent in 1947. The important point here is that the Kuznets
Has This Happened Before? 117

series does not rise during the 1930s. Rather the top portion of
the income distribution narrowed during the Great Depres-
sion. The Kuznets data, therefore, reveal nothing particularly
unusual about the late 1930s in comparison with the 1920s.”9
This quote agrees with the data presented in Table 6.1, except
for the last sentence. As noted above, the three income share
series presented in Table 6.1 show a declining share of income
for the bottom 95 or 90 percent (thus a rising share for the top
5 or 10 percent) in the 1920s. That trend is reversed beginning
in the early 1930s. Thus there is something noteworthy about
the 1930s compared with the 1920s—namely, income inequal-
ity rose in the 1920s and began to fall in the 1930s.
The scenario of rising inequality in the 1920s followed
by declining inequality in the 1930s and beyond is noted by
Margo in his review of the inequality analysis by Williamson
and Lindert. “Wage inequality drifted upwards after 1860,
peaking at some point in the late 1920s, after experiencing a
sharp, but transitory, decline during World War One. Inequal-
ity fell sharply and more or less continuously between 1929
and 1950.”10
One likely reason for the sharp but short-lived drop of in-
come inequality brought on by World War I is that agricultural
prices exploded upward, a typical wartime phenomena. From
a prewar wholesale price index (1967 = 100) of 43.7 in 1913, it
shot up to a peak of 96.4 in 1919. The agricultural sector was
far more important for the U.S. economy back then, account-
ing for 23 percent of national income and 27 percent of total
employment in 1919. The average monthly wage for farm labor
rose 141 percent from 1910 to 1920. But the end of the war and
a national recession in the early 1920s brought on a collapse in
agriculture. National income from agriculture plummeted 45
percent in only three years—1919 to 1922. The wholesale price
118 Has This Happened Before?

index for farm products dropped 40 percent in that period,

and the average monthly farm wage fell more than one-third.11
With farm employment hardly changed in those three years,
the effect on farm family annual incomes must have been se-
vere. And given the relatively large size of the farm sector back
then, the rising trend of income inequality from 1920 to 1929,
shown in Table 6.1, is hardly surprising. Another piece of evi-
dence supporting rising income inequality in the 1920s is the
reversal of long-run state income convergence shown by Barro
and Sala-i-Martin. Their measure of dispersion among states’
per capita incomes, after declining from 1870 to 1920 (declines
mean state incomes are growing closer together, as the neo-
classical model predicts), turned up in the 1920s (rises mean
state incomes are growing further apart, which can be a source
of rising national income inequality, as noted in Chapter II).12
It may not be a coincidence or an aberration that the shift from
spatial convergence of state incomes to divergence, from the
1980s forward, includes a period of national rising income in-
equality, as does the 1920s.
But did the collapse in agriculture have a negative im-
pact upon urban incomes, especially of the less skilled? Prob-
ably so, because migration of farm families to urban areas
increased the supply of such labor. The net loss of farm pop-
ulation through migration was 6.3 million from 1920 to 1929.
That is much higher than the net loss to migration in the 1930s
of 3.8 million.13

The Long Rise of the Average

Propensity to Consume
Simon Kuznets developed national income and consump-
tion data for the United States back to 1869. In Table 6.2,
Has This Happened Before? 119

what Kuznets calls “flow of goods to consumers” as a share

of national income is labeled the APC because that is what
he sought to measure in the early days of national income ac-
counts. Those data are ten-year averages. In the three decades
from 1884–93 to 1914–23, the APC rose 4.7 percentage points.
In the next decade, it rose a further 5.4 percentage points. So
the gain over four decades was ten percentage points, larger
than the rise of almost seven percentage points from 1984 to the
peak in 2005, shown in Figure 5.1 of Chapter V. Although the
rise was larger, it took longer to develop. Nonetheless, the data
for both periods contradict Friedman’s argument that over the
long run, the APC is stable.
“Estimates of savings in the United States made by
Kuznets for the period since 1899 revealed no rise in the per-
centage of income saved over the past half-century despite a
substantial rise in real income. According to his estimates,
the percentage of income saved was much the same over the
whole of the period.”14 It is true that the inferred saving rates,
1 – APC, of Kuznets in Table 6.2 show “no rise in the percent-
age of income saved over the past half-century,” but they do
show a fall in the saving rate from 16.1 percent in 1884–93 to
6.0 percent in 1924–33. Friedman does not note the specific
ten-year APC estimates of Kuznets referred to above.
Friedman does show estimates of the APC by Raymond
Goldsmith from time-series data for the United States for
periods of varying length. The longest, 1897–1949, has an es-
timated APC of 0.88. All the shorter period estimates range
from 0.87 to 0.89, except for one: 1929–41. For that period,
estimated APC is 0.94, six points above the estimate for 1897–
1949.15 Note that the outlier in Kuznets’s APC averages is for
1924–33, partly the same period, shown in Table 6.2. So the
remarkable stability of Goldsmith’s APCs must be qualified. If
120 Has This Happened Before?

the APC is much above the long-term average for ten to thirteen
years, then the saving rate was not “much the same over the
whole period.”16
Which estimates of the APC, Kuznets’s or Goldsmith’s,
better agree with Friedman’s assertion of stability? In order to
compare them, from the annual data Goldsmith presents in his
three-volume book A Study of Saving in the United States (1955),
the APC values have been calculated for the years that corre-
spond to Kuznets’s ten-year periods in Table 6.2 and are shown
as well. The point to take away from this is that the Goldsmith
APC values do not show a steady long-term rise of the APC
(and fall of the saving rate) from 1894–1903 through 1924–33.
Rather, the Goldsmith figures show an APC rise over a shorter
period: 1914–23 through 1924–33. Thus the Goldsmith figures
that Friedman presented are more agree­able with Friedman’s
hypothesis of a constant saving rate and a stable APC than are
the Kuznets figures.
Kuznets’s data show a long period of rising APC in the
first half of the twentieth century that also includes a period
of rising income inequality. The current data show that in-
come inequality has been rising for around thirty years (see
Table 2.2) and that the APC has been rising for much of that
same period—1984 to 2005. Kuznets’s data show that income
inequality rose from 1920 to 1928 (as do the Piketty and Saez
data) and that the APC rose for the four decades after 1884–93.
The fact that those two events overlap twice in our economic
history may not be a coincidence. The rough temporal coin-
cidence of rising APC and increasing income inequality for
two long periods cannot be explained with Friedman’s theory
of consumption. Also, the rise of income inequality during
1920–28 shown in Table 6.1 was preceded by a long rise of in-
come inequality briefly interrupted by World War I. That ear-
Has This Happened Before? 121

Table 6.2. Data on Average Propensity to Consume

(APC) and Inferred Saving Rate, Kuznets (1874–1933) and
Goldsmith (1897–1933)
Kuznets Goldsmith
Average Inferred Average Inferred
Period APC saving rate APC saving rate
1874–83 85.6% 14.4% N/A N/A
1884–93 83.9% 16.1% N/A N/A
1894–1903 85.2% 14.8% 90.9%a 9.1%a
1904–13 86.9% 13.1% 90.3% 9.7%
1914–23 88.6% 11.4% 88.1% 11.9%
1924–33 94.0% 6.0% 93.4% 6.6%
a
1897–1903
Source: Author’s compilation from Kuznets (1946); APC averages from Table 16, p.
53. Inferred saving rate calculated as 1 – APC. Author’s compilation from Goldsmith
(1955): disposable income in current dollars (Vol. III, Table N–3, columns 7 and 8,
p. 431) less personal saving in current dollars (Vol. I, Table T–6, columns 1–5) equals
consumption in current dollars.

lier trend is documented by Williamson and Lindert. “Both

measures [income tax data from top income groups] show
peak inequalities on the eve of America’s entry into World War
I and again just before the Great Crash.”17

The Rise of Household Debt

The evidence presented here shows that income inequality rose
during the 1920s. Chapter IV argues that the huge run-up in
household debt prior to the Great Recession was in part a re-
sponse to rising inequality that depressed household income
growth. Did debt rise during the 1920s? Yes. Figure 6.1 shows
aggregate household debt to GDP for the period 1920–31. Figure
122 Has This Happened Before?

Figure 6.1. Debt to GDP ratio and share of top 5 percent in income
distribution, 1920–31. Sources: Data for debt and GNP from U.S.
Census Bureau (1975), tables on pp. 289 and 224. The share of top 5
percent in income distribution is from Saez (2013).

