ESSAYS ON JOB DISPLACEMENT AND EARNINGS VOLATILITY
By
Kyong Hyun Koo

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
Economics — Doctor of Philosophy
2017

ABSTRACT
ESSAYS ON JOB DISPLACEMENT AND EARNINGS VOLATILITY
By
Kyong Hyun Koo
The first essay “The Effects of Job Displacement on Family Expenditures” explores the
effects of a husband’s job displacement on his family expenditures by using the Panel Study of
Income Dynamics during 1999-2013. Even with a lasting and sizable fall in family income
induced by a husband’s job displacement, the households reduced their expenditures only
moderately. More specifically, their expenditures fell just slightly in the periods around the job
displacement occurrence, then continued to further decline two to six or more years after the job
loss, but not as much as their concurrent income loss. These later declines in their expenditures
were largely driven by the fall in those for housing, health care and education. The relatively
small reduction in those expenditures was achieved through a reduction in wealth.

The second essay “The Evolution of Earnings Volatility During and After the Great
Recession” studies how men’s and women’s earnings volatility evolved during and after the
Great Recession, the most severe downturn of the postwar era. Using matched March Current
Population Survey data, I find that earnings volatility rose considerably after the Great Recession
began and, particularly for men, the volatility increase was as large as in the severe recession of
the early 1980s. These findings update the evidence on the counter-cyclicality of earnings
volatility. I show that such counter-cyclicality is due mainly to counter-cyclical volatility in

annual hours of work. In turn, the counter-cyclical volatility in work hours appears mainly
among workers who experience unemployment.

Finally, the third essay “How Did Men’s Earnings Volatility Change during the Great
Recession? A Comparison of Evidence between the PSID and CPS” also finds men’s earnings
volatility increased substantially during the Great Recession by using the Panel Study of Income
Dynamics (PSID). I further find that the counter-cyclicality was mainly driven by unemployed
workers. All these findings are largely consistent with those of my second essay, which examines
the same topic with the matched March Current Population Study (CPS). In particular, I find
that the PSID and CPS generate a fairly similar evolution of men’s earnings volatility from the
early 1980s through the Great Recession for the sample, excluding zero-earners and the 1 % top
and bottom positive-earners. For the sample including the zero-earners and outliers, however, the
PSID shows somewhat divergent results compared to the CPS, particularly for the period 19932004. This partly explains why previous studies have reported different trends in men’s earnings
volatility during the early 2000s between the two datasets.

Copyright by
KYONG HYUN KOO
2017

To my wife, Ye Won

iv

ACKNOWLEDGEMENTS

First of all, I would like to express my deep gratitude to Steven Haider, my dissertation
committee chair. He not only provided me with excellent guidance and inspiration for
developing my dissertation, but also was willing to be my counselor who kept encouraging and
supporting me whenever I faced challenges during my Ph.D. study. It would be unimaginable to
finish my dissertation without his advice and support.

I also greatly appreciate my other committee members, Christian Ahlin and Ben Zou. They
were always attentive to my academic progress and provided me with their thoughtful advice and
feedback. I will never forget their sincere cooperation, attitude, and all their encouraging words,
which enabled me to keep proceeding with my research. I would like to send out a special thanks
to Gary Solon. I owe him many insights that started the ball rolling for my dissertation idea while
he was at Michigan State University (MSU). I also thank Todd Elder for his valuable comments
on my research and all of his efforts to guide me on the job market.

There were also many of my other colleagues at MSU sharing their insightful views on my
research; I thank Kyu Yub Lee, Kyoungbo Sim, Hyeongjun Bang, Soobin Kim, Cheol Keun Cho,
Hyunsoo Kim, Doosoo Kim, Yung Shin Jang, Hwanoong Lee, and all other participants at my
seminar at MSU for their helpful comments. I am also very grateful to Ruth Mendel and Byung
Lee for their advice/suggestions on how to edit my dissertation. With their help, my dissertation
has been written and crafted with conciseness and clarity. I would also like to thank Rich Bearup,
the director of Friendship House at East Lansing. Through weekly talks that I had with him

v

during the last year of my Ph.D. study, I could be truly encouraged and comforted so that I may
maintain my focus until the end of my Ph.D. study.

I cannot emphasize enough how much I am indebted to my family for being what I am now. I
am truly thankful to my parents, Jae Kwan Koo and Ye Sook Kim, who are dedicated to helping
their children, albeit they have already given everything to their children through their lives.
Without their love and prayers, there would be nothing that I could achieve. I thank my brother,
Min Hyun Koo, as well, for taking care of our parents in Korea while I was abroad studying at
MSU. I want him to know how much I am proud of and rely on him. Above all, I greatly
appreciate my wife, Ye Won Shin for her continual support and love. Since I met her 12 years
ago, whatever I dream for, she has trusted and encouraged me. My long journey to obtain a
doctoral degree in economics also could be embarked because of her exceptional encouragement
10 years ago. Now I am really glad that I could finally share all the joy of completing this
journey with my wife, and also with our lovely daughter, Ohnew, together as a family. I just
want my wife to know that the reason why I feel this moment to be very special is because of the
fact that she is right beside me.

Lastly and most importantly, I thank God with all my heart for his sincere guidance and
unfailing love that He has shown me throughout my entire life.

vi

TABLE OF CONTENTS

LIST OF TABLES ....................................................................................................................... ix
LIST OF FIGURES ..................................................................................................................... xi
CHAPTER 1 The Effects of Job Displacement on Family Expenditures ............................... 1
1.1 Introduction ................................................................................................................... 1
1.2 Literature Review.......................................................................................................... 5
1.3 Econometric Approach ................................................................................................. 8
1.4 Data ............................................................................................................................. 10
1.4.1 Constructing Annual Family Expenditures .................................................. 11
1.4.2 Job Displacement ......................................................................................... 13
1.4.3 Sample Selection.......................................................................................... 15
1.4.4 Descriptive Statistics for the Non-displaced and Displaced ........................ 17
1.5 Results ......................................................................................................................... 18
1.6 Summary and Discussion............................................................................................ 30
1.6.1 Displaced Families’ Expenditure Smoothing by Reducing Their Own
Wealth .......................................................................................................... 31
1.6.2 The Sluggish Fall in Displaced Families’ Expenditures .............................. 33
1.7 Conclusion .................................................................................................................. 36
APPENDICES .................................................................................................................. 37
APPENDIX A FIGURES FOR CHAPTER 1 ..................................................... 38
APPENDIX B TABLES FOR CHAPTER 1 ....................................................... 41
APPENDIX C SUPPLEMENTAL TABLES FOR CHAPTER 1 ....................... 47
APPENDIX D ADDITIONS FOR CHAPTER 1 ................................................ 56
REFERENCES ................................................................................................................. 61
CHAPTER 2 The Evolution of Earnings Volatility
During and After the Great Recession ............................................................. 64
2.1 Introduction ................................................................................................................. 64
2.2 Data ............................................................................................................................. 66
2.2.1 Matched CPS ............................................................................................... 66
2.2.2 Sample Selection.......................................................................................... 67
2.2.3 Measuring Earnings Volatility ..................................................................... 69
2.3 Results ......................................................................................................................... 71
2.3.1 Men’s and Women’s Earnings Volatility during and after the Great
Recession ..................................................................................................... 71
2.3.2 Subgroup Analysis by Education and Age .................................................. 79
2.3.3 Unemployment, Work Hours, and Earnings Volatility in the Great
Recession. .................................................................................................... 82
2.4 Conclusion .................................................................................................................. 87
APPENDICES .................................................................................................................. 89
APPENDIX A FIGURES FOR CHAPTER 2 ..................................................... 90
APPENDIX B SUPPLEMENTAL FIGURES FOR CHAPTER 2 ................... 100

vii

APPENDIX C SUPPLEMENTAL TABLES FOR CHAPTER 2 ..................... 116
APPENDIX D ADDITIONS FOR CHAPTER 2 .............................................. 123
REFERENCES ............................................................................................................... 127
CHAPTER 3 How Did Men’s Earnings Volatility Change during the Great Recession?
A Comparison of Evidence between the PSID and CPS .............................. 130
3.1 Introduction ............................................................................................................... 130
3.2 Data and Measurement ............................................................................................. 132
3.3 Results ....................................................................................................................... 135
3.4 Conclusion ................................................................................................................ 141
APPENDICES ................................................................................................................ 143
APPENDIX A FIGURES FOR CHAPTER 3 ................................................... 144
APPENDIX B SUPPLEMENTAL TABLE FOR CHAPTER 3 ....................... 149
REFERENCES ............................................................................................................... 151

viii

LIST OF TABLES

Table 1.B1 Descriptive Statistics for Non-displaced and Displaced Families (PSID 1999-2013)42
Table 1.B2 The Effects of Husband’s Job Displacement on Family Income (PSID 1999-2013) 44
Table 1.B3 The Effects of Husband’s Job Displacement on Family Expenditures (PSID 19992013) .......................................................................................................................... 45
Table 1.B4 The Effects of Husband’s Job Displacement on Family Wealth in 2013$ Terms
(PSID 1999-2013) ..................................................................................................... 46
Table 1.C1 Summary Statistics for Family Expenditures (PSID 1999-2013) .............................. 48
Table 1.C2 The Number of the Job Displacement Dummies (PSID 1999-2013) ........................ 50
Table 1.C3 The Effects of Husband’s Job Displacement on Family Food Expenditure (PSID ... 50
Table 1.C4 The Effects of Husband’s Job Displacement on Home Equity, Home Price, and
Housing Expenditures in 2013$ Terms ..................................................................... 51
Table 1.C5 The Probability of A Displaced Family’s Living in a Different, Low-Priced House
Compared to t-4 ......................................................................................................... 52
Table 1.C6 The Probability of A Displaced Family’s Living a Different, Low-Priced House .... 52
Table 1.C7 The Probability of A Displaced Family’s Living a Different, Higher-Priced House,
Compared to t-4 ......................................................................................................... 53
Table 1.C8 The Probability of A Displaced Family’s Living a Different, Higher-Priced House 53
Table 1.C9 The Effects of Husband’s Job Displacement on Sub-components of Family Income in
2013$ Terms .............................................................................................................. 54
Table 1.C10 The Effects of the Great Recession and Multiple Job Losses on Total Family
Expenditures of Displaced Husbands ...................................................................... 55
Table 2.C1 Match Rate ............................................................................................................... 117
Table 2.C2 Summary Statistics for Matched Men ...................................................................... 118
Table 2.C3 Summary Statistics for Matched Women ................................................................ 119
Table 2.C4 Earnings Volatility for Men and Women ................................................................. 120

ix

Table 2.C5 Men’s Earnings Volatility by Education and Age (DES for Sample B) .................. 121
Table 2.C6 Women’s Earnings Volatility by Education and Age (DES for Sample B) ............ 122
Table 3.B1 Descriptive Statistics for Sample A and B ............................................................... 150

x

LIST OF FIGURES

Figure 1.A1 The Effects of Husband’s Job Displacement on Annual Earnings, Family Income,
and Expenditures ....................................................................................................... 39
Figure 1.A2 The Effects of Husband's Job Displacement on Family Expenditures in Dollar
Terms ......................................................................................................................... 40
Figure 2.A1 Fraction of Zero-earners in Matched March CPS .................................................... 91
Figure 2.A2 Men’s Earnings Volatility (DG and DES; Ages 25-59) ........................................... 91
Figure 2.A3 Men’s Earnings Volatility (SS; Ages 25-59) ............................................................ 92
Figure 2.A4 Women’s Earnings Volatility (DES & DG; Ages 25-59) ........................................ 92
Figure 2.A5 Women’s Earnings Volatility (SS; Ages 25-59) ...................................................... 93
Figure 2.A6 Men’s Earnings Volatility by Education and Age (DES for Sample B) .................. 94
Figure 2.A7 Women’s Earnings Volatility by Education and Age (DES for Sample B) ............. 95
Figure 2.A8 The Yearly Fraction of Male and Female Workers Having Experienced
Unemployment .......................................................................................................... 96
Figure 2.A9 The Effect of Unemployment on Men’s and Women’s Earnings Volatility ............ 97
Figure 2.A10 Decomposition of DG for Men and Women in Sample A ..................................... 98
Figure 2.A11 Decomposition of Var(Δlog(hours)) for Men and Women in Sample A ............... 99
Figure 2.B1 Yearly Match Rate for Matched March CPS .......................................................... 101
Figure 2.B2 Men’s Earnings Volatility by Education and Age (SS for A &B).......................... 102
Figure 2.B3 Women’s Earnings Volatility by Education and Age (SS for A &B) .................... 103
Figure 2.B4 Men’s Earnings Volatility by Education and Age (DES for A) ............................. 104
Figure 2.B5 Women’s Earnings Volatility by Education and Age (DES for A) ........................ 105
Figure 2.B6 Men’s Earnings Volatility by Four Education Subgroups (DES for A) ................. 106
Figure 2.B7 Men’s Earnings Volatility by Four Education Subgroups (DES for B) ................. 106