6.2 shows the same for the period 1983–2012. Figures 6.1 and 6.2
are duplicates of two graphs shown in “Inequality, Leverage and
Crises: The Case of Endogenous Default” by Kumhof, Ranciere,
and Winant. The authors note about this comparison:

In the periods prior to both major crises, rapidly

growing income inequality was accompanied by a
Has This Happened Before? 123

Figure 6.2. Debt to GDP ratio and share of top 5 percent in income
distribution, 1984–2012. Sources: Debt is mortgage debt and con-
sumer credit of households from Federal Reserve Board (2014a),
Table L100. GDP is from Council of Economic Advisers (2013).
Share of top 5 percent in income distribution is from Saez (2013).

sharp increase in aggregate household debt. . . . Be-

tween 1920 and 1928, the top 5% income share in-
creased from 27.4% to 34.8%. During the same pe-
riod, the ratio of household debt to GDP more than
doubled. . . . Between 1983 and 2007 income in­
equality experienced a sharp increase as the share
of total income commanded by the top 5% of the
124 Has This Happened Before?

income distribution increased from 21.8% in 1983

to 33.8% in 2007.18

Two other IMF economists argue that the recent rise of

income inequality in the United States is similar to such a rise
in the 1920s. “In both cases there was a boom in the financial
sector, poor people borrowed a lot, and a huge financial crisis
ensued.” And further, “The recent global economic crisis, with
its roots in U.S. financial markets, may have resulted, in part
at least, from the increase in inequality.” Then they conclude,
“It would be a big mistake to separate analyses of growth and
income distribution.”19
Another similarity is the trend in mortgage debt. Figure
6.3 shows residential mortgage debt outstanding from 1910 to
1934. Also shown is the price index of single-family houses for
twenty-two cities, based on sales for the same period. In the
ten years from 1910 to 1920, mortgage debt doubled, and in the
next ten years, 1920 to 1930, it more than tripled. The year 1930
was the peak year for mortgage debt, reaching 30.2 billion. It
was not until 1947 that the 1930 peak was surpassed.
House prices rose 38 percent from 1910 to 1920, slowing
somewhat after that. They peaked in 1925, edged down more to
1927, and then post 1928 went into a freefall through 1933. The
price level in 1934 was down to the 1914 level (Figure 6.3). As
described by Hockett:

When real estate prices leveled off and then began

falling in 1928, however, short-term mortgages
could no longer be refinanced in full. Again, things
were much as they are today. Resultant forced sales
and foreclosures, which reached the rate of over
1,000 per day once some 50 % of all home mort-
Has This Happened Before? 125

Figure 6.3. Mortgage debt outstanding and house prices, 1910–34.

The dashed line shows residential mortgage debt in billions of
dollars. Solid line: Price index of single family houses in twenty-two
cities. Source: U.S. Census Bureau (1975), House Price Indexes,
1890–1947, Series N 259–61; residential mortgage debt outstanding,
1890–1947, Series N 262–72.

gages in the country had gone into default, brought

prices steadily lower. The real estate market fell into
the familiar “downward spiral.” The parallel with
today could not be more striking.20

As Hockett points out, one reason house prices fell so

much was that mortgage finance was rather short term in
the early part of the twentieth century, with down payments
typically more than 50 percent of value. Once house prices
126 Has This Happened Before?

stopped soaring, many homeowners could not refinance, the

usual way to extend one’s payback period. The thirty-year fully
amortized mortgage was in fact a development prompted by
the 1920s real estate collapse. Both the final year of the Hoover
administration and the Roosevelt New Deal were responsible
for bringing stability to mortgage finance with such improve-
ments.21
Figure 6.4 presents mortgage debt and house prices for
the recent past, 1991 to 2012. The growth of mortgage debt ac-
celerated post 1995 to the peak in 2007. However, that recent
expansionary period was not as strong as the tripling of mort-
gage debt from 1920 to 1930, shown in Figure 6.3. The rise of
house prices was more rapid in the 1990s and 2000s than it
was during the 1910 to 1925 period. The run-up of mortgage
debt is greater in the earlier period, whereas the rise of house
prices before the Great Depression was less sharp. Nonethe-
less, with 50 percent of all mortgages in the nation in default,22
the real estate market collapse back then resembles the recent
burst of the housing price bubble.
Non-mortgage consumer debt increased by a factor of
two in the 1920s. Its peak of $7.6 million in 1929 was not ex-
ceeded for ten years thereafter. Although that seems a trivial
amount, consumer credit terms were quite strict and default
more costly for the debtor than today.23 Krugman links in-
come inequality to consumer debt in our current recession and
speculates that “a return to pre-Depression levels of inequality
. . . followed by a return to depression economics could be
just a coincidence. Or it could reflect common causes of both
phenomena.”24 After rejecting the underconsumption story of
the current recession. Krugman argues, “A better case can be
made for the opposite proposition—that rising in­equality has
led to too much consumption rather than too little and, more
Has This Happened Before? 127

Figure 6.4. Mortgage liabilities and house prices. Dotted line: Mort-
gage liabilities, households, and nonprofit organizations. Solid line:
House Price Index, fourth quarter 1991 = 100. Sources: House price
index data from Federal Housing Finance Agency (2014). Mortgage
liabilities from Federal Reserve Board (2014a), Table L100, p. 61,
March 2014.

specifically, that the widening gaps in income have caused

those left behind to take on too much debt.”25
To what extent, then, was the period leading up to the
Great Depression similar to that leading up to the Great Re-
cession? They are similar in the rise of income inequality and
in household debt, especially mortgage debt. It is true, though,
that income inequality had been rising for more than thirty
years in the current period before the debt burden brought
down the economy in 2008, whereas income inequality had
been rising for less than ten years when the debt burden con-
tributed to the crash in 1929. The likely reason debt became
unsustainable so much sooner in the 1920s is that credit condi-
128 Has This Happened Before?

tions were far stricter back then.26 The two periods are similar
in the long-term rise of the APC and in the replacement of
state income convergence with state income divergence. Also,
the earlier period had a housing price bubble that burst in 1928.
Rising income inequality was a cause of rising household
debt to unsustainable levels, which was an important factor in
bringing on the Great Recession. As we have seen, rising in-
come inequality in the 1920s was accompanied by rising mort-
gage debt. The regression results presented support the claim
for the recent period. There is no statistical evidence linking
rising income inequality to rising debt for the earlier period,
but the historical evidence presented in this chapter does lend
some support to the thesis of a causal link.
 VII
Conclusion

I
t is clear that rising income inequality has had delete-
rious effects upon household debt and saving. The over-
hang of debt and revived saving by households post 2007 will
be a drag on economic expansion for some years. The
adverse aftereffects of the Great Recession, although officially
ended in June 2009, will be much longer than those of reces-
sions of the recent past. From the third quarter of 2009, the first
quarter of official recovery from the Great Recession, through
the fourth quarter of 2014, real consumption expenditures
increased only 2.2 percent on average. That is well below the
recent historical experience of nearly 4 percent.1 Rather than
conclude with views on how to jump-start or at least improve
our limping economy, some authors are noted who have done
a good job of prescribing policies for economic recovery. The
book The Way Forward: Moving from the Post-Bubble Post-Bust
Economy to Renewed Growth and Competitiveness (2011) by
Daniel Alpert, Robert Hockett, and Nouriel Roubini lays out
an impressive plan. Roubini was one of the few economists,
Shiller another, who foresaw the financial crash. Paul Krug-
man’s book End This Depression Now! (2012) also tells how to
130 Conclusion

do it.2 It is not rocket science. First-year students in macro­-

economics courses can design solutions. When unemploy-
ment is high, inflation is low, and interest rates near zero have
no stimulative effect, it is clear that effective demand from con-
sumers, businesses, and exporters is insufficient to restore full
employment. Therefore the federal government must boost
effective demand through deficit spending on massive infra-
structure projects that have immediate and large employment
effects. Tax cuts will not do in a high unemployment scenario
because the unemployed are not paying any taxes, and so they
gain no spending power. Those that do gain from tax cuts might
save some or all, fearing times will get worse, or they might spend
some of the first-round stimulus on imports. That will not help
near-term recovery much.
In addition to telling how to bring on economic recov-
ery, Krugman explains why a big short-term deficit to finance
expansion of effective demand will not raise the long-term
debt nearly as much as an austerity program such as the one
imposed in the United Kingdom. That is because the rise of
federal tax revenues in a full economic recovery can offset
the deficit’s effect on debt. We should have learned that after
World War II. The heavy wartime borrowing by the govern-
ment pushed the debt well above GDP, but the postwar eco-
nomic boom brought the debt back down below GDP. The
absolute debt may not have changed much, but its burden on
the economy was reduced. He rightly labels inflation as “the
phantom menace.”3 What faces us now is a higher danger of
deflation than of inflation in such a slack economy.
Going beyond an easy-to-design short-term recovery
program, the more daunting issue is how to reverse or at least
stop the decades-long rise of income inequality, which, as ar-
gued here, has been a major cause of the Great Recession and
Conclusion 131