xi

Figure 2.B8 Men’s Earnings Volatility by Four Education Subgroups (SS for A & B) ............ 107
Figure 2.B9 Men’s Earnings Volatility by Seven Age Subgroups (DES for A) ........................ 108
Figure 2.B10 Men’s Earnings Volatility by Seven Age Subgroups (DES for B)....................... 109
Figure 2.B11 Men’s Earnings Volatility by Seven Age Subgroups (SS for A & B) .................. 110
Figure 2.B12 Women’s Earnings Volatility by Four Education Subgroups (DES for A) .......... 111
Figure 2.B13 Women’s Earnings Volatility by Four Education Subgroups (DES for B) .......... 111
Figure 2.B14 Women’s Earnings Volatility by Four Education Subgroups (SS for A & B) ..... 112
Figure 2.B15 Women’s Earnings Volatility by Seven Age Subgroups (DES for A) ................. 113
Figure 2.B16 Women’s Earnings Volatility by Seven Age Subgroups (DES for B) ................. 114
Figure 2.B17 Women’s Earnings Volatility by Seven Age Subgroups (SS for A & B) ............ 115
Figure 3.A1 Men’s Earnings Volatility for Sample A (DG and DES) ....................................... 145
Figure 3.A2 Earnings Volatility for Sample A and B (DES) ..................................................... 146
Figure 3.A3 Earnings Volatility for Sample A and B (SS)......................................................... 147
Figure 3.A4 Decomposition of Earnings Volatility: Unemployed vs. Not unemployed ............ 148

xii

CHAPTER 1
The Effects of Job Displacement
on Family Expenditures
1.1 Introduction
The existing literature has extensively documented the persistent effects of job displacement
on a worker’s subsequent earnings. Although the estimates vary depending on the data set and
period, displaced workers have been found to have 10 to 30 percent less earnings than those they
would have without displacement, even six or more years after displacement (Couch and Placzek
2010; Jacobson, Lalonde, and Sullivan 1993; Ruhm 1991; Stevens 1997; Topel 1990; von
Wachter, Song, and Manchester 2007).
To better understand the welfare and policy implications of job displacements, further
investigation of how job displacement changes people’s consumption behavior is necessary
given that consumption is a better indicator of people’s well-being than is income. In general,
displaced workers’ consumption behavior cannot be inferred solely on the basis of income
change because a change in income does not necessarily lead to the same or proportionate
change in consumption. As predicted by life-cycle models of consumption and the permanent
income hypothesis, displaced workers who are forward-looking should smooth their
consumption by saving and/or borrowing. In practice, the ability of displaced workers to smooth
consumption may depend on the adequacy of their savings or borrowing ability and the
availability of any other insurance mechanisms to offset income loss, such as public/private

1

transfers and labor supply adjustments of other family members (Blundell, Pistaferri, and Preston
2008; Blundell, Pistaferri, and Saporta-Eksten 2016; Dynarski and Gruber 1997).
In contrast to the substantial literature on the income effects of job displacement, there is
relatively scant literature examining the link between job displacement and consumption by
using micro expenditure data. The majority of studies focus on only short-term change in
workers’ expenditures one or two years after displacement, or, if they do track the long-term
expenditure response of displaced workers, rely only on a limited number of expenditure
categories, such as food and some other non-durables (Browning and Crossley 2008; Dynarski et
al. 1997; Gruber 1997; Saporta-Eksten 2014; Stephens 2001).1 Hence, there is a need to explore
how displaced workers change their consumption behaviors, particularly over a longer period
during which they experience persistent income loss.
To fill this research gap, the current paper studies how a husband’s job displacement affects a
broad set of his family expenditures over an extensive period of time from roughly two years
before the displacement through six or more years after afterwards using the 1999-2013 Panel
Study of Income Dynamics (PSID). Furthermore, this study tracks concurrent changes in diverse
components of family wealth and income of a displaced husband that underlie the family’s
expenditure behaviors in order to explicitly capture the contribution of savings/assets and other
family income sources that offset his earnings loss due to job displacement. This comprehensive
analysis on the long-term effects of job displacement is possible mainly because of the
availability of rich information on an extensive range of indicators of family expenditures and
wealth, which have been biennially collected since 1999 by the PSID.
Like a number of previous studies, this study directly analyzes expenditures rather than
consumption, although consumption is the concept more clearly linked to individual well-being.
1

A more detailed review of the literature is presented in Section 1.2.

2

While there is little difference between the two for non-durable goods, a large difference can
exist for durable goods. Simply put, a one-time expenditure on a durable such as a car could
provide substantial consumption flows for multiple years. Despite this distinction, I analyze
expenditures for three reasons. First, income, expenditures, and changes in assets are linked
through an accounting identity, and I make use of this identity to provide additional insight into
how a family with a displaced husband (hereafter referred to as a displaced family) respond to
his job loss. Second, accurately converting expenditures into consumption requires much
additional data and assumptions, which can lead to substantial measurement error. Third,
changes in expenditures can still provide important insights into how consumption is changing.
The key findings of this paper can be summarized as follows. First, during the period 19982012, the annual total income of displaced families is reduced by 16 percent, on average, over
the long period of time noted above (roughly two years before to six or more years after
displacement) compared to what they would have without a husband’s displacement. This large,
persistent income loss was mainly because the husband went through a substantial loss in annual
earnings due to job displacement over the same period of time, and there was no significant role
of other insurance mechanisms such as spouses’ labor earnings and public/private transfers in
compensating the earnings loss of displaced husbands.
Second, even with a large income loss, displaced families practiced considerable expenditure
smoothing for a broad set of major expenditure categories related to food, housing, transportation,
health care, and education (hereafter referred to as total expenditures). In particular, they did not
decrease the expenditures significantly until the year of displacement, notwithstanding the
roughly 12% of their annual income loss that they concurrently had by then. Roughly two years
after displacement, their expenditures began to fall significantly by five percent, and decreased

3

further for the subsequent years, so that they had nine percent lower expenditures roughly six or
more years after displacement, compared to what they would spend without displacement. This
sluggish fall in the total expenditures of displaced families is not apparent in their food
expenditures alone, which have been heavily relied on by a number of previous studies to
represent a family’s whole consumption behavior. This study, in fact, like that of Stephens
(2001), finds that their food expenditures began to fall even before the occurrence of a husband’s
displacement and also manifested a similar magnitude of decrease as the initial one even six or
more years after displacement. Unlike the food expenditures, however, the other non-food
expenditures of displaced families appeared not to have this early decrease in the year of, or prior
to displacement found in food expenditures; their expenditures for housing, health care and
education, in particular, began to fall quite slowly, which mainly drove the entire sluggish fall in
their total expenditures. The decline in their food expenditure turns out to have explained only 15%
of the lagged decrease in their total expenditures that began roughly two years after displacement.
Lastly, I find that the main tool of the large expenditure smoothing of displaced families was
to decrease their own wealth relative to what they would have without the displacement. They
began to decrease their total wealth, even before the occurrence of displacement, and ended up
with $172,000 less family wealth, on average, by roughly six or more years after displacement,
than what they would have without it. This amount of wealth reduction equals 45% of the sample
average of total family wealth in this study, and also covers roughly 55% of their disposable
income loss during the same time. Furthermore, the speed of the decline in their family wealth
was decreasing over time, which is consistent with their sluggish decrease in their total
expenditures described above: A large decrease in their wealth occurred in the initial periods
around the occurrence of job displacement to the extent that it roughly offset the total loss in

4

their disposable family income during that same initial period. In turn, the annual decrease in
their wealth for subsequent periods became less prominent covering only 40% of their annual
disposable income loss that they had since roughly two years after displacement.
The remainder of the paper is structured as follows: Section 1.2 reviews relevant previous
studies and highlights the contributions of this study to the existing literature; then, main
econometric approaches and data set are described in Sections 1.3 and 1.4 respectively; Section 5
presents empirical results, and Section 6 summarizes the main findings of the paper and
discusses respective implications. The conclusion is presented in Section 7.

1.2 Literature Review
As noted earlier, there are a relatively small number of studies exploring the relationship
between job displacement and consumption behaviors in contrast to the extensive literature
investigating the adverse effect of displacement on workers’ wages or earnings. 2 In particular,
many of the studies exploring the consumption effect of joblessness have focused on analyzing
the short term dynamics of expenditures. Dynarski and Gruber (1997) examine the immediate
family expenditure response to the year-to-year change in earnings due to a head’s
unemployment shock, and find evidence of considerable expenditure smoothing. Gruber (1997)
documents a significant role of unemployment insurance benefit in unemployed workers’
expenditure smoothing. According to his estimates, the unemployment shock to a head resulted
in 6% less family food expenditures within a year, on average, compared to the previous
expenditure level before the shock. Browning and Crossley (2008) find that permanently laid-off
2

See Couch and Placzek (2010) and Kletzer (1998) for summaries of papers related to the income effects of
displacement.

5

workers reduced their total expenditures by 4 to10% within 4 to 9 months after displacement,
which represented a loss of two-year normal expenditure growth of temporarily laid-off workers,
who were considered a control group fully insured against displacement shocks by the authors.
Stephens (2001) and Saporta-Eksten (2014) explore the long-term relationship between job
displacement and nondurable expenditures. Both authors find that the households with displaced
heads practiced sizable expenditure smoothing through the whole post-displacement period.
However, they do not further explore the link between the expenditure smoothing and dissaving
/borrowing behaviors explicitly among household with displaced heads, as I do in the current
study.
Specifically, Stephens (2001) considers the long-run dynamics of family consumption
resulting from a husband’s earliest displacement. Unlike the current study using a broad set of
expenditure data, he relies solely on food expenditure data, which was the only nondurable
expenditure data available in the PSID for 1968-1992, and find a much more moderate and less
volatile decrease in food expenditures of displaced husbands’ families that continued for more
than 6 years after displacement compared to the severe decrease in family income they
experienced during the same period.
Using more extensive family nondurable expenditure data in the PSID 1999-2009, SaportaEksten (2014) also explores the long-term dynamics of nondurable expenditures before and after
job displacement. He finds there was an overall larger decrease in nondurable expenditures
before and after displacement than that in food expenditures found by Stephens (2001). But in
terms of changing patterns in nondurable expenditures around displacement occurrence, his
findings, are similar to those of Stephens (2001) using food expenditures alone.

6

Saporta-Eksten's (2014) study differs from the current study in several respects. He uses a
smaller set of expenditure data excluding durable expenditures and a shorter sample period for
the analysis. 3 Furthermore, the author uses only pooled OLS for the analysis without isolating
the earliest displacement from other subsequent ones for each household’s head. This is an
important departure from the current study and that of Stephens' (2001) ; in contrast to SaportaEksten's (2014) study, both Stephens (2001) and I focus on the first-observed displacement of a
husband and its long-term effects, and mainly use the fixed effect estimation to control
unobserved time-invariant component for each household, which has been considered important
for mitigating endogeneity issues in the literature (Jacobson, Lalonde, and Sullivan 1993;
Stephens 2001; Stevens 1997). For that reason, Stephens' (2001) findings are more comparable
to those from this study than the findings from Saporta-Eksten (2014).
The current study is also related to another strand of literature that empirically tests life-cycle
theories and the permanent income hypothesis. A number of studies have used micro data to
focus on exploring agents’ consumption responses against well expected changes in income, and
the empirical results diverge (Browning and Collado 2001; Hsieh 2003; Parker 1999; Souleles
1999). Browning and Crossley (2001) argue that such divergent results may largely stem from
the differences in magnitudes of the expected income change across those studies; the greater the
expected income change relative to life income, the greater its effect on people’s current
consumption. Accordingly, empirical studies using larger expected income variations would
report more favorable results to the predictions of the standard life-cycle theories and the
permanent income hypothesis.

3

The author uses the expenditure data only for nondurables, thereby excluding from the analysis other
expenditure data for major durables related to housing and vehicles.

7

In a more comprehensive approach related to testing life-cycle theories and the permanent
income hypothesis, Blundell, Pistaferri, and Preston (2008) identified transitory and permanent
income shocks from households’ earnings variations by assuming a general earnings equation.
They showed that permanent income shocks were much less insured than transitory ones, thereby
leading to a larger change in family expenditures.
Along this line of literature, what people expect the future effects of a permanent income
shock will be and how such expectation affect their consumption behaviors over a long period
are interesting empirical research question, though relatively unexplored. Given that job
displacement is a well-documented source of permanent income shocks, the current study
provides useful information for examining those questions, through documenting long-term
changes in expenditure/saving behaviors of displaced families using rich data on family
expenditures and wealth.

1.3 Econometric Approach
I use a simple reduced-form approach, which has been extensively used by other relevant
studies exploring the long-run effects of job displacement (Stephens 2001; Stevens 1997). The
main model specification is,
𝑌𝑖𝑖 = 𝐷𝑖𝑖 β + 𝑋𝑖𝑖 γ + αi + δt + 𝜖𝑖𝑖 ,

where 𝑌𝑖𝑖 is an economic outcome for family i in year t, such as total annual family income

and its sub-components, total annual family expenditures and its subcomponents, and total
family wealth and its sub-components.

8

I use both log and level specifications for the dependent variables in the analysis. On the one
hand, using log dependent variables enables me to interpret the regression coefficients in terms
of percentage points, which makes it easier to assess the relative impacts of job displacement and
is also preferred in the literature on theoretical grounds for analysis regarding income and
expenditure dependent variables (Deaton and Muellbauer 1980; Mincer 1974; Zeldes 1989). On
the other hand, using level dependent variables is useful when analyzing dependent variables that
are additively related, as are many of the dependent variables in my analysis. For example, I can
analyze how each disaggregated sub-component contributes to the total job displacement effects
on an aggregate dependent variable in dollar terms because the linear regression coefficient for
the aggregate level dependent variable is merely the sum of the corresponding regression
coefficients for all the aggregate variable’s exclusive and exhaustive sub-components. Similarly,
I can examine how job displacement affects the family budget constraint because a family’s
change in wealth equals the difference between income and expenditures. A further advantage of
using level dependent variables is that I can include zero or negative values of the dependent
variables in the regressions, which is not possible when using log dependent variables.
The vector 𝐷𝑖𝑖 refers to the dummies indicating the first job displacement that occurred to the

husband of a household i in the past, currently, or in the future at year t. Therefore, the

coefficients for the job displacement dummies basically capture the mean log or level differences
in family incomes, expenditures, or wealth of displaced heads’ families from those they would
have without experiencing their heads’ displacement. 𝑋𝑖𝑖 is a vector of time-varying variables,
which is expected to capture dynamic changes in family expenditure preference; 𝑋𝑖𝑖 includes

logarithmic family food standard variable 4, quadratics of the head’s age, 5 the number of family

4

These are variables which were generated by the PSID from 1968 to 1992, based on the number of family
members by gender and age. By using the same methods, I create the same variables for 1999-2013.

9

members between the ages of 0 and 5, 6 and12, 13 and17, 18 and 64, and 65 or older. Îąi stands
for family-specific time-invariant components affecting family behavior. A large body of

literature has underscored the importance of controlling the unobserved effect because it is likely
to be correlated with the probability of job displacement (Couch and Placzek 2010; Jacobson,
Lalonde, and Sullivan 1993; Stevens 1997). To control the unobserved effect, I employ the fixed
effect estimation method. δt is a vector of year dummies controlling the aggregate time effect,

which may affect family economic activities globally. 𝜖𝑖𝑖 is the error term, which is clustered by
family to obtain standard errors robust to both serial correlation and heteroskedasticity.

1.4 Data
The Panel Study of Income Dynamics (PSID) is one of the longest nationally representative
panel data, and has collected a variety of socioeconomic information annually from 1968 to 1996,
and biennially from 1997 on. In particular, it contains a set of variables related to employment
history for household heads, from which whether and when a head experienced job displacement
can be identified. Hence, the PSID have been extensively used to study the long-term effects of
job displacement on family earnings and income in the literature. Although the PSID collected a
wide set of diverse family income information, its documentation of family expenditure
information was quite limited before 1999; family food expenditure and rents/mortgage
payments were the only categories that the PSID collected for most years during this period from
1968 to 1997.

5

Substituting a head’s potential experience for a head’s age in the regressions for diverse income dependent
variables also gives very comparable results

10

Since 1999, however, this limitation of the PSID has been alleviated significantly. The PSID
began to collect data on a biennial basis after 1997, enlarging its sample size and the number of
variables covered by the survey. Along these lines, it began in 1999 to collect extensive
information on family expenditures for major durables and non-durables related to housing,
transportation, health care, education, as well as food. 6 Moreover, diverse family wealth
information has also been collected biennially since 1999, covering a broad range of family
wealth indicators, such as home equity, cash in hand (checking/savings account), net values of
vehicles, business or farm, other estates, stocks, annuities, and non-collateralized debts.
Therefore, the PSID for 1999-2013 provides a distinctive opportunity to directly explore a topic
that has heretofore received scant attention in the literature: how a husband’s displacement
impacts major family expenditures as well as diverse family income sources in the long-run, and,
furthermore, how it changes the net values of a wide range of family savings and assets during
the same time.

1.4.1 Constructing Annual Family Expenditures
In the 1999-2013 PSID, there are a number of inconsistencies in timing and time units across
the extensive expenditure variables, unlike the variables related to family income and wealth. To
minimize these inconsistencies, I develop a proxy for annual expenditures during the previous
year of each interview year for each expenditure category. 7 I then calculate total annual family
6

Since 2005, the PSID has encompassed broader expenditure categories, such as expenses for home
repair/maintenance, furnishing, clothing, and entertainment/vacation. For consistency of analysis, those variables are
not used in the main analysis of this study.
7
The variables related to annual family income in the PSID 1999-2013 mostly refer to the year prior to each
survey year. However, all the variables related to family wealth refer to the time around the interview month which
is mostly between March and May for each survey year, while the reference time for expenditure variables differs
widely across expenditure categories. In this study, I consider each wealth variable as a proxy for the amount of
wealth accumulated by the end of the year prior to each survey year. Similarly, I construct annual expenditure
variables for each expenditure category to obtain a proxy capturing, as closely as possible, what a family spent for

11

expenditures for food, housing, transportation, health care, and education by summing up all the
corresponding sub-component expenditure categories.
The specific sub-components summed up for each category are the following: the food
expenditures include the annual expenditure for food-away-from-home, as well as that for foodat-home, which also includes the annual monetary value for delivered food and food stamps. The
housing expenditures are calculated by summing up annual rent or mortgage payments, home
utility expenses, such as electricity and heating, home insurance premiums, and property taxes.
The transportation expenditure refers to the summation of average annual expenses for
purchasing or leasing car(s), other annual car operation costs such as fueling, car repair or
maintenance, car insurance premiums, and annual public transportation costs for using buses,
taxis, and trains. Note that the mortgage payments within the housing expenditures and carpurchasing expenses within the transportation expenditures encompass some expenditures for
durables as well as non-durables.
As for the health care expenditures, I sum the annual expenses for doctor appointments,
surgery, dental treatment, hospitalization, nursing care, prescription medicine and health
insurance. Lastly, the education expenditures encompass both annual school- related costs and
child care costs. All the expenditures described above are deflated to 2013 dollar terms using
CPI-U-RS from the Bureau of Labor Statistics.
Table 1.C1 shows descriptive statistics for annual family expenditures for food, housing,
transportation, health care, and education, with all the corresponding sub-component categories
for the main sample. Housing expenditures account for the largest portion of these expenditures,
about $21,000 on average. The second and third largest categories are transportation and food

the year prior to each survey year. Refer to Appendix A for a more detailed description of the procedures for
generating annual expenditure variables.

12

expenditures, totaling $15,000 and $10,000, on average, respectively. The largest category of
total annual family expenditures, encompassing food, housing, transportation, health care, and
education adds up to $56,000 on average. 8

1.4.2 Job Displacement
Since 1968, the PSID has contained a series of questions that ask household heads about the
reasons why they left their previous jobs or employers, if their jobs were changed recently.

9

Based on the answers to those questions, job separation due to company closings, being laid off
or fired is defined in this study as job displacement. 10 Thus, I focus on the job separations which
can be considered exogenous shocks to workers. Note that I employ the conventional definition
of job displacement found in the previous studies on job displacement using the PSID (Stephens
2001; Stevens 1997).
By using those questions to investigate the causes of heads’ recent job separations, I identify
a husband’s experience of job displacement and its timing for each household in the following

8

One issue in constructing larger expenditure categories by adding up multiple sub-components is that the greater
the number of sub-components that are summed up, the worse the missing data problems become. To mitigate this
missing data problem, I perform selective imputations for missing data of some variables such as home insurance,
property tax, and car insurance, which have a relatively large amount of missing data. Furthermore, I do not include
water/sewage expenses in housing expenditures because those expenses have a large amount of missing data, while
accounting for a relatively small portion of the total expenditure for housing. Table 1.C1 also shows the statistics of
the variables containing the imputed variables above, and they appear to be comparable to those of the non-imputed
ones. The total expenditures shown in the bottom of Table 1.C1 are constructed by using the imputed variables to
alleviate the missing data problem.
9
More specifically, in the 1999 and 2001 waves, the PSID asked currently-working heads about the reasons for
their job changes if they had worked in other jobs in the year prior to the interview year. It also asked currently-nonworking heads about their most recent job, or up to two jobs that they had had in the year prior to the interview year,
and recorded reasons why that job or jobs had been terminated. Since 2003, the PSID has begun to investigate up to
four recent jobs which a head had had in the year prior to the interview year whether the head is currently working
or not, and recorded the reasons for job separation(s) if the head no longer had the job(s) at the time of the interview.
10
According to Boisjoly, Duncan, and Smeeding (1998), of the respondents answering that they were laid off or
fired in the PSID for 1968-1992, roughly 16% had actually been fired.