our sluggish recovery. One thing is clear: the cause is mostly

not economic—it is mostly political. So the solution, if ever
devised, will be mostly political. Another financial and eco-
nomic crash just as bad or worse than the last might focus
Congress on correcting causes of income inequality.
The emphasis now is on what is new here. As noted in
the Introduction and in Chapter V, the mainstream theory of
consumption does not accord any importance to the distribu-
tion of income. But that theory cannot explain recent trends
in relative consumption and saving. This analysis shows that
rising income inequality was central to what has happened
to relative consumption (the long-term rise of the APC) and
consumer indebtedness. The state panel regressions of Chap-
ter IV show that household debt rose with rising income in-
equality. The income quintile panel regressions show that the
net change in household liabilities is positively related to in-
creases of income inequality. The Consumer Expenditure Sur-
vey data show large gains in the relative share of consumer
spending on shelter, health, and education while the share of
non-housing necessities shrank. That is the case across all in-
come quintiles. As argued in Chapter V, a revised theory of
consumption should be developed that affords a central place
to income distribution and household debt as well as to Modi-
gliani and Brumberg’s wealth-to-income ratio and to the per-
suasive insights of behavioral economics. To many, historical
evidence carries more weight than econometric results, and to
them, the evidence presented in Chapter VI may be the most
persuasive evidence that rising income inequality can lead
to unsustainable household debt that brings on a severe eco-
nomic crash. The year 2008 looks like 1929.
Finally, I present a plea for a more fact-based econom-
ics than an authority-based economics. In the 1960s, when it
132 Conclusion

became clear that indeed the APC, and of course the saving
rate, were stable despite rapid income growth, Keynes’s con-
cern that income growth would lower the APC (and thus raise
the saving rate) was jettisoned by macroeconomists. That was
the right choice, and it was based on evidence. Starting in the
mid-1980s, the APC began its long-term rise (and thus the
long-term fall in the saving rate), which continued for twenty
years. In the 2013 edition of his popular economics text, Greg-
ory Mankiw argues that the APC in Friedman’s consumption
model will rise only “when current income temporarily falls
below permanent income,” as quoted in Chapter V. He also
argues that the APC in Modigliani and Brumberg’s consump-
tion model will be constant because the wealth-to-income
ratio will be constant over “long periods of time,” also quoted
in Chapter V. So although the actual APC has been rising
for twenty years, a period no one would consider temporary,
Mankiw explains why the APCs of the major authorities on
the theory of consumption will be constant or at least stable.
“Hence, in long-time series, one should observe a constant av-
erage propensity to consume, as in fact Kuznets found.”4 Did
he look at the data? Why have today’s economists failed to jet-
tison the mainstream theory of consumption in the face of so
much evidence to the contrary?
 Notes

I
Introduction
1. Van Treeck (2012), p. 24.
2. Wolff (2010).
3. Stiglitz (2012); Rajan (2010); Krugman (2012); Palley (2012).
4. Cynamon and Fazzari (2013).
5. Piketty and Saez (2003, 2004).

II
Trends in Income Distribution
1. Information from 2009, calculated from Tables A0 and A1 at Piketty
and Saez’s web page, http://elsa.berkeley.edu/~saez/TabFig2007.xls.
2. Kaldor (1955–56).
3. Orszag (2011).
4. Stiglitz (2012), p. 64.
5. Stiglitz (2012), pp. 334–35.
6. Alvaredo, Atkinson, Piketty, and Saez (2011).
7. Palley (2002).
8. Dew-Becker and Gordon (2005), p. 125.
9. Horn et al. (2009).
10. Reich (2010).
11. Piketty and Saez (2003).
12. Barro and Sala-i-Martin (1991).
13. Drennan, Lobo, and Strumsky (2004).
14. James Galbraith (2012).
134 Notes to Pages 19–34

15. Alvaredo, Atkinson, Piketty, and Saez (2011).

16. Saez (2013).
17. Piketty and Saez (2004), p. 24.

III
Possible Causes of Rising Income Inequality
1. See Hacker and Pierson (2010); Kuttner (2007, 2013); Blank (2011); John-
son and Kwak (2011); Stiglitz (2012); Krugman (2012); Horn et al. (2009).
2. Piketty and Saez (2004).
3. Violante (2008), p. 520.
4. Blank (2011).
5. Mankiw (2013a).
6. Kaplan and Rauh (2013).
7. Dew-Becker and Gordon (2008), pp. 44–45.
8. Piketty and Saez (2003).
9. See Stiglitz (2012), p. 56; and Hacker and Pierson (2010).
10. Acemoglu and Autor (2011).
11. Mishel, Shierholz, and Schmitt (2013).
12. Piketty (2014).
13. Blank (2011).
14. Blank (2011), p. 12.
15. Blank (2011), p. 2.
16. DeNavas-Walt, Proctor, and Smith (2013).
17. Hacker and Pierson (2010).
18. Hacker and Pierson (2010), p. 187.
19. Hacker and Pierson (2010).
20. Hacker and Pierson (2010).
21. Stone (2010).
22. Hacker and Pierson (2010).
23. Hacker and Pierson (2010).
24. Hacker and Pierson (2010).
25. Drutman (2010).
26. Hacker and Pierson (2010).
27. Hacker and Pierson (2010).
28. Black (1997).
29. Stiglitz (2012), p. 35.
30. Mankiw (2013a), p. 23.
31. Stiglitz (2012), p. 37.
32. Gabaix and Landier (2006).
Notes to Pages 34–62 135

33. Dew-Becker and Gordon (2008), p. 45.

34. Bivins and Mishel (2013), Table 2.

IV
Consumers’ Shift to Debt
1. See Boushey and Weller (2006); Dutt (2006); Hockett and Dillon (forth-
coming); Mian and Sufi (2009a); van Treeck (2012); Kumhof, Ranciere, and
Winant (2013); Reich (2010); Pollin (1988); Barba and Pivetti (2009); Horn et al.
(2009); James Galbraith (2012); United Nations (2009); and IMF-ILO (2010).
2. Palumbo and Parker (2009).
3. Federal Reserve Board (2012).
4. Kumhof, Ranciere, and Winant (2013).
5. Mayer and Pence (2008).
6. Mian and Sufi (2009a).
7. Mayer and Pence (2008).
8. Mayer and Pence (2008), p. 2.
9. Federal Reserve Bank of New York (2011).
10. U.S. Courts (n.d.), “Bankruptcy Statistics.”
11. Sullivan, Warren, and Westbrook (2006), p. 220.
12. Hockett and Dillon (forthcoming), p. 39.
13. Boushey and Weller (2006).
14. See Bureau of Labor Statistics (n.d.), “Consumer Expenditure Sur-
vey”; Garner, McClelland, and Passero (2009); and Passero (2009).
15. The Rockefeller Foundation (2010).
16. Communication by email with Jacob Hacker, creator of the Economic
Security Index.
17. Stiglitz (2010b), p. 24.
18. See Carroll, Otsuka, and Slacalek (2011); Dynan (2009); and Mian and
Sufi (2010).
19. See “The Long Climb” (2009); “From Ozzie to Ricky” (2009); and
Rajan (2010).
20. Stiglitz (2010a), p. 4.
21. See Federal Reserve Board (2014b) and Shiller (2011).
22. Shiller (2011).
23. Mian and Sufi (2009b), pp. 1–2.
24. See Mian and Sufi (2009b), p. 3; Federal Reserve Board (2012); and M.
Brown et al. (2013).
25. Mian and Sufi (2009b), p. 4.
26. Mian and Sufi (2009b), Abstract.
136 Notes to Pages 62–93

27. Wolff (2010), p. 21.

28. Wolff (2010), p. 22.

V
Consumption Theory and Its Critics
1. Varian (1990), p. 18.
2. Varian (1990), p. 295.
3. Kuznets (1955).
4. Okun (1975).
5. Krugman (2014).
6. Newman, Gayar, and Spencer (1954), pp. 158, 160, 163.
7. Schumpeter (1954).
8. Cottrell (1998).
9. Schumpeter (1954), p. 622.
10. Keynes (1951), pp. 113, 115–16.
11. Keynes (1951), p. 119.
12. Keynes (1951), p. 121.
13. Keynes (1951), p. 123.
14. Cottrell (1998), p. 63.
15. Keynes quoted in C. Brown (2004, p. 293).
16. See C. Brown (2004), p. 293, note 4.
17. Kuznets (1946).
18. Modigliani and Brumberg (1954).
19. Mankiw (2013b), p. 479.
20. Mankiw (2013b), p. 479.
21. Friedman (1957).
22. Mankiw (2013b), p. 483.
23. Friedman (1957), p. 18.
24. Friedman (1957), p. 19.
25. Hoover (2003), p. 425.
26. Krusell and Smith (1998), p. 890.
27. Friedman (1957), pp. 224–25.
28. Friedman (1957), p. 224.
29. Krueger and Perri (2006), p. 186.
30. Krueger and Perri (2006), p. 187.
31. Krueger and Perri (2006), p. 164.
32. Van Treeck (2012), p. 9.
33. Kopczuk, Saez, and Song (2010), p. 125.
34. Iacoviello (2008), pp. 929, 957.
Notes to Pages 94–109 137