13

way: 11 In a certain interview year, if a husband responds to these questions by stating that he had
stopped working in a job he held the prior year, due to a company-closing, being laid off or fired,
the husband is considered to have experienced job displacement between January of the year
prior to the interview year and the interview month of the interview year (mostly from March to
May). If a husband experienced multiple job displacements through multiple interview years,
only the first observed job displacement is accounted for in this study, as Stephens (2001) and
Stevens (1997) do. A primary reason for focusing on the first job displacement is because a
household head’s subsequent displacements following the head’s first one were found to play a
central role in generating the persistent earnings loss of displaced heads (Stevens 1997). This
finding implies that the subsequent job losses may be an important consequence of the first one.
Thus, this study focuses on exploring the effect of the earliest job displacement of a husband
since he started to be interviewed by the PSID.
This identification strategy for the first occurrence of job displacement works well before
1997, when the PSID collected the data annually. Since 1997, however, when the PSID began to
collect the data biennially for each household head, it has systemically generated certain periods
(mostly between June and December of odd-numbered years between 1997 and 2013), during
which it is not possible to determine whether or not the head experienced job displacement. This
is because the PSID has continued to document a head’s reasons for job separations only for
separations occurring in the previous year or current year of each interview year, even after it
began to survey biennially after 1997.
This data limitation raises several issues that should be kept in mind when the regression
results are interpreted in Section 1.5 of this paper. First, the effects of job displacement occurring

11

The sample of this study, which is discussed later, consists of only households whose head has a spouse, and for
these households, the husband is considered a default head by the PSID

14

in the second half of odd-numbered years during the 1997-2013 period cannot be identified.
Hence, the estimated effect of job displacement in this study should be regarded more as an
average effect accounting only for the displacements occurring in even numbered years or in the
first quarter or first half of odd-numbered years between 1997 and 2013. Second, the sample of
non-displaced husbands from the period of 1999 to 2013 identified in this study would include a
fraction of husbands who were actually displaced in the second half of odd-numbered years
during that earlier commencing period, 1997 to 2013. This would underestimate the effect of a
husband’s job displacement on family expenditures, and the degree of underestimation would
depend on the percentage of the misidentified non-displaced compared to the correctly-identified
non-displaced. 12 Third, some instances of identifications of the first job displacement occurrence
may be incorrect as they may have occurred in the second half of earlier odd-numbered years
between 1997 and 2013. 13 14 15

1.4.3 Sample Selection
The main sample of this study consists of households which have both a husband (defined as
the household head by the PSID) and his wife, whose ages are between 25 and 65 for each

12

In fact, the potentially misidentified group appears to account for just a small portion of the whole nondisplaced group. Based on the observations in the second half of even-numbered years during 1997-2013, only 13%
of heads experienced job separations, and among those separations, only 13% were due to job displacement. Hence
the degree of underestimation is expected to be moderate. To further examine the degree of underestimation, I also
performed a quasi-experiment, by only using the data of odd-numbered years (or even-numbered years) from the
annual data set of the PSID during 1968-1997; the results show that the degree of underestimation is quite small.
More detailed results are presented in Table 1.C3.
13
These cases also seem to be minimal in number, according to the discussion in Footnote 12.
14
As a way of mitigating the misidentification problems for displaced husbands described here, I specify the
husbands who were likely to experience displacement in the second half of the odd-numbered years between 1997
and 2013. This likelihood is judged by using other information on whether the head had some unemployed periods
or received unemployment benefits around that time. I then drop some of the heads’ observations which come after
the period when they were suspected to be displaced. Even with this further sample restrictions, I find that all the
main empirical results in this study hold robustly.
15
See Table 1.C2 for detailed numbers of the first job displacement dummies by year, which are used by this
study.

15

sample year. 16 More specifically, for each male head (or husband) observed by the PSID at least
twice during the period 1968-2013, I specify the earliest survey year when he reported a spouse
satisfying the age criteria above. Then, from that survey year on, I include him and his family in
the sample of this study unless he divorces, separates, misses/leaves the survey, or either his or
his wife’s age become over 65. Furthermore, I use only the households whose heads’ labor
earnings and total family income were positive for the whole observed period. Thus, the main
sample of this study refers to relatively stable families in terms of their family structure,
continuity of survey participation, and husband’s attachment to the labor market. This sample
stability is expected to be helpful for identifying the pure effects of the first job displacement of a
husband, as Stephens (2001) discusses.
Among the households selected based on the criteria described above, only the families
which were observed during the period 1999-2013 are used in the analyses because the main
analysis of this study is on the various family expenditures and wealth data of the PSID during
the same period. Moreover, to obtain a more consistent measure of job displacement occurrences,
I drop the families which reported their male heads’ displacement in the PSID waves prior to
1997, when the PSID collected data on an annual basis. As a result, the displaced husbands in the
main sample of this study refer to those who were first observed to have job displacement in the
1997 PSID wave or later ones, while satisfying all the sample selection criteria listed above.
Along the same lines, the non-displaced husbands in the main sample of this study indicate those
16

I use only national representative samples, the SRC (Survey Research Center) and immigrant samples of the
PSID. That is, I exclude the SEO (Survey of Economic Opportunity) sample, which oversamples low-income
families, and the Latino sample, which was dropped after 1995. The SRC sample refers to the nationally
representative families who were selected in 1968 and their subsequent split-off families. The immigrant sample
refers to national representative immigrants who were added to the PSID since 1997 to reflect the large increase in
immigrants in the U.S. The PSID provides longitudinal (and therefore time-varying) family weights accounting for
both the core and immigrant samples. From these weights, I calculate two types of fixed family weight for each
family: first, I select the last observed family weight within a family. Second, I average out the family weights over
time within a family. Both weights lead to quite comparable results. All the results shown hereafter are ones based
on the first type of family weight.

16

who satisfy the same selection criteria as the displaced ones, except that they had never reported
the experience of job displacement.
I further drop from the main sample the observations that have missing data for any variables
used in the regression analyses of this paper. Consequently, the number of total observations of
the main sample is 9,924 for 2,245 families.

1.4.4 Descriptive Statistics for the Non-displaced and Displaced
Table 1.B1 presents the summary statistics for non-displaced and displaced families for the
PSID waves 1999-2013. Those for the latter group are calculated by pooling all the ex post
observations after displacement, and those for the former group are based on the observations of
the households whose heads have never been displaced, or have not yet been displaced during
the period 1998-2012 (based on the PSID waves 1999-2013).
Table 1.B1 shows that the displaced husbands and their spouses were, on average, slightly
older than the non-displaced ones. The mean schooling years for the former were also somewhat
lower than those of the latter. In addition, the displaced husbands were slightly less likely to be
white. The other demographic variables, such as the number of children show fairly comparable
mean values for both groups.
This study finds no significant difference in the probability of working in the manufacturing
industry in a comparison of the two groups. The displaced husbands were more likely to work in
blue-color jobs, however, and less likely to have a job covered by a union contract or to be a
union member.
When comparing income levels, the difference between the two groups becomes more salient.
The displaced husbands experienced rather large reductions in annual earnings and total family

17

income, on average, after being displaced. Consequently, the proportion of the spouse’s earnings
in total family income increased considerably after a head’s job displacement. A significant
increase in the mean of public transfer is observed too, but the magnitude of the compensation is
fairly small compared to the total income loss due to job displacement.
The next panel in Table 1.B1 shows the statistics for family expenditures. With the large
reduction in family income, the displaced families also seem to have reduced various family
expenditures on average, but the magnitudes of decreases were relatively moderate compared to
the average income loss. In addition, considerable reductions in the average value of various subcomponents of family wealth are found in the panel at the bottom of Table 1.B1. Taken together,
these observations imply that much of the expenditure smoothing was executed by the displaced
families and their substantial dissaving/borrowing (or decrease in active saving) played an
important role in alleviating the reduction in their expenditures.
Referring only to those unconditional means, however, is not sufficient to isolate the pure
effects of job displacement. To clarify the role of job displacement further, I employ the fixed
effect estimation method, which controls husbands’ age effects, change in age compositions of
family members, year effects, and each family’s time-invariant unobservable components, as
discussed in Section 1.3. The regression results follow in the next section.

1.5 Results
Figure 1.A1 presents both the main findings of this paper and how they compare to previous
results in the literature. Specifically, the left panel A shows the average evolution of a displaced
husband’s annual earnings, his family’s annual total income, and food and total expenditures

18

during the period of 1998 to 2012 (the 1999-2013 waves of the PSID); the point estimates are in
panel A of Table 1.B2 and Table 1.B3 which are later discussed. For comparison, the right panel
B shows what Stephens (2001) finds by using the PSID for 1968-1992. As can be seen, all the
corresponding graphs to those in panel A of Figure 1.A1, except for total family expenditure, are
found in panel B of the same figure. For both panels, point t on the x-axis indicates the year
when the first displacement occurred, while the dotted horizontal line at zero on the y-axis
implies the level of income or expenditures that a husband or his family would have without
displacement. Note that in the remainder of this paper, all the quantitative comparison statements
for displaced families made from regression results are based on this counter-factual horizontal
line, although, for the sake of brevity, I will generally not repeat this comparison criterion
explicitly.
The two graphs at the bottom of panel A of Figure 1.A1 first show that a displaced husband
had a sizable and persistent earnings loss during the period 1998-2012, on average, and his
earnings loss also led to a considerable fall in his total family income for the same period.
Specifically, a displaced husband’s annual earnings began to decline by 12% at two to two and a
half years before displacement (t-2), and further decreased from the year of, or within a half year
before displacement (t) through one and a half to two years after displacement (t+2), ending up
with a 36% lower level compared to what he would earn without displacement. After three and a
half to four years after displacement (t+4), his earnings showed a recovery pattern somewhat, but
still remained at a 23% lower level even five and a half or more years after displacement (t+6
plus). The displaced family’s total annual income also showed a similar dynamic pattern,
although the magnitude of decrease was smaller in general; a displaced family had 8%, 13%,
23%, 20%, and 18% less annual total income during the periods t-2, t, t+2, t+4, and t+6 plus,

19

respectively, relative to what it would have without its male head’s displacement. All those
decreases in a displaced husband’s earnings and his family income are statistically significant
(see Column (1) and (3) in Panel A of Table 1.B2), and the decreasing patterns also appear
comparable with those previously found by Stephens (2001) using the PSID 1968-1992 (see the
graphs with square- and triangle shaped points in panel B of Figure 1.A1). 17
The smaller percentage fall in the total income of a displaced family compared to that of a
displaced husband’s earnings is partly due to the fact that a husband’s earnings only account for
65-70% of his total family income in the main sample of this study (composition effects).
Moreover, the lesser fall might be partly induced by some increases in other family income
sources, such as the spouse’s earnings and public/private transfers, which offset a fraction of the
earnings loss of a displaced husband (insurance effects). The two effects are difficult to
disentangle from each other based solely on the results from the log dependent variable analyses
shown in panel A of Figure 1.A1 or Table 1.B2. In the level dependent variable analyses
presented in panel B of Table 1.B2, however, we can see how each family income source was
affected by a husband’s displacement and ended up being aggregated into total family income in
dollar terms, so that we may compare the composition and insurance effects more clearly.
Specifically, columns (1), (2), and (3) in panel B of Table 1.B2 demonstrate the evolution of
job displacement effects on husband’s annual earnings, the sum of husband’s and wife’s annual
earnings, and their total family annual income in 2013 dollar terms, respectively. Unlike the log
dependent analyses seen in panel A, the level dependent variables for those columns in panel B
have additive relationships with each other, as do all the corresponding regression coefficients as
well. Hence, the differences in the coefficients in columns (1) and (2), and columns (2) and (3) in

17

Note that I use the before-tax total family income, while Stephens (2001) uses the after-tax one, which was
imputed by the PSID until the 1992 wave. The after-tax family income was not provided in the 1999-2013 PSID.

20

panel B indicate each individual effect of a husband’s displacement on his wife’s earnings and
the other family income sources, respectively, in dollar terms. 18
According to the results presented in panel B of Table 1.B2, the gap in the percentage fall
between a displaced husband’s earnings and his family income appears to be mainly due to the
composition effect rather than the insurance effect. Comparing the level differences in
coefficients between the columns in panel B of Table 1.B2, both a wife’s earnings and other
family income of a displaced husband turn out to have decreased slightly in dollar terms rather
than increased during the period 1998 to 2012, although those differences are not statistically
significant. This observation suggests that other family income sources played a minimal role in
buffering the earnings shock of displaced husbands during the period 1998-2012. 19
Even with such a substantial reduction in the total family income of displaced families as
shown in Figure 1.A1/Table 1.B2, they decreased their annual total expenditures only
moderately during the period 1998-2012, as is also seen in the graph with diamond-shaped points
in panel A of Figure 1.A1. 20 In particular, a significant fall in their total expenditures appeared
far later compared to that in their total income. For instance, the displaced families did not
decrease their total expenditures significantly during the period between t-2 and t (see column (1)
in panel A of Table 1.B3 for the standard errors), despite the fact that they had a significant fall
in their total annual income by 11% on average for the same period, as seen in Table 1.B2.
Roughly two years after a husband’s displacement (t+2), the total family expenditure started to
18

Other family income sources include the earnings of family member(s) other than husband and wife,
public/private transfers, interests/dividends from family assets.
19
This observation seems inconsistent with Stephens (2002)’s finding based on the 1968-1992 PSID that a
displaced husband’s wife increased her labor supply significantly and persistently before and after her husband’s
displacement. One factor driving this inconsistency might be the difference in the growth of female labor market
participation in the United States between the periods 1968-1992 and 1999-2013; while there was a strong
increasing trend in the female labor force participation rate (FLFP) during the former period, the growth in FLFP
stagnated substantially during the latter period (Lee 2014).
20
As explained earlier, total expenditures include expenditures for food, housing, transportation, health care, and
education.