35. See Frank and Levine (2005); Frank (2007); Barba and Pivetti (2009);
Palley (2010); and Van Treeck (2012).
36. See Dynan, Skinner, and Zeldes (2004); Frank (2007); and Stiglitz (2012).
37. Kopczuk, Saez, and Song (2010), p. 125.
38. See Cynamon and Fazzari (2008), p. 25; and Pollin (1988).
39. Zellner (1960).
40. Rajan (2010), p. 42.
41. Van Treeck (2012), pp. 1–2. Van Treeck references Duesenberry (1949),
Frank (1985), and Frank and Levine (2005).
42. Van Treeck (2012), p. 4.
43. Bordo and Meissner (2012), p. 1; Rajan (2010); Kumhof, Ranciere, and
Winant (2013).
44. Malinen (2013), p. 2.
45. Malinen (2013), p. 13.
46. C. Brown (2004), p. 299.
47. C. Brown (2004), p. 303.
48. Bunting (2009), p. 293.
49. Bunting (2001), pp. 157–58.
50. See Veblen (1934, 1965).
51. Chase (1934).
52. Schumpeter (1954).
53. Bowles and Park (2004), p. 21.
54. Veblen (1898), p. 1.
55. Frank (2011), p. 17.
56. Bergstrom (2012).
57. Veblen (1898).
58. Duesenberry (1949), p. 26.
59. See Duesenberry (1949), quotes on pp. 28, 34, 101, 40, 45, 111–12, re-
spectively.
60. Frank (1985), p. 37.
61. Luttmer (2005), p. 990.
62. Frank and Levine (2005), pp. 2, 3.
63. Frank (2007), pp. 75–76.
64. Akerlof (2002), p. 411.
65. Akerlof (2002), p. 412, 428.
66. Friedman (1957), p. 40.
67. Klyuev and Mills (2006).
68. See Cook (2005); also Tsionas and Christopoulos (2002).
69. See Krueger and Perri (2006); C. Brown (2004); and Iacoviello (2008).
70. Rajan (2010).
138 Notes to Pages 112–32

VI
Has This Happened Before?
1. Eccles (1951), p. 77.
2. John Kenneth Galbraith (1954), p. 182.
3. Kuznets (1946); Friedman (1957).
4. Alvaredo, Atkinson, Piketty, and Saez (2011).
5. Williamson and Lindert (1980).
6. Smolensky and Plotnick (1993), pp. 7–8.
7. Goldin and Katz (n.d.), pp. 25–26.
8. Smolensky and Plotnick (1993), p. 8.
9. Goldin and Katz (n.d.), pp. 4–5.
10. Margo (1999), p. 2.
11. National Industrial Conference Board (1949).
12. Barro and Sala-i-Martin (1991).
13. National Industrial Conference Board (1949).
14. Friedman (1957), pp. 3–4.
15. Friedman (1957), p. 126.
16. Friedman (1957), p. 4.
17. Williamson and Lindert (1980), p. 77.
18. Kumhof, Ranciere, and Winant (2013), pp. 9–10.
19. Berg and Ostry (2011), p. 13.
20. Hockett (2010), p. 1268.
21. Hockett (2010).
22. Hockett (2010).
23. Olney (1999).
24. Krugman (2012), p. 83.
25. Krugman (2012), pp. 83–84.
26. Olney (1999).

VII
Conclusion
1. Bureau of Economic Analysis (n.d.), Table 2.3.1.
2. Alpert, Hockett, and Roubini (2011); Krugman (2012).
3. Krugman (2009).
4. Mankiw (2013b), p. 483.
 Bibliography

Acemoglu, Daron, and David Autor. 2011. “Skill Tasks and Technologies: Im-
plications for Employment Earnings.” In O. Ashenfelter and D. Card,
eds., The Handbook of Labor Economics, vol. 4b. Amsterdam: Elsevier.
Akerlof, George. 2002. “Behavioral Macroeconomics and Macroeconomic
Behavior.” American Economic Review 923:411–33.
Alpert, Daniel, Robert Hockett, and Nouriel Roubini. 2011. The Way For-
ward: Moving from the Post-Bubble Post-Bust Economy to Renewed
Growth and Competitiveness. New York: The New America Foundation.
Alvaredo, Facundo, Tony Atkinson, Thomas Piketty, and Emmanuel Saez.
2011. The World Top Incomes Database. http://topincomes.parisschool
ofeconomics.eu/#Home.
Barba, Aldo, and Massimo Pivetti. 2009. “Rising Household Debt: Its Causes
and Microeconomic Implications—a Long Period Analysis.” Cambridge
Journal of Economics 53:113–37.
Barro, Robert J., and Xavier Sala-i-Martin. 1991. “Convergence Across States
and Regions.” Brookings Papers on Economic Activity 1.
Berg, Andrew, and Jonathan Ostry. 2011. “Inequality and Unsustainable
Growth: Two Sides of the Same Coin?” International Monetary Fund
Staff Discussion Note.
Bergstrom, Ted. 2012. Review of Robert Frank’s “The Darwin Economy: Lib-
erty, Competition, and the Common Good.” Journal of Economic Liter-
ature 503:795.
Bivins, Josh, and Lawrence Mishel. 2013. “The Pay of Corporate Executives
and Financial Professionals as Evidence of Rents in Top 1 Percent In-
comes.” Journal of Economic Perspectives 273:57–78.
Black, John, ed. 1997. Oxford Dictionary of Economics. Oxford: Oxford Uni-
versity Press.
140 Bibliography

Blank, Rebecca M. 2011. Changing Inequality. Berkeley: University of Cali-

fornia Press.
Bordo, Michael D., and Christopher M. Meissner. 2012. “Does Inequality
Lead to Financial Crisis?” NBER Working Paper No. 17896.
Boushey, Heather, and Christian E. Weller. 2006. “Inequality and House-
hold Economic Hardship in the United States of America.” DESA Working
Paper No. 18, U.N. Department of Economic & Social Affairs, pp. 1–26.
Bowles, Samuel, and Yongjin Park. 2004. “Emulation, Inequality, and Work
Hours: Was Thorstein Veblen Right?” Economic Journal 115507:F397–F412.
Brown, Christopher. 2004. “Does Income Distribution Matter for Effective
Demand? Evidence from the United States.” Review of Political Economy
163:291–307.
Brown, Meta, Andrew Haughwout, Donghoon Lee, and Wilbert van der
Klaauw. 2013. “The Financial Crisis at the Kitchen Table: Trends in
Household Debt and Credit.” Current Issues in Economics and Finance
19(2):1–10.
Bunting, David. 2001. “Keynes’ Law and Its Critics.” Journal of Post Keynes-
ian Economics 241:149–63.
———. 2009. “The Saving Decline: Macro-Facts, Micro-Behavior.” Journal of
Economic Behavior & Organization 701–2:282–95.
Bureau of Economic Analysis, U.S. Department of Commerce. n.d. “Gross
Domestic Product.” Available at http://www.bea.gov/national/index
.htm#gdp.
———. n.d. “National Economic Accounts.” http://www.bea.gov/national/
index.htm.
———. n.d. “Table 2.3.1. Percent Change from Preceding Period in Real Per-
sonal Consumption Expenditures by Major Type of Product.” Available
at http://www.bea.gov/national/index.htm#personal.
———. n.d. “Table S.3.a. Households and Nonprofit Institutions Serving
Households.” Available at http://www.bea.gov/national/nipaweb/NI_
FedBeaSna/Index.asp.
Bureau of Labor Statistics, U.S. Department of Labor. n.d. “Consumer Ex-
penditure Survey, 1984–2010.” http://www.bls.gov/cex/#tables.
———. n.d. “Consumer Price Index.” http: //www.bls.gov/cpi.
———. 2015. “News Release, Productivity and Costs” [press release], March 5.
Carroll, Christopher D., Misuzu Otsuka, and Jiri Slacalek. 2011. “How Large
Are Housing and Financial Wealth Effects? A New Approach.” Journal of
Money, Credit, and Banking 431:55–79.
Case-Shiller House Price Index. 2012. http://www.econ.yale.edu/~shiller/
data.htm.
Bibliography 141

Chase, Stuart. 1934. Foreword to Theory of the Leisure Class. New York:
Modern Library.
Cook, Steven. 2005. “The Stationarity of Consumption-Income Ratios: Evi-
dence from Minimum LM Unit Root Testing.” Economics Letters 89:55–60.
Cottrell, A. 1998. “Keynes, Ricardo, Malthus and Say’s Law.” In James Ahrak-
por, ed., Keynes and the Classics Reconsidered, pp. 63–75. Boston: Kluwer
Academic.
Council of Economic Advisers. 2011. Economic Report of the President.
Washington, DC: U.S. Government Printing Office.
———. 2013. Economic Report of the President. Washington, DC: U.S. Gov-
ernment Printing Office.
Cynamon, Barry, and Steven M. Fazzari. 2008. “Household Debt in the Con-
sumer Age: Source of Growth—Risk of Collapse.” Capitalism and Society
3(2):1–30.
———. 2013. “Inequality, the Great Recession, and a Stagnant Recovery.”
http://www.eaepeparis2013.com/papers/Full_Paper_Barry-Cynamon
-Fazzari.pdf.
DeNavas-Walt, Carmen, and Bernadette D. Proctor, U.S. Census Bureau.
2014. Income, Poverty, and Health Insurance Coverage in the United
States: 2013. Current Population Reports, P60-249. Washington, DC:
U.S. Government Printing Office. http://www.census.gov/content/dam/
Census/library/publications/2014/demo/p60-249.pdf.
DeNavas-Walt, Carmen, Bernadette D. Proctor, and Jessica C. Smith,
U.S. Census Bureau. 2011. Income, Poverty, and Health Insurance Cov-
erage in the United States: 2010. Current Population Reports, P60-239.
Washington, DC: U.S. Government Printing Office. http://www.census
.gov/prod/2011pubs/p60-239.pdf.
———. 2013. Income, Poverty, and Health Insurance Coverage in the United
States: 2012. Current Population Reports, P60-245. Washington, DC:
U.S. Government Printing Office. http://www.census.gov/prod/2013
pubs/p60-245.pdf.
Dew-Becker, Ian, and Robert J. Gordon. 2005. “Where Did the Productivity
Growth Go? Inflation Dynamics and the Distribution of Income.” Brook-
ings Papers on Economic Activity 2:67–122.
———. 2008. “Controversies About the Rise of American Inequality: A Sur-
vey.” NBER Working Paper Submission, April 21.
Drennan, Matthew P., Jose Lobo, and Deborah Strumsky. 2004. “Unit Root
Tests of Sigma Income Convergence Across US Metropolitan Areas.”
Journal of Economic Geography 4(5):583–95.
Drutman, Lee Jared. 2010. “The Business of America Is Lobbying: The Ex-
142 Bibliography

pansion of Corporate Political Activity and the Future of American Plu-

ralism.” Ph.D. dissertation, University of California, Berkeley.
Duesenberry, James S. 1949. Income, Saving and the Theory of Consumer Be-
havior. Cambridge, MA: Harvard University Press.
Dutt, Amitav K. 2006. “Maturity, Stagnation and Consumer Debt: A Steind­
lian Approach.” Metroeconomica 57(3):339–64.
Dynan, Karen E. 2009. “Changing Household Financing Opportunities and
Economic Security.” Journal of Economic Perspectives 23(4):49–68.
Dynan, Karen E., Jonathan Skinner, and Stephen Zeldes. 2004. “Do the Rich
Save More?” Journal of Political Economy 112(2):397–442.
Eccles, Marriner. 1951. Beckoning Frontiers. New York: Alfred A. Knopf.
Federal Housing Finance Agency. 2014. House Price Index. http://www.fhfa
.gov/DataTools/Downloads/Pages/House-Price-Index.aspx.
Federal Reserve Bank of New York. 2011. “Household Debt and Credit.” August.
Federal Reserve Board. 2009. “Survey of Consumer Finances, 2007.” http://
www.federalreserve.gov/pubs/oss/oss2/2007/scf2007home.html.
———. 2012. “Survey of Consumer Finances, 2010.” http://www.federalreserve
.gov/econresdata/scf/scf_2010.htm.
———. 2014a. Financial Accounts of the United States: Flow of Funds, Bal-
ance Sheets, and Integrated Macroeconomic Accounts—Historical Annual Ta-
bles, 2005–2013. http://www.federalreserve.gov/releases/z1/Current/annuals/
a2005-2013.pdf.
———. 2014b. “Household Debt Service and Financial Obligations Ratios.”
http://www.federalreserve.gov/datadownload/Download.aspx?rel=FOR
&series=767f957af5d4d8e28f95408bfb53e557&filetype=spreadsheetml&la
bel=include&layout=seriescolumn&from=03/01/1980&to=03/31/2014.
Frank, Robert H. 1985. Choosing the Right Pond. New York: Oxford Univer-
sity Press.
———. 2007. Falling Behind. Berkeley: University of California Press.
———. 2011. The Darwin Economy: Liberty, Competition, and the Common
Good. Princeton, NJ: Princeton University Press.
Frank, Robert H., and Seth Levine. 2005. “Expenditure Cascades.” https://
www.aeaweb.org/annual_mtg_papers/2007/0107_1300_0202.pdf.
Friedman, Milton. 1957. A Theory of the Consumption Function. National Bu-
reau of Economic Research. Princeton, NJ: Princeton University Press.
“From Ozzie to Ricky.” 2009. Economist, October 3–9, pp. 5–8.
Gabaix, Xavier, and Augustin Landier. 2006. “Why Has CEO PAY Increased
So Much?” Quarterly Journal of Economics 123(1):49–100.
Galbraith, James. 2012. Inequality and Instability. Oxford: Oxford University
Press.
Bibliography 143

Galbraith, John Kenneth. 1954. The Great Crash. New York: Houghton Mifflin.
Garner, Thesia I., Robert McClelland, and William Passero. 2009. “Strengths
and Weaknesses of the Consumer Expenditure Survey from a BLS Per-
spective.” NBER Summer Institute, Conference on Research on Income
and Wealth, Cambridge, Massachusetts, July 13.
Goldin, Claudia, and Lawrence F. Katz. n.d., n.p. “Decreasing and Then In-
creasing Inequality in America: A Tale of Two Half-Centuries.”
Goldsmith, Raymond W. 1955. A Study of Saving in the United States. 3 vols.
Princeton, NJ: Princeton University Press.
Hacker, Jacob S. 2015. Economic Security Index. http://www.economicsecu
rityindex.org/?p=home.
Hacker, Jacob S., and Paul Pierson. 2010. “Winner-Take-All Politics: Public
Policy, Political Organization, and the Precipitous Rise of Top Incomes
in the United States.” Politics and Society 38(2):152–204.
Hockett, Robert. 2010. “A Fixer-Upper for Finance.” Washington University
Law Review 87(6):1213–91.
Hockett, Robert, and Daniel Dillon. Forthcoming. “Income Inequality and
Market Fragility: Some Empirics in the Political Economy of Finance.”
Northeastern Economic Journal.
Hoover, Kevin D. 2003. “A History of Postwar Monetary Economics and
Macroeconomics.” In Warren J. Samuels, Jeff E. Biddle, and John B.
Davis, eds., A Companion to the History of Economic Thought, pp. 411–27.
Oxford: Blackwell.
Horn, Gustav, Katharina Droge, Simon Sturn, Till van Treeck, and Rudolph
Zweiner. 2009. “From the Financial Crisis to the World Economic Crisis:
The Role of Inequality.” IMK Report No. 41, Policy Brief.
Iacoviello, Matteo. 2008. “Household Debt and Income Inequality.” Journal
of Money, Credit and Banking 40(5):929–65.
IMF-ILO. 2010. “The Challenges of Growth, Employment and Social Cohe-
sion.” Proceedings from joint IMF-ILO conference, Oslo, Norway, Sep-
tember 13.
Johnson, Simon, and James Kwak. 2011. Thirteen Bankers. New York: Ran-
dom House.
Kaldor, Nicolas. 1955–56. “Alternative Theories of Distribution.” Review of
Economic Studies 23(2):83–100.
Kaplan, Steven N., and Joshua Rauh. 2013. “It’s the Market: The Broad-
Based Rise in the Return to Top Talent.” Journal of Economic Perspectives
27(3):35–56.
Keynes, John Maynard. 1936. The General Theory of Employment, Interest
and Money. London: Macmillan.
144 Bibliography

———. 1951. Essays in Biography. New York: Horizon Press.