21

statistically significantly decline by 5%, and further decreased with stronger statistical
significance for the subsequent period, so that the displaced heads’ families showed 9% less total
expenditures roughly four or more years after displacement (i.e., t+4 and t+6 plus), relative to
what they would spend without a husband’s displacement.
It is notable that this sluggish fall in total expenditures of displaced families is difficult to
detect in what Stephens (2001) finds about their food expenditure behaviors alone by using the
PSID 1968-1992. As seen in the graph with x-shaped points in panel B of Figure 1.A1, he finds a
significant earlier drop in their food expenditures even before the occurrence of displacement (t2 and t-1). In addition, in his study, the degree of decrease in their food expenditures changed
little over the whole post-displacement period, after their food expenditures had fallen by 9% in
the year of displacement (t). Consequently, displaced families showed a flat trend in their food
expenditures between the periods t and t+6 plus. Stephens (2001) regards this flat pattern as
evidence of consumption smoothing of displaced families who anticipated much of their future
income loss.
As for the food expenditures alone, the results of this study are comparable to those of
Stephens' (2001) for displaced families. For example, the graph with x-shaped points in panel A
of Figure 1.A1 (or column (2) in panel A of Table 1.B3) shows that displaced families had a
significant early drop in their food expenditures over the initial period of time around the
occurrence of displacement during the period 1998-2012; their food expenditures fell by 5% and
8% at t-2 and t, respectively, and those declines were statistically different from zero. Moreover,
even roughly six or more years after displacement (t+6 plus), the decrease in their food
expenditures appeared to be as much as in the earlier period t, although the full trajectory of their

22

food expenditures between t and t+6 plus seems not as flat as observed in Stephens' (2001)
results during the corresponding period. 21
The initial decline of food expenditures, and their leveling off thereafter suggests that the
sluggish fall in total expenditures including food expenditures would be driven by other nonfood expenditures. Indeed, the other regression results using log dependent variables for nonfood sub-expenditures, shown in column (3)–(5) in panel A of Table 1.B3, displays a tendency
for those non-food sub-expenditures to decline more slowly compared to the food expenditures.
As discussed earlier regarding panel A of Table 1.B2, however, it is hard to clearly see how each
sub-expenditure category contributed to the dilatory decrease in total expenditures through the
log dependent variable analyses. Moreover, the log dependent variable analysis has to drop some
observations with zero values for each sub-expenditures regression, which leads to different
sample sizes across the sub-expenditures analyses, as shown in panel A of Table 1.B3. This
constraint also impedes a fair comparison among them.
To mitigate the limitations of the log dependent variable analysis described above, I put the
corresponding results of the level dependent variable analyses in the lower panel B of Table 1.B3,
as I do in Table 1.B2. Specifically, columns (1)-(5) in panel B of Table 1.B3 show the level
effects of a husband’s job displacement in 2013 dollar terms on total annual family expenditures
and its exhaustive sub-components, such as annual family expenditures for food, housing,
transportation, and health care/education, in that order. Correspondingly, I also depict five panels
A through E in Figure 1.A2 to visualize each result in columns (1)-(5) in panel B of Table 1.B3,
with 95% confidence intervals. As the measure of total family expenditures is the summation of

21

The smaller decreases in food expenditures of displaced families between t+2 and t+4 compared to the previous
period appeared to be partly affected by a significant increase in their food stamp usage since t+2, during the period
1998-2012. Based on my replication of Stephens (2001) using the 1968-1992 PSID, however, no significant role of
food stamp usage is found in food expenditure smoothing for displaced families, during the period 1967-1991.

23

the other sub expenditures above, each coefficient of job displacement dummies in column (1) of
panel B is also the simple summation of the corresponding coefficients of the other subexpenditure regressions from column (2) through (5) of panel B. Thus, it is possible to compare
each sub-expenditures’ contributions directly from panel B of Table 1.B3, as can similarly be
done in panel B of Table 1.B2.
Overall, panel B of Table 1.B3 shows comparable dynamic patterns for each sub-expenditure
as well as total expenditures of displaced families with what we see in upper panel A of the same
table. For example, column (2) in panel B of Table 1.B3 (or panel B in Figure 1.A2) shows an
early significant fall in displaced families’ food expenditures as does the same column in upper
panel A; for both early periods t-2 and t, displaced families statistically significantly reduced
their food expenditures by $800 and $900, respectively. When it comes to the total expenditures
for the same period (column (1) in panel B of Table 1.B3), however, the declines appear not to
be statistically different from zero, which is also consistent with the results shown in the same
column of the upper panel. Referring to the corresponding estimates in columns (3) through (5)
in panel B of Table 1.B3 (or panels C, D, and E in Figure 1.A2), the statistically insignificant fall
in total expenditures at t-2 and t, turns out to have been mainly because displaced families did
not have a significant fall in the other non-food sub-expenditures during the same periods, for
housing, transportation, health care, and education. 22
From t+2 through t+6 plus in panel B of Table 1.B3 (or in Figure 1.A2), in contrast,
displaced families began to significantly decrease their total expenditures by $3,000-$6,000, and
in particular, this lagged fall in their total expenditures were mainly due to the sluggish

22

For t-2, the displaced families’ expenditures for transportation and health/education of displaced families
increased rather than decreased point estimate-wise. This pattern suggests that the main purpose of the earlier
decrease in their food expenditures might be to adjust the expenditure share across multiple expenditure categories
in response to the increased risk of future income loss.

24

downswing in their expenditures for housing, health care, and education during the same time.
Specifically, at t+2, displaced families’ total expenditures decreased by $3,000, and the decline
in their sub-expenditures for housing and health care/ education mostly accounted for the
reduction in total expenditures. For the subsequent periods, t+4 and t+6 plus, displaced families
spent $6,000 less on total expenditures, on average, and their sub-expenditures for housing, and
health care/education accounted for 50-30% and 30-50% of the total decrease in their total
expenditures, respectively. It is notable that only 15% of the entire decline in their total
expenditures from t+2 through t+6 plus can be attributable to the fall in their food expenditures,
which a number of previous studies have relied on heavily to proxy people’s consumption
behavior.
These dollar amounts of declines in total expenditures of displaced families shown in column
(1) in panel B of Table 1.B3 can be further compared to their family income losses in dollar
terms for each corresponding period previously shown in panel B of Table 1.B2. The comparison
illustrates a large gap between the two: On the one hand, for example, displaced families had had
$30,000 lower total income (before tax) annually, on average, over roughly eight years for the
post-displacement period (i.e., from t through t+6 plus). 23 This amount nearly equals the loss in
their disposable family income of $24,000. 24 On the other hand, they decreased their annual total
expenditures only by $4,000, on average, for the same post-displacement period, based on
column (1) in panel B of Table 1.B3. Therefore, the average gap between their disposable
income loss and total expenditures decline appeared to be about $20,000 in an accounting sense.

23

In this study, 8.7 years are the conditional sample mean of the maximum number of years that passed since the
husband’s first displacement, among the husbands having been first displaced roughly six or more years ago.
24
The average tax rate for the sample of this study is 20 percent based on NBER TAXSIM9, which is a tax
calculating program provided by NBER (http://users.nber.org/~taxsim/taxsim-calc9/index.html),

25

This gap can be explained either by a fall in the other sub-expenditures which are not
included in the measure of total family expenditures of this study, or by their self-funding
activities through reducing the net value of their own family wealth. While the former factor
cannot be analyzed because of the lack of other expenditure information in the 1999-2013 PSID,
the latter factor, self-funding behaviors of displaced families, can be explored by using extensive
family wealth data in the same data set.
Table 1.B4 shows in detail how a husband’s job displacement affected various components
of his family wealth from t-2 through t+6 plus. As in the previous analyses for the
income/expenditures effects of job displacement, all the interpretations of the coefficients for job
displacement dummies will be based on counter-factual wealth level which a family would have
without its husband’s displacement. But, for brevity, I will also frequently omit to explicitly
mention the comparison criterion when describing the following results.
Column (1) in Table 1.B4 first shows the evolution of total family wealth level of displaced
families in 2013 dollar terms. The displaced families appeared to considerably reduce their own
wealth for their expenditures smoothing: Their total family wealth declined by $172,000 by the
end of t+6 plus, on average, compared to what they would have without displacement, and this
amount of reduction accounts for 45% of the average total family wealth of families in this
sample. More specifically, they decreased the net value of their total family wealth by $11,000
annually, on average, from the end of t-2 through the end of t+6 plus. 25 This average annual
decline in their total family wealth roughly covers 55% of the unexplained gap of $20,000
between their disposable income loss and expenditures fall from t through t+6 plus, as
previously computed based on the results in Table 1.B2 and Table 1.B3.

25

The difference in family wealth effects of job displacement between t-2 and t+6 plus based on column (1) in
Table 1.B4 is -$113,921, whose standard error is 45,538.

26

Moreover, the decreasing pattern in their total family wealth is consistent with the sluggish
fall in their total family expenditures discussed in Table 1.B3; displaced families decreased their
total family wealth more intensively during the initial periods before the end of t, than they did
during the latter periods after the end of t. Specifically, they had already reduced their total
wealth by $58,000 at the end of t-2, and further decreased it by $29,000 by the end of t point
estimate-wise, and thus their total family wealth was $87,000 less at the end of t, on average,
compared to what they would have without displacement. Hence, of the total average reduction
in their total family wealth of $172,000 by the end of t+6 plus, roughly 50% took place by the
end of t.
Notably, this early decrease in their total family wealth appears to have covered most of their
family income loss which they had during the same period. Between the ends of t-2 and t, for
instance, they decreased their total wealth by $14,500 per year on average, point estimate-wise,
according to column (1) in Table 1.B4, and this decline roughly equals the loss in their
disposable annual total income of $15,000 at t, which is calculated based on the results in
column (3) in panel B of Table 1.B2. 26 This observation suggests that not only the expenditures
included in this study, but also other expenditures not covered by this study were likely reduced
only slightly by displaced families during the initial period around their husband’s job
displacement occurrence.
In addition, the speed of the decline in total family wealth of displaced families decelerated
for the subsequent periods after t, and this slowdown also corroborates the finding of their
dilatory reduction in total expenditures. Specifically, for about eight years, between the ends of t

26

The difference in family wealth effects of job displacement between t-2 and t based on column (1) in Table
1.B4 is -$28,716, whose standard error is 23,114.

27

and t+6 plus, they decreased their total family wealth by $10,000 per year, on average. 27 This
amount covers only 40% of the average annual loss of $27,000 in the total disposable family
income that they had had since t+2. 28 Inversely, this computation implies that the other 60% of
the disposable income loss of displaced families was offset, in an accounting sense, through a
decrease in their expenditures since t+2. This conjecture is in keeping with the previous finding
that displaced families decreased their total expenditures relatively slowly; a significant fall in
their total expenditures had begun to appear in t+2. 29
Columns (2) through (8) in Table 1.B4 further present the individual regression results for
each sub-component of total family wealth, such as the net value of housing, vehicles, cash in
hand (checking/saving), stocks, business/farm, individual retirement account (IRA), and other
assets, in 2013 dollar terms. 30 As in panel B of Table 1.B3, all the coefficients in column (1) of
Table 1.B4 are the mere summation of the corresponding ones in column (2) through (8) in the
same table. Hence, we can compare each sub-component’s contribution to the decrease in total
family wealth for each period.
As for the early decrease in total family wealth by the end of t, for example, its decrease in
displaced families was largely driven by falls in the (net) values of their home equity, stocks, and
family business/farm. In particular, the early reductions in the net values of their stocks and
business/farm were so large that the average dollar amounts of both declines by the end of t were
up to 50- 60% of each corresponding mean dollar value of the entire sample in this study. In turn,
27

The difference in family wealth effects of job displacement between t and t+6 plus based on column (1) in
Table 1.B4 is -$85,205, whose standard error is 36,103.
28
It is calculated based on the estimates in panel B of Table 1.B2
29
This further implies that the total expenditures of this study related to food, housing, transportation, health care
and education have relatively low income elasticity compared to the other expenditures that are not covered by this
study: as seen earlier, displaced families decreased their expenditures for food, housing, transportation, health care
and education only by $5,000 per year, on average, from t+2 through t+6 plus. This accounts for less than 30% of
$17,000, which is the average difference between disposable income loss and reduction in their family wealth.
30
The vehicles include cars, trucks, a motor home, a trailer, or a boat, while the other assets include the net
values of other estate and non-collateralized debts