Klyuev, Vladamir, and Paul Mills. 2006. “Is Housing Wealth an ATM?
The Relationship Between Household Wealth, Home Equity With-
drawal and Saving Rates.” International Monetary Fund Working Paper
06/162.
Kopczuk, Wojciech, Emmanuel Saez, and Jae Song. 2010. “Earnings Ine-
quality and Mobility in the United States: Evidence from Social Security
Data Since 1937.” Quarterly Journal of Economics 125(1):91–128.
Krueger, Dirk, and Fabrizio Perri. 2006. “Does Income Inequality Lead to
Consumption Inequality? Evidence and Theory.” Review of Economic
Studies 73:163–93.
Krugman, Paul. 2009. “Role Reversal.” New York Times, November 22.
———. 2012. End This Depression Now! New York: W. W. Norton.
———. 2014. “Why We Are in a New Gilded Age.” New York Review of Books,
May 8.
Krusell, Per, and Anthony A. Smith, Jr. 1998. “Income and Wealth Heteroge-
neity in the Macroeconomy.” Journal of Political Economy 106(5):867–96.
Kumhof, Michael, Romain Ranciere, and Pablo Winant. 2013. “Inequality,
Leverage and Crises: The Case of Endogenous Default.” International
Monetary Fund Working Paper, WP/13/249.
Kuttner, Robert. 2007. The Squandering of America: How the Failure of Our
Politics Undermines Our Prosperity. New York: Alfred A. Knopf.
———. 2013. Debtors’ Prison. New York: Alfred A. Knopf.
Kuznets, Simon S. 1946. National Income: A Summary of Findings. New York:
National Bureau of Economic Research.
———. 1955. “Economic Growth and Income Inequality.” American Eco-
nomic Review 45(1):1–28.
“The Long Climb.” 2009. Economist, October 3–9, p. 8.
Luttmer, Erzo F. P. 2005. “Neighbors as Negatives: Relative Earnings and
Well-Being.” Quarterly Journal of Economics, August, pp. 963–1002.
Malinen, Tuomas. 2013. “Is There a Relationship Between Income Inequality
and Leverage?” Helsinki Center of Economic Research, Discussion Paper
No. 362.
Mankiw, N. Gregory. 2013a. “Defending the One Percent.” Journal of Eco-
nomic Perspectives 27(3):21–34.
———. 2013b. Macroeconomics, 8th edition. Houndmills, UK: Palgrave Mac-
millan.
Margo, Robert A. 1999. “The History of Wage Inequality in America, 1820 to
1970.” Social Science Research Network, Working Paper No. 286.
Bibliography 145

Mayer, Christopher J., and Karen Pence. 2008. “Subprime Mortgages: What,
Where, and to Whom?” NBER Working Paper 14083.
Mian, Atif, and Amir Sufi. 2009a. “The Consequences of Mortgage Credit
Expansion: Evidence from the U.S. Mortgage Default Crisis.” Quarterly
Journal of Economics, November, pp. 1449–96.
———. 2009b. “House Prices, Home Equity-Based Borrowing, and the
United States Household Leverage Crisis.” National Bureau of Economic
Research, Working Paper no. 15283.
———. 2010. “Household Leverage and the Recession of 2007 to 2009.”
NBER Working Paper, No. 15896.
Mishel, Lawrence, Heidi Shierholz, and John Schmitt. 2013. “Don’t Blame
the Robots: Assessing the Job Polarization Explanation of Growing Wage
Inequality.” Working Paper, Economic Policy Institute.
Modigliani, Franco, and Richard Brumberg. 1954. “Utility Analysis and the
Consumption Function: An Interpretation of Cross-Section Data.” In
K. K. Kurihara, ed., Post-Keynesian Economics, pp. 3–45. New Brunswick,
NJ: Rutgers University Press.
National Industrial Conference Board (NICB). 1949. “National Income–
Percentage of Income Received by Specified Proportion of Recipients
Arranged According to Size of Income: 1910 to 1937.” In U.S. Census, His-
torical Statistics of the United States, 1789–1945, Series A 176–94. Wash-
ington, DC: U.S. Government Printing Office.
Newman, Philip C., Arthur D. Gayar, and Milton H. Spencer, eds. 1954.
Source Readings in Economic Thought. New York: W. W. Norton.
Okun, Arthur M. 1975. Equality and Efficiency: The Great Tradeoff. Washing-
ton, DC: Brookings Institution.
Olney, Martha L. 1999. “Avoiding Default: The Role of Credit in the Con-
sumption Collapse of 1930.” Quarterly Journal of Economics 114(1):319–35.
Orszag, Peter. 2011. “As Kaldor’s Facts Fall, Occupy Wall Street Rises.”
Bloomberg View, October 18.
Palley, Thomas I. 2002. “Economic Contradictions Coming Home to Roost?
Does the U.S. Economy Face a Long-Term Aggregate Demand Genera-
tion Problem?” Journal of Post Keynesian Economics 25(1):9–32.
———. 2010. “The Relative Permanent Income Theory of Consumption: A
Synthetic Keynes-Duesenberry-Friedman Model.” Review of Political
Economy 22(1):41–56.
———. 2012. From Financial Crisis to Stagnation. Cambridge: Cambridge
University Press.
Palumbo, Michael J., and Jonathan A. Parker. 2009. “The Integrated Finan-
146 Bibliography

cial and Real System of National Accounts for the United States: Does It
Presage the Financial Crisis?” NBER Working Paper 14663.
Passero, William. 2009. “The Impact of Income Imputation in the Con-
sumer Expenditure Survey.” Monthly Labor Review, August, pp. 25–42.
Piketty, Thomas. 2014. Capital in the Twenty-first Century. Cambridge, MA:
Belknap Press of Harvard University Press.
Piketty, Thomas, and Emmanuel Saez. 2003. “Income Inequality in the
United States, 1913–1998.” Quarterly Journal of Economics 118(1):1–39.
———. 2004. “Income Inequality in the United States, 1913–2002.” University
of California, Berkeley, Department of Economics. http://elsa.berkeley
.edu/~saez/piketty-saezOUP04US.pdf.
Pollin, Robert. 1988. “The Growth of U.S. Household Debt: Demand-Side
Influences.” Journal of Macroeconomics 10(2):231–48.
Rajan, Raghuram G. 2010. Fault Lines: How Hidden Fractures Still Threaten
the World Economy. Princeton, NJ: Princeton University Press.
Reich, Robert B. 2010. Aftershock. New York: Vintage Books.
The Rockefeller Foundation. 2010. “More Americans Are Financially In-
secure Now Than in the Past 25 Years.” http://www.rockefellerfounda
tion.org/news/publications/more-americans-are-financially-insecure.
Saez, Emanuel. 2013. “Striking It Richer: The Evolution of Top Income in
the United States.” Pathways, Winter 2008, 6–7; http://elsa.berkeley
.edu/~saez/TabFig2012prel.xls.
Schumpeter, Joseph A. 1954. History of Economic Analysis. New York: Ox-
ford University Press.
Shiller, Robert J. 2011. “The Sickness Beneath the Slump.” New York Times,
June 12.
Smolensky, Eugene, and Robert Plotnick. 1993. “Inequality and Poverty in
the United States: 1900 to 1990.” Institute for Research on Poverty, Dis-
cussion Paper #998–93.
Stiglitz, Joseph E. 2010a. Free Fall. New York: W. W. Norton.
———. 2010b. The Stiglitz Report: Reforming the International Monetary and
Financial Systems in the Wake of the Global Crisis. New York: New Press.
———. 2012. The Price of Inequality. New York: W. W. Norton.
Stone, Katherine V. W. 2010. “John R. Commons and the Origins of Legal
Realism, or the Other Tragedy of the Commons.” In A. Brophy and
D. Hamilton, eds., Transformations in American Legal History, vol. 2, pp.
326–43. Cambridge, MA: Harvard University Press.
Sullivan, Teresa A., Elizabeth Warren, and Jay Lawrence Westbrook. 2006.
“Less Stigma or More Financial Distress: An Empirical Analysis of the
Bibliography 147