28

the decrease in the net values of their vehicle and checking/saving accounts was relatively
moderate by the end of t, while the net values of their IRA appeared to even increase during the
same time, although the estimate for the IRA increase shows relatively weak statistical
significance.
These decreasing patterns across the subcomponents of family wealth, however, had changed
since the end of t. On the one hand, specifically, the further decrease in total family wealth of
displaced families was largely driven by the additional fall in the net values of their
checking/saving accounts, stocks and IRA taking place since the end of t. In particular, the net
values of their checking/saving accounts and IRA decreased so prominently from the end of t
through the end of t+6 plus relative to the preceding period, that their contributions to the entire
decrease in their total family wealth at the end of t+6 plus became much larger. On the other
hand, in contrast, there was no further significant decrease in the (net) values of their home
equity and business/farm from the end of t through the end of t+6 plus, although both of them
played a central role in the early decrease in their total family wealth by the end of t. As a result,
the contribution of the decline in the (net) values of their home equity and business/farm to the
decrease in their total family wealth became much smaller at the end of t+6 plus than the earlier
period.
The relatively early decline by the end of t in the (net) values of home equity, stocks, and
business/farm of displaced families can be viewed as their early downsizing of the share of
relatively illiquid/risky assets in response to their increased future income risk, by putting less
money into them than they would have done without the displacement. However, given that their
housing expenditures continued to decrease over time as seen in column (3) of Table 1.B3, the
early decrease in their home equity and its subsequent flat path through t+6 plus appears

29

somewhat puzzling. Through additional analyses on the change in their home price, which is the
summation of home equity and debt for home, and the probability of their moving to a lower/higher priced home before and after the displacement, I find that their home price began to
decline at t-2 as their home equity did, but also continued to decrease over time through t+6 plus,
as opposed to the non-decreasing trend in their home equity during the same time. 31 Moreover,
such patterns in their home price were partly due to a lower likelihood of a move to higherpriced homes by t, and a higher likelihood of a move to lower-priced homes after t through t+6
plus, relative to their counterparts who had not experienced job displacement. 32 These findings
partly explain the seemingly inconsistent pattern of changes in home equity and housing
expenditures of displaced families: The displaced families became more likely to move to lowerpriced homes with lower-home debt over time after displacement, than they would have done
without displacement. As a result, their average home price kept going down over time, which
made the fall in their average housing expenditures (mainly the mortgage payments) more salient
as time lapsed, whereas their home equity did not further decrease because their home debt also
decreased commensurate with the drop in home price.

1.6 Summary and Discussion
The results of this study translate into two main findings: First, the households with displaced
husbands practiced considerable expenditure smoothing mainly by reducing a large share of their
own wealth during the period 1998-2012. Second, the total expenditures of displaced families
fell sluggishly, although there began to be a significant fall in total family income much earlier,
31

Refer to Table 1.C4 for detailed regression results.

32

Refer to Table 1.C5-Table 1.C8 for detailed regression results.

30

even before the occurrence of displacement. In this section, I briefly review the main results
related to each finding, and then further discuss their respective implications.

1.6.1 Displaced Families’ Expenditure Smoothing by Reducing Their Own
Wealth
This study shows that during the period 1998-2012, a displaced family had18% less total
annual income, on average, over six or more years after a husband’s displacement (i.e., from t to
t+6 plus), relative to what it would have without a husband’s displacement. For the same time,
despite the substantial loss in family income, the displaced family decreased its annual total
expenditures only by 7%, on average, which total expenditures encompass a broad set of major
expenditure categories related to food, housing, transportation, health care, and education. As a
result of this large expenditure smoothing, displaced families ended up with $172,000 less family
wealth, on average, at the end of t+6 plus, which accounts for 45% of the average net value of
family wealth in the whole sample of this study. Particularly for the post-displacement period
from the end of t through the end of t+6 plus, for example, families decreased their total wealth
by $11,000 annually, on average, and this amount roughly offset 55% of the loss of its
disposable annual income during the same time.
These findings of this study are consistent, overall, with Stephens (2001), in that a relatively
moderate fall in expenditures of displaced families was observed. Specifically, Stephens (2001),
by using the 1968-1992 PSID, finds a moderate fall in food expenditures over a long period for
displaced families, compared to the large income loss they persistently had. The author views the
finding as evidence of consumption smoothing of displaced families. Along the same line, the
current study reexamines the long-term expenditure behaviors of displaced families by using

31

richer information about their family expenditures, and the results largely confirm Stephens'
(2001) argument: even with a broad set of expenditures categories including housing,
transportation, health care, and education, as well as food, displaced families turned out to
practice considerable expenditure smoothing during the post-displacement period. This finding
of the current study again calls attention to the importance of investigating the expenditure
effects of job displacement for assessing the true welfare loss caused by job displacement. As
Stephens (2001) argues, simply referring to the income loss induced by displacement is likely to
overestimate the actual long-run impact of job displacement on people’s welfare.
Through further analyses on wealth change of displaced families, however, this study
highlights the fact that such a large expenditure smoothing of displaced families was enabled at
the cost of a large share of their family wealth that they would have without displacement. This
finding invites consideration of two scenarios where a husband’s job displacement could
severely damage a family’s welfare: first, a family with a low level of wealth, thus presumably
also having a relatively tight borrowing constraint, would have to decrease their family
expenditures more drastically; and, second, even a family with a relatively high level of wealth
likely becomes more financially vulnerable over time once it experiences a husband’s job
displacement, because a displaced family tends to have an early large depletion of, and gradual
subsequent declines in its family wealth, as seen in Table 1.B4.
This study further suggests that the existing public assistance program did not play a
substantial role in mitigating the two problems described above in Table 1.B2. In the Table 1.C9,
an extended version of panel B in Table 1.B2, which presents the level job displacement effects
for each sub-component of family income, it is observed that there was statistically significant
increases in public transfers in dollar terms for displaced families, on average, at t, t+2, and t+4,

32

but the magnitudes of the increases were trivial. In addition, a large part of the increase was
concentrated on the initial periods around the displacement occurrence at t, and thus displaced
families were likely to have less public transfer over time after displacement.

1.6.2 The Sluggish Fall in Displaced Families’ Expenditures
Another noteworthy finding of this paper is that the decrease in displaced families’
expenditures became more salient over time. In fact, they did not decrease the total family
expenditures significantly during the initial period around displacement, although their total
income was significantly lowered by 12%, during the same time. A significant fall in total
expenditures of 5% began to appear roughly two years after displacement (t+2), and the total
expenditures decreased further point estimate-wise for the subsequent years, resulting in 9% less
annual total expenditures from t+4 through t+6 plus. I further show that this lagged fall would
appear in general even with a broader measure of family total expenditures than that of the
current study, by tracking their decreasing patterns of total family wealth and comparing it with
their concurrent income loss in an accounting sense. 33
This sluggish fall in displaced families’ total expenditures is a novel finding, which is not
found in their food expenditure behavior alone, as shown in both the current study and Stephens
(2001) (see Figure 1.A1). Furthermore, the interpretation of the lagged fall in total expenditures
of displaced families seems more complex than that of the flatter trend in their food expenditures
with an early significant fall, under classical assumptions of standard life cycle
theories/permanent income hypothesis. Why, for example, did the displaced families not begin to
reduce their total expenditures earlier so as to make their consumption path more even, even
33

Specifically, they appeared to decrease their wealth level enough to roughly cover the entire income loss they
had by t, whereas the decrease in their wealth for subsequent periods only covered 30% of the income loss that they
had during the same time.

33

though a large number of them appeared to expect future income loss and prepare for it in
advance by reducing some part of their expenditures and wealth even before displacement?
There are at least two possible explanations for this lagged fall that would answer this
question. First, the lagged decrease in their total expenditures could be attributable to the
difference in adjustment costs across sub-expenditure categories of the total expenditures. For
example, the food expenditures can be adjusted relatively easily while the expenditures for
housing generally require more time and money costs for adjustment. Health care and education
expenses are also relatively difficult to adjust if some family members are getting ongoing
treatments or already enrolled in schools. This different adjustment costs can lead to a large
difference in the timing of change across food expenditures and the other sub-expenditures
shown in Table 1.B3. Thus, no matter how precisely the displaced families expect the future
income loss due to their heads’ displacement, they might optimally choose to postpone a
decrease in expenditures for some categories with high adjustment costs in order to minimize it,
which lead to a lagged decrease in their total expenditures.
The second possible explanation is that their expectation of persistent (or permanent) income
loss due to job displacement might change systematically over time. In other words, they might
underestimate the persistent effect of job displacement on their earnings at initial stages, on
average (from an ex post perspective), yet continue to update their expectations as time lapses,
which leads to a further decrease in their total expenditures.
There are a couple of factors which can cause such systematic expectation errors on the
income effect of job displacement. First, a large economic shock could change people’s
expectations. The sample years of this study, 1998-2012, include two recession periods: 20022003 and 2008-2009, the latter a particularly severe recession that has been termed as the Great

34

Recession (Elsby et al. 2011). Therefore, the families in the main sample of this study were
likely to be affected by such a large economic shock and might, as a result, change their future
expectations.
Second, undergoing typical processes of persistent earnings losses after the first job
displacement may necessarily entail people’s expectation adjustments. For example, Stevens
(1997) finds that multiple job losses following the first job loss account for much of the
persistent reductions in earnings due to job displacement. The experience of such subsequent job
losses would lower people’s expectations of their future income paths gradually, which, in turn,
would lead to gradual falls in their expenditures over time.
Table 1.C10 shows some evidence that both factors above, the Great Recession shock and
multiple job losses, played a role in generating the lagged decrease in total family expenditures
of displaced heads. Specifically, column (1) in Table 1.C10 refers to the effect of displacement
on total family expenditures which is the same as column (1) in Table 1.B3. Columns (2) and (3)
in Table 1.C10 show how the Great Recession and subsequent job displacements, respectively,
affected their total expenditures during the post-first-displacement period. Overall, both factors
appeared to affect the lagged reduction in their total expenditures point estimate-wise; in
particular, the Great Recession made a relatively significant contribution to the decrease in their
total expenditures at t+4, while the subsequent job losses of their heads appeared to play a
statistically significant role in further decreasing their total expenditures, in the current year of or
a half year before the occurrence of the subsequent job losses. When I include all the interaction
terms for the Great Recession and the heads’ subsequent job losses in my original model
(column (4)), it can be observed that the lagged decrease in their total expenditures due to heads’
first displacement becomes smaller in terms of both point estimate and statistical significance.

35

This suggests that displaced families might change their future expectation of life family income
while experiencing unexpected additional shocks after their heads’ first displacement.

1.7 Conclusion
During the period 1998-2012, families with a displaced husband practiced substantial
expenditure smoothing over a long period from roughly two years before through six or more
years after displacement, notwithstanding the sizable persistent income loss that they
experienced over the same period of time. In particular, their expenditures decreased only
slightly in the initial periods around the year of displacement, but the decline in their
expenditures became more salient as time went on. It is noteworthy that this large expenditure
smoothing was enabled mainly by a considerable decrease in savings/assets of displaced families.
For about six or more years since the year of displacement, the average annual decrease in their
wealth covers roughly 55% of the average annual loss in their disposable income. In turn, the
average decline in their annual expenditures for food, housing, transportation, health care and
education explains about 20% of the average annual loss in their disposable income during the
same time.

36

APPENDICES

37

APPENDIX A

FIGURES FOR CHAPTER 1

38

Figure 1.A1 The Effects of Husband’s Job Displacement on Annual Earnings, Family Income,
and Expenditures
B. PSID 1968-1992 (Stephens 2001)

A. PSID 1999-2013
0

0

-10%

-10%

-20%

-20%

-30%

-30%

-40%

-40%
t-2

t

t+2

t+4

t-2

t+6

t-1

t

t+1

Total Expenditures

Food Expenditures

Food Expenditures

Husband's Earnings

Total Family Income

Total Family Income

39

t+2

t+3

t+4

t+5

Husband's Earnings

t+6

Figure 1.A2 The Effects of Husband's Job Displacement on Family Expenditures in Dollar
Terms
4,000
2,000

2013$

0
-2,000
-4,000
-6,000
-8,000
-10,000
-12,000
t-2

t

t+2

t+4

t+6

2013$

A. Total Expenditures

4,000

4,000

2,000

2,000

0

0

-2,000

-2,000

-4,000

-4,000

-6,000

-6,000

-8,000

-8,000

-10,000

-10,000

-12,000

-12,000
t-2

t

t+2

t+4

t+6

t-2

2013$

B. Food Expenditures

4,000

2,000

2,000

0

0

-2,000

-2,000

-4,000

-4,000

-6,000

-6,000

-8,000

-8,000

-10,000

-10,000

-12,000

-12,000
t

t+2

t+4

t+2

t+4

t+6

C. Housing Expenditures

4,000

t-2

t

t+6

t-2

D. Transportation Expenditures

t

t+2

t+4

E. Health and Edu. Expenditures

40

t+6

APPENDIX B

TABLES FOR CHAPTER 1

41

Table 1.B1 Descriptive Statistics for Non-displaced and Displaced Families (PSID 1999-2013)
Variables
Head’s Age
Head’s Education (Years)
Wife’s Age
Wife’s Education (Years)
Percentage of White Heads
Number of Children (0-17)
Number of Young Children (0-5)
Percentage of Manufacturing
Industry
Percentage of Blue-Collar
Workers
Percentage of Jobs Covered by
Union Contract
Percentage of Union Members
Family Annual Income (2013$)
Head’s Earnings
Wife's Earnings
Public Transfer
Private Transfer
Other Income
Total Annual Expenditure (2013$)
Food
Housing
Transportation
Health Care
Education

Non-displaced 1999-2013
Mean(SD)
N
42.74
8,561
(10.04)
14.25
8,470
(2.70)
41.12
8,561
(9.87)
14.39
8,473
(2.50)
0.86
8,561
(0.35)
1.23
8,561
(1.20)
0.42
8,561
(0.71)
0.19
8,561
(0.39)
0.31
8,561
(0.46)
0.14
8,561
(0.34)
0.12
8,561
(0.33)
133,159
8,561
(180,928)
86,846
8,561
(155,902)
32,131
8,561
(36,682)
983
8,561
(3,973)
3,000
8,561
(16,454)
10,200
8,561
(42,905)
56,683
8,561
(31,025)
10,482
8,561
(5,269)
21,723
8,561
(16,699)
16,234
8,561
(12,746)
3,861
8,561
(4,900)
4,383
8,561
(9,842)

42

Displaced 1999-2013
Mean(SD)
N
44.07
1,363
(9.77)
14.03
1,355
(2.64)
42.57
1,363
(9.55)
14.04
1,338
(2.76)
0.80
1,363
(0.40)
1.33
1,363
(1.27)
0.40
1,363
(0.68)
0.20
1,363
(0.40)
0.41
1,363
(0.49)
0.08
1,363
(0.26)
0.07
1,363
(0.25)
103,650
1,363
(75,160)
64,291
1,363
(60,857)
28,110
1,363
(31,661)
3,022
1,363
(6,794)
2,224
1,363
(10,044)
6,002
1,363
(17,176)
52,358
1,363
(26,299)
10,016
1,363
(4,614)
19,843
1,363
(12,901)
14,684
1,363
(11,088)
3,929
1,363
(4,009)
3,886
1,363
(9,210)

Table 1.B1 (cont’d)
Variables
Total Family Wealth (2013$)
Home Equity
Total Vehicle(s) Value
Checking/Saving
Stock
Business/Farm
IRA
Other Wealth

Non-displaced 1999-2013
Mean(SD)
N
396,807
8,561
(1,004,994)
123,594
8,561
(190,764)
22,393
8,561
(27,246)
32,764
8,561
(109,175)
55,685
8,561
(311,796)
66,722
8,561
(558,822)
55,198
8,561
(160,614)
40,452
8,561
(277,948)

43

Displaced 1999-2013
Mean(SD)
N
262,111
1,363
(507,842)
99,337
1,363
(181,102)
16,204
1,363
(17,150)
20,649
1,363
(61,321)
33,396
1,363
(130,543)
15,873
1,363
(153,372)
50,089
1,363
(152,592)
26,563
1,363
(170,223)

Table 1.B2 The Effects of Husband’s Job Displacement on Family Income (PSID 1999-2013)
(1)
Husband's
Earnings

(2)
Husband's and
Wife's Earnings

(3)
Total
Family Income

2 - 2.5 years before
(t-2)

-0.122**
(0.051)

A. Log Dependent Variable
-0.071**
(0.031)

-0.083***
(0.023)

Current year 0.5 year before (t)

-0.228***
(0.056)

-0.180***
(0.044)

-0.132***
(0.033)

1.5 - 2 years after
(t+2)

-0.357***
(0.066)

-0.256***
(0.043)

-0.229***
(0.034)

3.5 - 4 years after
(t+4)

-0.295***
(0.074)

-0.214***
(0.053)

-0.196***
(0.044)

5.5+ years after
(t+6 plus)

-0.232***
(0.073)

-0.188***
(0.053)

-0.182***
(0.044)

9,924
2,245

9,924
2,245

9,924
2,245

2 - 2.5 years before
(t-2)

-13,340***
(3,962)

B. Level Dependent Variable (2013$)
-13,929***
(4,149)

-16,959***
(4,886)

Current year 0.5 year before (t)

-17,385***
(6,024)

-19,080***
(6,287)

-18,197***
(6,931)

1.5 - 2 years after
(t+2)

-29,475***
(5,786)

-32,502***
(6,103)

-35,198***
(6,630)

3.5 - 4 years after
(t+4)

-27,761***
(6,885)

-30,753***
(7,587)

-32,872***
(7,967)

5.5+ years after
(t+6 plus)

-28,863***
(9,591)

-33,444***
(10,249)

-35,105***
(10,561)

83,777
(146,838)

115,361
(152,815)

129,144
(170,737)

9,924
2,245

9,924
2,245

9,924
2,245

Job Displacement

Observations
Number of Families

Mean
(SD)
Observations
Number of Families

Note: Robust standard errors are presented in parentheses (*** p<0.01, ** p<0.05, * p<0.1).

44

Table 1.B3 The Effects of Husband’s Job Displacement on Family Expenditures (PSID 19992013)
(1)

(2)

(3)

(4)

(5)

Total
Expenditure

Food

Housing

Transportation

Health+Edu

Job Displacement
2 - 2.5 years before
(t-2)

0.014
(0.022)

A. Log Dependent Variable
-0.049*
0.020
0.041
(0.029)
(0.032)
(0.052)

Current year 0.5 year before (t)

-0.023
(0.028)

-0.075**
(0.030)

-0.006
(0.036)

0.019
(0.055)

-0.019
(0.117)

1.5 - 2 years after
(t+2)

-0.049*
(0.028)

-0.051
(0.035)

-0.061
(0.038)

0.064
(0.057)

-0.140
(0.117)

3.5 - 4 years after
(t+4)

-0.087**
(0.034)

-0.037
(0.034)

-0.063
(0.044)

-0.062
(0.067)

-0.105
(0.138)

5.5+ years after
(t+6 plus)

-0.089**
(0.036)

-0.086**
(0.036)

-0.005
(0.055)

-0.033
(0.069)

-0.306**
(0.144)

9,924
2,245

9,922
2,245

9,825
2,242

9,817
2,245

9,687
2,232
767
(827)

Observations
Number of Families

0.141
(0.090)

2 - 2.5 years before
(t-2)

378
(1,299)

B. Level Dependent Variable (2013$)
-776***
-303
690
(263)
(727)
(751)

Current year 0.5 year before (t)

-1,622
(1,593)

-943***
(269)

-892
(751)

470
(719)

-257
(1,068)

1.5 - 2 years after
(t+2)

-3,033*
(1,637)

-583*
(324)

-1,797**
(727)

788
(901)

-1,441
(992)

3.5 - 4 years after
(t+4)

-5,571***
(2,046)

-590*
(347)

-2,538***
(823)

-645
(1,005)

-1,798
(1,253)

5.5+ years after
(t+6 plus)

-5,984**
(2,482)

-1,057***
(390)

-1,691
(1,085)

-225
(1,002)

-3,011*
(1,677)

Mean
(SD)

56,095
(30,460)

10,419
(5,187)

21,467
(16,247)

16,023
(12,544)

8,186
(11,346)

9,924
2,245

9,924
2,245

9,924
2,245

9,924
2,245

9,924
2,245

Observations
Number of Families

Note: Robust standard errors are presented in parentheses (*** p<0.01, ** p<0.05, * p<0.1).

45

Table 1.B4 The Effects of Husband’s Job Displacement on Family Wealth in 2013$ Terms
(PSID 1999-2013)
(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Total Family
Wealth

Home
Equity

Total
Vehicle(s)
Value

Checking
/Saving

Stocks

Business/Farm

IRA

Other

2 - 2.5 years
before (t-2)

-58,218**
(27,639)

-19,198**
(7,601)

1
(1,495)

-11,168***
(3,881)

-7,024
(14,748)

-32,225**
(13,765)

13,709
(10,947)

-2,314
(12,918)

Current year 0.5 year before (t)

-86,934***
(27,117)

-21,497**
(10,181)

-2,186*
(1,252)

-9,083*
(4,955)

-25,988**
(13,132)

-34,659***
(12,620)

18,914
(12,097)

-12,435
(10,879)

1.5 - 2 years after
(t+2)

-121,837***
(31,696)

-15,720
(10,616)

-1,574
(1,417)

-18,976***
(4,620)

-22,302
(13,995)

-36,480**
(16,612)

-6,832
(9,673)

-19,952
(12,211)

3.5 - 4 years after
(t+4)

-118,482***
(37,534)

-16,283
(10,752)

-4,918***
(1,633)

-23,147***
(5,699)

-29,680*
(15,947)

-35,764**
(18,217)

6,458
(16,115)

-15,148
(13,429)

5.5+ years after
(t+6 plus)

-172,139***
(48,725)

-20,759
(13,398)

-5,467**
(2,136)

-28,830***
(7,350)

-49,890**
(21,570)

-39,312
(25,502)

-12,350
(15,368)

-15,530
(17,748)

378,480
(953,822)

120,294
(189,652)

21,551
(26,188)

31,116
(104,047)

52,652
(293,877)

59,803
(522,772)

54,503
(159,548)

38,562
(265,904)

9,924
2,245

9,924
2,245

9,924
2,245

9,924
2,245

9,924
2,245

9,924
2,245

9,924
2,245

9,924
2,245

Job Displacement

Mean
(SD)

Observations
Number of Families

Note: Robust standard errors are presented in parentheses (*** p<0.01, ** p<0.05, * p<0.1).