Extraordinary Increase in Bankruptcy Filings.” Stanford Law Review

59(2):213–56.
Temin, Peter. 2010. “The Great Recession and the Great Depression.” Daeda-
lus 139(4):115–24.
Tsionas, Eftymios, and Dmitris Christopoulos. 2002. “Non-Stationarity
in the Consumption-Income Ratio: Further Evidence from Panel and
Asymmetric Unit Root Tests.” Economics Bulletin 3(122):1–5.
United Nations. 2009. Report of the Commission of Experts of the President
of the United Nations General Assembly on Reforms of the International
Monetary and Financial System. New York: United Nations. http://www
.un.org/ga/econcrisissummit/docs/FinalReport_CoE.pdf.
U.S. Census Bureau. 1975. Historical Statistics of the United States, Colonial
Times to 1970. Washington, DC: U.S. Government Printing Office.
———. 2009. “Historical Income Tables—Households.” http://www.census
.gov/hhes/www/income/data/historical/household/index.html.
U.S. Courts. n.d. “Bankruptcy Statistics.” http://www.uscourts.gov/Statistics
/BankruptcyStatistics.aspx.
Van Treeck, Till. 2012. “Did Inequality Cause the U.S. Financial Crisis?”
IMK Working Paper 91, pp. 1–39.
Varian, H. R. 1990. Intermediate Microeconomics: A Modern Approach, 2nd
edition. New York: W. W. Norton.
Veblen, Thorstein. 1898. “Why Is Economics Not an Evolutionary Science.”
Quarterly Journal of Economics 12:2–17.
———. 1934. The Theory of the Leisure Class. New York: Modern Library.
———. 1965. The Theory of Business Enterprise. New York: A. M. Kelley.
Violante, Giovanni L. 2008. “Skill-Biased Technological Change.” In The
New Palgrave Dictionary of Economics, 2nd edition, ed. Steven N. Dur-
lauf and Lawrence E. Blume. New York: Palgrave Macmillan.
Williamson, Jeffrey G., and Peter H. Lindert. 1980. American Inequality: A
Macroeconomic History. New York: Academic Press.
Wolff, Edward N. 2010. “Recent Trends in Household Wealth in the United
States: Rising Debt and the Middle-Class Squeeze—An Update to 2007.”
Levy Economics Institute, Working Paper No. 589.
Zellner, Arnold. 1960. “Tests of Some Basic Propositions in the Theory of
Consumption.” American Economic Review 50(2):565–73.
This page intentionally left blank
 Index

agriculture, 76, 117–18 Bordo, Michael, 96

Akerlof, George, 106–7 Boston, 43
Alpert, Daniel, 129 Bowles, Samuel, 100–101
Alt-A mortgages, 42, 44, 54 Brookings Institution, 32
American Economic Review, 78 Brown, Christopher, 97
American Economic Review Papers Brumberg, Richard, 4, 5, 6, 73, 75,
and Proceedings, 78 99, 102, 103, 104, 110, 131;
American Enterprise Institute consumption viewed by, 74,
(AEI), 32 83–87, 109, 132
antitrust laws, 6 bubbles: in housing, 1–2, 3, 37,
appliances, 59 43–44, 53–54, 55, 56, 86, 87,
arbitration, 27, 29 110, 111–12, 126, 128; in tech-
Arizona, 42, 44 nology stocks, 55, 85, 110
arms races, 102 Bunting, David, 98–99, 108
Asia, 21 Bureau of Labor Statistics (BLS),
austerity, 130 49, 51
average propensity to consume Business Roundtable, 30
(APC). See consumption
California, 42–43, 44
Bakersfield, Calif., 42–43 call centers, 21
bankruptcy, 29, 45–46 campaign spending, 6, 28, 31
Barro, Robert, 17–18, 118 Canada, 21–22, 27
Beckoning Frontiers (Eccles), 112 capital flows, 21
behavioral economics, 6, 75, 99, capital income, 9, 11, 12, 19, 24–25
104, 131 Capital in the Twenty-first Century
Bivins, Josh, 34 (Piketty), 24–25, 78
blacks, 43 card check certification, 27
Blank, Rebecca, 26 car loans, 70
150 Index

Carroll, Christopher, 54 Cook, Steven, 108

Carter, Jimmy, 28 copyright, 29, 30
Cato Institute, 32 corporate governance, 34
Census Bureau, 51 Cottrell, Allin, 79, 81
Century Fund, 32 credit: availability of, 1, 44, 55, 62,
Chamber of Commerce, 30 111–12, 127–28; income ine-
Changing Inequality (Blank), 26 quality mitigated by, 70, 87,
child care, 60 93, 94–95, 109; interest rates,
Christopoulos, Dmitris, 108–9 1, 5, 44, 54, 55, 108, 111–12,
civil rights, 31 130
Cold War, 74, 77 credit cards, 38, 62, 70
Commons, John R., 29–30, 33 credit scores, 41–42
competition, 102 Current Population Survey (CPS),
conspicuous consumption, 3, 100 50, 70
consumer debt. See household
debt Darwin, Charles, 101–2
Consumer Expenditure Survey Debt Service Ratio (DSR), 38
(CEX), 49–50, 51, 70–71, 72, deficit spending, 130
91, 131 deflation, 130
consumer interests, 30–31 deleveraging, 2, 112
Consumer Price Index (CPI), democracy, 30, 74, 76
59–60 deregulation, 44, 95, 108
consumption: conspicuous, 3, 100; derivatives, 33
debt-supported, 1, 2; Fried- Detroit, 43
man’s theory of APC, 4, 5, Dew-Becker, Ian, 23, 34
6, 75, 83, 87–92, 94, 97–99, Dickey-Fuller test, 109
102–10, 119, 120, 132; income Dillon, Daniel, 48–49
inequality linked to rise in, dividends, 15
5, 75, 86–87, 97, 107–8, 111, Duesenberry, James, 75, 99, 102–7,
126, 131; inequality of, 40, 108
91–94; Keynes’s theory of
APC, 4–5, 74, 81–83, 90, 132; eastern Europe, 21
long-term rise of APC, 5, 85, Eccles, Marriner, 112
87, 118–21, 128, 132; main- Economic Security Index (ESI),
stream theories of, 3–7, 51–53
73–110, 130; Modigliani and education: cost of, 3, 6, 58–61, 131;
Brumberg’s theory of APC, of workforce, 23, 24
74, 83–87, 109, 132; revised effective demand, 4, 79–81, 130
theory of, 109–10, 131 End This Depression Now! (Krug-
container shipping, 21 man), 129–30
Index 151

environmental interests, 31 Germany, 21–22

Equality and Efficiency (Okun), Gilded Age, 29, 77
77–78 Gini coefficient, 8, 15, 16, 17, 77, 91
estate taxes, 31 globalization, 21–22
executive compensation, 23, 34 Goldin, Claudia, 115–16
Goldsmith, Raymond, 116, 119–21
Fannie Mae (Federal National Mort- Gordon, Robert, 23, 34
gage Association), 54–55 government spending, 20
Fault Lines (Rajan), 94–96 Great Crash, The (Galbraith), 113
Federal Register, 31 Great Depression, 7, 8–9, 15, 19, 77,
Federal Reserve Board, 36, 38 100, 111, 112–18, 127
filibusters, 28, 30 Great Recession, 39, 69, 71, 73, 127,
financial deregulation, 44, 95, 108 128; causes of, 1–2, 3, 7, 46,
Financial Obligation Ratio (FOR), 63, 111; sluggish recovery
38 from, 2, 63, 129, 130–31; top
first contract arbitration, 27 wage earners during, 11–12
Florida, 42, 44 Greenspan, Alan, 55
foreclosures, 111, 124–25
France, 21–22 Hacker, Jacob, 24, 27, 28–29
Frank, Robert H., 75, 101–2, 104, health care: cost of, 3, 6, 58–61, 131;
105 employment in, 27
Freddie Mac (Federal Home Loan health insurance, 59
Mortgage Corporation), Heritage Foundation, 32
54–55 Hispanics, 43
Friedman, Milton, 73; consump- History of Economic Analysis
tion viewed by, 4, 5, 6, 75, (Schumpeter), 100
83, 87–92, 94, 97–99, 102–10, Hockett, Robert, 48–49, 124–25,
119, 120, 132; inequality dis- 129
regarded by, 4, 6, 107, 115; home equity lines of credit, 54,
saving viewed by, 75, 89, 90, 56, 70
97–98 home equity loans, 37, 54, 56, 70
full employment, 77, 100 home insurance, 38, 59
furniture, 59 Hoover, Herbert, 126
Hoover, Kevin, 89
Galbraith, James, 18 Hoover Institution, 32
Galbraith, John Kenneth, 113 hourly earnings, 8
Garn–St Germain Act (1982), 54 household debt: bankruptcies and,
General Theory of Employment, 43–46; causes of, 1, 2–3, 5,
Interest, and Money 53–63; economic insecurity
(Keynes), 4–5, 81, 106 and, 6, 50–53; Great Depres-
152 Index

household debt (continued) linked to, 112–18; house-

sion linked to, 112–18; hous- hold debt linked to, 46–50,
ing prices and, 43–46; 63–64, 66–72, 93, 121–24,
by income group, 39–40; 127, 128, 131; how to reverse,
income inequality linked 130–31; income stagnation
to, 46–50, 63–64, 66–72, 93, linked to, 8–9; institutional
121–24, 127, 128, 131; national causes of, 26–28; job polar-
data on, 35–39, 49–50, ization linked to, 24–25;
69–72; non-mortgage, 126; measures of, 8, 15, 16, 17,
rise of, 121–28; state data 50, 70, 72, 77, 91, 97; politi-
on, 46–49, 63–69; subprime cal causes of, 28–34; rising
mortgages and, 40–43 consumption linked to, 5,
household income, 8, 9, 15, 16, 19, 75, 86–87, 97, 107–8, 111, 126,
37, 66; ratios of, 17, 35, 38–41, 131; SBTC linked to, 22–24;
131, 132 trends in, 8–19, 120
housing: cost of, 1–2, 3, 6, 35, 37, income stagnation, 1, 6, 8–9, 12–17,
42–45, 47, 48, 54–61, 66, 68, 111
124–27, 131; insurance for, 38, incomplete markets, 110
59; as liquid asset, 1, 35–36, industrialization, 76–77
44, 54, 56–62, 108; mainte- inflation, 59–60, 117, 130
nance costs for, 59 information technology (IT),
Houston, 43 22–23
Hume, David, 111 infrastructure, 130
Institutionalist School, 99
Iacoviello, Matteo, 93, 94 interest income, 15
immigration, 21 interest groups, 30
imports, 60, 130 interest rates, 1, 5, 44, 54, 55, 108,
Income, Saving and the Theory 111–12, 130
of Consumer Behavior International Monetary Fund, 108
(Duesenberry), 102–3 inverted-U hypothesis, 76–77, 78
income convergence, 17, 18, 118
income divergence, 18 Japan, 21–22
income inequality: academic indif- job polarization, 21, 24–25
ference to, 1, 2, 3, 4, 76–78, Journal of Political Economy, 78
131; causes of, 6, 131; credit junior mortgages, 56
and, 70, 87, 93, 94–95, 109;
demographic causes of, Katz, Lawrence, 115–16
25–26; economic causes of, Keynes, John Maynard, 4–5, 6,
20–25; globalization linked 74–75, 80–83, 87, 99, 106–7,
to, 21–22; Great Depression 132
Index 153

Krueger, Dirk, 91–93, 94 median voter theorem, 28

Krugman, Paul, 2, 78, 126–27, Medicaid, 59
129–30 Medicare, 29, 59
Kumhof, Michael, 96, 122–24 Meissner, Christopher, 96
Kuznets, Simon: falling inequal- Memphis, Tenn., 42–43
ity hypothesized by, 73–74; Miami, 43
income inequality docu- Mian, Atif, 62
mented by, 5, 113, 114, 115, Midwest, 43
116–17; inverted-U hypoth- minorities, 43
esis of, 76–77, 78; Mankiw’s Mishel, Lawrence, 24, 34
misinterpretation of, 132; Modigliani, Franco, 4, 5, 6, 73, 75,
rising APC documented by, 99, 102, 103, 104, 110, 131;
5, 82, 108, 118–19, 120, 121 consumption viewed by, 74,
83–87, 109, 132
labor income, 8, 9–10, 49 Monetary Control Act (1980), 54
labor laws, 6, 27–28, 76 monopoly, 30, 33
labor unions, 27–28, 30–31 mortgages, 37–39, 70, 111, 128;
Las Vegas, 43 down payments for, 125;
Latin America, 21 housing prices and, 124–25,
Levine, Seth, 105 127; junior and second, 56;
libertarianism, 102 refinancing of, 37, 41, 54,
life-cycle theory of consumption, 56, 124, 126; short-term vs.
4, 84, 98 long-term, 124–25; sub-
Lindert, Peter, 115, 116, 117, 121 prime, 40–44, 54–55, 95.
living standards, 12 See also household debt
lobbying, 6, 30–32 mutual funds, 28
Lucas, Robert, Jr., 78, 89
National Association of Manufac-
macroeconomics, 78–79, 89, 106 turers, 30
Madison, Wis., 43 National Industrial Conference
Malinen, Tuomas, 96–97 Board (NICB), 113–15, 116
Malthus, Thomas, 4, 6, 73, 74, Nevada, 42, 44
79–81, 99 New Classical economics, 106
managerial and professional work- New Deal, 126
ers, 11, 116 New York, 43
Mankiw, Gregory, 23, 33, 84, 86, non-enactment, 29, 30
87–88, 132 nonprofit sector, 35–36
manufacturing, 27, 116
Margo, Robert, 117 off-shoring, 21
Mayer, Christopher, 42 Okun, Arthur, 74, 77–78
154 Index

Olin Foundation, 32 representative agent, 89–90, 99

Orzsag, Peter, 10 retail trade, 27
Otsuka, Misuzu, 54 Ricardo, David, 4, 6, 73, 74, 79,
over-the-counter market, 33 81, 99
Roosevelt, Franklin, 126
Palley, Thomas, 2 Roubini, Nouriel, 129
Panel Study of Income Dynam-
ics, 51 Saez, Emmanuel, 19, 23–24, 113,
Pareto efficiency, 73, 76–78 114, 115, 116, 120
Park, Yongjin, 100–101 safety net, 76
patents, 29, 30, 33–34 Sala-i-Martin, Xavier, 17–18, 118
Pence, Karen, 42 saving, 99, 131, 132; consumption
permanent income hypothesis, 4, vs., 2, 5, 35–36, 109; declin-
75, 87–94, 95, 97–98, 105–6, ing rate of, 108, 119; Duesen-
107 berry’s theory of, 102–7;
Perri, Fabrizio, 91–93, 94 forced, 106; Friedman’s the-
pharmaceuticals, 29 ory of, 75, 89, 90, 97–98;
Pierson, Paul, 24, 27, 28–29 Keynes’s theory of, 74; Mal-
Piketty, Thomas, 19, 23–25, 78, 113, thus’s theory of, 80–81
114, 115, 116, 120 Say’s law, 74, 79–80
Plotnick, Robert, 115, 116 Schumpeter, Joseph, 79–80, 100
policy drift, 30 second mortgages, 56
political contributions, 6, 28, 31 securitization, 40, 54
Price of Inequality, The (Stiglitz), 33 service sector, 21, 27
Principles of Political Economy Shiller, Robert, 129
(Malthus), 80 skill-biased technological change
prisoner’s dilemma, 102 (SBTC), 21, 22–24
productivity, 8, 12–19, 22, 27 Slacalek, Jiri, 54
Smith, Adam, 101–2, 111
Quarterly Journal of Economics, Smolensky, Eugene, 115, 116
78, 101 software production, 21
Stiglitz, Joseph, 2, 10, 24, 33–34,
Rajan, Raghuram, 2, 94–96, 109 53–54, 55
Ranciere, Roman, 96, 122–24 stock options, 29, 34
rationality assumption, 75, 102, 110 stock prices, 34
refinancing, 37, 41, 54, 56, 124, 126 strike replacements, 27–28
relative income hypothesis, 73, 75, Study of Saving in the United
94, 96, 102, 105–6 States, A (Goldsmith), 120
rental income, 15 subprime mortgages, 40–44,
rent-seeking, 23, 32–34 54–55, 95
Index 155

Sufi, Amir, 62 Twentieth Century Fund, 32

Survey of Consumer Finances two-earner families, 26
(SCF), 36
Survey of Income and Program unemployment, 47, 52, 64, 66,
Participation (SIPP), 51 68–69, 100, 130
unionization, 27–28, 30–31
tariffs, 21 United Kingdom, 21–22, 130
taxes, 20, 29, 33; estate, 31; as fiscal urbanization, 118
policy tool, 130; progressive, utilities, 59
77; property, 38, 59
technology: labor supply and, 10; vacation homes, 37, 70
skill-biased changes in, 21, van Treeck, Till, 2, 95–96
22–24 Veblen, Thorsten, 75, 99–102, 103
telecommunications, 21 vehicle loans, 70
Temin, Peter, 111
Theil index, 8, 15, 16, 17, 50, 70, wage compression, 115–16
72, 97 wage income, 11
Theory of Business Enterprise, The warehousing, 27
(Veblen), 100 Way Forward, The (Alpert, Hock-
Theory of the Consumption Func- ett, and Roubini), 129
tion, A (Friedman), 90, 102 West Virginia, 42
Theory of the Leisure Class wholesale trade, 27
(Veblen), 100 Williamson, Jeffrey, 115, 116, 117, 121
think tanks, 31–32 Winant, Pablo, 96, 122–24
tort reform, 29 “Winner-Take-All Politics” (Hacker
trade, 21 and Pierson), 28–29
transfer payments, 15, 20, 29 Wolff, Edward, 62–63
transitory income, 75, 87, 92, 93, 94 women: in labor force, 13–15,
transportation: cost of, 21, 58–60; 25–26; organizations for, 31
employment in, 27 World Trade Center attack, 55
Treatise on Political Economy (Say), World Trade Organization, 21
79 World War I, 19, 115, 117, 120
Tsionas, Eftymois, 108–9 World War II, 15, 19, 77, 130