46

APPENDIX C

SUPPLEMENTAL TABLES FOR CHAPTER 1

47

Table 1.C1 Summary Statistics for Family Expenditures (PSID 1999-2013)
Annual Expenditure Variables
(2013$)
Food
Food-at-home
Food-away-from-home
Food Stamps
Housing
Rent
Mortgage Payments
Home Utility
Electricity
Heating
Home Utility
Home Insurance
Property Tax
Housing (w/ imputed
subcomponents)
Home Insurance w/ imputation
Property Tax w/ imputation
Transportation
Total Car Expenditure
For Buying Car#1
For Buying Car#2
For Buying Car#3
For Leasing Car#1
For Leasing Car#2
For Leasing Car#3

Mean(SD)
10,419
(5,187)
7,398
(3,793)
2,906
(2,875)
115
(819)
21,502
(16,596)
1,757
(5,146)
13,353
(14,270)
3,178
(2,099)
1,513
(1,100)
1,136
(1,210)
2,648
(1,594)
825
(891)
2,905
(3,472)
21,467
(16,247)
831
(865)
2,878
(3,419)
15,187
(11,926)
14,856
(11,691)
3,476
(6,230)
1,080
(2,951)
212
(1,167)
400
(1,766)
111
(942)
11
(311)

PSID 1999-2013
# of Cond. Mean for Non-zeros
Zeros
(SD)

N
9,924

2

9,924

44

9,924

414

9,924

9,583

8,945

99

9,924

8,268

9,924

2,785

9,603

198

9,924

895

9,924

2,193

9,924

203

9,119

1,944

9,472

1,893

9,924

99

9,924

1,944

9,924

1,928

9,396

107

9,396

201

9,924

5,163

9,924

7,667

9,924

9,182

9,924

9,391

9,924

9,777

9,924

9,911

48

10,420
(5,185)
7,424
(3,773)
3,037
(2,871)
3,970
(2,823)
21,706
(16,541)
12,027
(7,599)
18,419
(13,695)
3,240
(2,072)
1,652
(1,044)
1,440
(1,191)
2,699
(1,566)
1,016
(885)
3,518
(3,528)
21,651
(16,194)
1,005
(854)
3,468
(3,470)
15,368
(11,880)
15,182
(11,608)
7,328
(7,321)
4,964
(4,556)
2,829
(3,286)
7,060
(2,832)
7,064
(2,742)
7,226
(3,268)

# of
Missing
0
0
0
0
979
0
0
321
0
0
0
805
452
0
0
0
528
528
0
0
0
0
0
0

Table 1.C1 (cont’d)
Annual Expenditure Variables
(2013$)
For Additional Car(s)
Car Operation
Car Insurance
Gasoline/Fuel
Car Repair
Car Parking/Pooling
Public Transportation
Bus/Train
Taxi
Other
Transportation
(w/ imputed car-insurance)
Car Expenses (w/ imputed
car-insurance)
Car Operation (w/ imputed
car-insurance
Car Insurance w/
imputation
Education
Child Care
School
Health Care
Doctor/Surgery/Dental
Prescription/Drugs
Hospital/Nursing Home
Health Insurance
Total (Food, Housing,
Transportation,
Health, & Education)

Mean(SD)

PSID 1999-2013
# of Cond. Mean for Non-zeros
Zeros
(SD)

N

2,336
(4,341)
7,205
(6,545)
2,130
(2,805)
2,997
(2,398)
2,007
(4,752)
84
(531)
331
(2,242)
101
(566)
25
(213)
206
(2,112)
16,023
(12,544)
15,692
(12,276)
8,065
(7,323)
2,977
(4,302)
4,315
(9,760)
1,112
(3,453)
3,203
(9,255)
3,870
(4,789)
809
(1,370)
376
(765)
370
(1,519)
2,316
(3,716)

9,924

6,374

9,396

203

9,396

268

9,924

247

9,924

4,990

9,924

9,017

9,924

8,851

9,924

9,276

9,924

9,636

9,924

9,550

9,924

107

9,924

201

9,924

203

9,924

287

9,924

4,258

9,924

7,555

9,924

5,532

9,924

406

9,924

1,463

9,924

1,427

9,924

6,512

9,924

2,321

56,095
(30,460)

9,924

0

49

6,666
(4,990)
7,365
(6,528)
2,190
(2,821)
3,074
(2,380)
4,064
(6,112)
876
(1,494)
2,864
(6,020)
1,406
(1,627)
824
(924)
5,129
(9,280)
16,204
(12,498)
16,018
(12,190)
8,234
(7,305)
3,061
(4,333)
7,636
(11,967)
5,085
(5,858)
7,109
(12,741)
4,026
(4,819)
938
(1,434)
436
(807)
1,120
(2,480)
3,018
(3,985)
56,095
(30,460)

# of
Missing
0
528
528
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0

0

Table 1.C2 The Number of the Job Displacement Dummies (PSID 1999-2013)
Survey Year
1999

2001

2003

2005

2007

2009

2011

2013

Total

1st Job Displacement Dummy
2 - 2.5 years before (t-2)
Current year - 0.5 year before (t)

24
22

35
31

25
39

45
30

58
48

56
87

43
66

0
44

286
367

1.5 - 2 years after (t+2)

29

22

27

36

27

41

80

61

323

3.5 - 4 years after (t+4)

0

24

21

23

32

30

43

73

246

5.5+ years after (t+6 plus)

0

0

19

32

53

82

107

134

427

Table 1.C3 The Effects of Husband’s Job Displacement on Family Food Expenditure (PSID
1968-1997)
Dependent Variables: Log Annual Food Expenditure
(1)
All years
Job Displacement Dummies

(2)
Oddnumbered
years

(3)
Oddnumbered
years

(4)
Evennumbered
years

(5)
Evennumbered
years

1-2 year(s) before
(2 years before for (2)-(5))

-0.046***
(0.014)

-0.069***
(0.024)

-0.057**
(0.024)

-0.037
(0.027)

-0.035
(0.027)

Current year or 1 year after
(Current year for (2)-(5))

-0.074***
(0.015)

-0.084***
(0.025)

-0.079***
(0.024)

-0.078***
(0.025)

-0.075***
(0.025)

2-3 years after
(2 years after for (2)-(5))

-0.078***
(0.016)

-0.090***
(0.026)

-0.085***
(0.026)

-0.114***
(0.030)

-0.112***
(0.029)

4-5 years after
(4 years after for (2)-(5))

-0.081***
(0.017)

-0.098***
(0.027)

-0.093***
(0.026)

-0.073**
(0.029)

-0.069**
(0.029)

6+ years after

-0.077***
(0.019)

-0.085***
(0.028)

-0.080***
(0.027)

-0.084***
(0.030)

-0.077***
(0.029)

Yes

No

No

Yes
Yes

No

No

Yes

Dropping all the heads displaced in evennumbered years
Including all the heads displaced in evennumbered years and assuming they’re not
displaced
Dropping all the heads displaced in oddnumbered years
Including all the heads displaced in oddnumbered years and assuming they’re not
displaced

Observations
39,595
16,339
19,928
15,927
Number of Families
4,028
3,403
4,003
3,210
Note: Robust standard errors are presented in parentheses (*** p<0.01, ** p<0.05, * p<0.1).

50

19,667
3,848

Table 1.C4 The Effects of Husband’s Job Displacement on Home Equity, Home Price, and
Housing Expenditures in 2013$ Terms
(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

Home
Equity

Home
Price

Housing
Expenditures

Rent

Mortgage
Payments

Electricity

Heating

Property
Tax

Home
Insurance

2 - 2.5 years
before (t-2)

-19,198**
(7,601)

-19,019**
(9,032)

-303
(727)

-116
(277)

-239
(746)

49
(61)

24
(76)

-36
(124)

15
(50)

Current year 0.5 year before (t)

-21,497**
(10,181)

-27,698**
(12,556)

-892
(751)

-165
(313)

-663
(751)

1
(67)

102
(159)

-184
(141)

17
(85)

1.5 - 2 years after
(t+2)

-15,720
(10,616)

-37,123***
(13,077)

-1,797**
(727)

232
(346)

-1,511**
(730)

32
(74)

-38
(96)

-408***
(141)

-103*
(59)

3.5 - 4 years after
(t+4)

-16,283
(10,752)

-44,247***
(13,092)

-2,538***
(823)

261
(396)

-2,211***
(850)

13
(92)

50
(109)

-529***
(166)

-121*
(73)

5.5+ years after
(t+6 plus)

-20,759
(13,398)

-49,490***
(17,303)

-1,691
(1,085)

580
(527)

-1,621
(1,017)

32
(96)

-4
(119)

-471*
(258)

-208**
(84)

Mean
(SD)

120,294
(189,652)

248,141
(261,710)

21,467
(16,247)

1,757
(5,146)

13,353
(14,270)

1,513
(1,100)

1,136
(1,210)

2,878
(3,419)

830.8
(864.9)

9,924
2,245

9,924
2,245

9,924
2,245

9,924
2,245

9,924
2,245

9,924
2,245

9,924
2,245

9,924
2,245

9,924
2,245

Job Displacement

Observations
Number of Families

Note: Robust standard errors are presented in parentheses (*** p<0.01, ** p<0.05, * p<0.1).

51

Table 1.C5 The Probability of A Displaced Family’s Living in a Different, Low-Priced House
Compared to t-4
(1)
1:Live in a different,
lower-priced house at
t-2, compared to t-4

(2)
1:Live in a different,
lower-priced house
at t, compared to t-4

(3)
1:Live in a different,
lower-priced house at
t+2, compared to t-4

(4)
1:Live in a different,
lower-priced house at
t+4, compared to t-4

(5)
1:Live in a different,
lower-priced house at
t+6, compared to t-4

-0.022
(0.018)

0.025
(0.029)

0.029
(0.042)

0.116**
(0.057)

0.207**
(0.086)

8,774
2,241

6,917
1,972

5,236
1,647

3,893
1,353

2,821
1,116

Job Displacement

Job Displacement
Occurred at t
Observations
Number of Families

Note 1: Table 1.C5 is based on Table 1.C6.
Note 2: Robust standard errors are presented in parentheses (*** p<0.01, ** p<0.05, * p<0.1).

Table 1.C6 The Probability of A Displaced Family’s Living a Different, Low-Priced House

Job Displacement

(1)
1:Live in a different,
lower-priced house
compared to 2 years
ago

(2)
1:Live in a different,
lower-priced house
compared to 4 years
ago

(3)
1:Live in a different,
lower-priced house
compared to 6 years
ago

(4)
1:Live in a different,
lower-priced house
compared to 8 years
ago

(5)
1:Live in a different,
lower-priced house
compared to 10 years
ago

2 - 2.5 years before
(t-2)
Current year 0.5 year before (t)
1.5 - 2 years after
(t+2)
3.5 - 4 years after
(t+4)
5.5+ years after
(t+6 plus)

-0.022
(0.018)
-0.008
(0.019)
0.010
(0.022)
0.008
(0.025)
-0.003
(0.025)

0.005
(0.028)
0.025
(0.029)
0.035
(0.036)
0.029
(0.038)
0.046
(0.045)

-0.004
(0.033)
0.034
(0.037)
0.029
(0.042)
0.060
(0.049)
0.056
(0.062)

0.049
(0.041)
0.069
(0.049)
0.125**
(0.055)
0.116**
(0.057)
0.114
(0.071)

0.110*
(0.059)
0.109*
(0.060)
0.162**
(0.074)
0.197**
(0.079)
0.207**
(0.086)

3,893
1,353

2,821
1,116

Observations
8,774
6,917
5,236
Number of Families
2,241
1,972
1,647
Note: Robust standard errors are presented in parentheses (*** p<0.01, ** p<0.05, * p<0.1).

52

Table 1.C7 The Probability of A Displaced Family’s Living a Different, Higher-Priced House,
Compared to t-4
(1)
1:Live in a different,
higher-priced house
at t-2, compared to
t-4

(2)
1:Live in a different,
higher-priced house
at t, compared to
t-4

(3)
1:Live in a different,
higher-priced house
at t+2, compared to
t-4

(4)
1:Live in a different,
higher-priced house
at t+4, compared to
t-4

(5)
1:Live in a different,
higher-priced house
at t+6, compared to
t-4

-0.007
(0.028)

-0.084**
(0.041)

-0.088
(0.062)

-0.124
(0.097)

-0.289**
(0.129)

8,774
2,241

6,917
1,972

5,236
1,647

3,893
1,353

2,821
1,116

Job Displacement

Job Displacement
Occurred at t
Observations
Number of Families

Note1: Table 1.C7 is based on Table 1.C8.
Note2: Robust standard errors are presented in parentheses (*** p<0.01, ** p<0.05, * p<0.1).

Table 1.C8 The Probability of A Displaced Family’s Living a Different, Higher-Priced House

Job Displacement
2 - 2.5 years before
(t-2)
Current year 0.5 year before (t)
1.5 - 2 years after
(t+2)
3.5 - 4 years after
(t+4)
5.5+ years after
(t+6 plus)

(1)
1:Live in a
different, higherpriced house
compared to 2
years ago

(2)
1:Live in a
different, higherpriced house
compared to 4
years ago

(3)
1:Live in a
different, higherpriced house
compared to 6
years ago

(4)
1:Live in a
different, higherpriced house
compared to 8
years ago

(5)
1:Live in a
different, higherpriced house
compared to 10
years ago

-0.007
(0.028)
-0.037
(0.023)
-0.021
(0.027)
-0.003
(0.035)
0.020
(0.031)

-0.052
(0.040)
-0.084**
(0.041)
-0.091**
(0.045)
-0.026
(0.055)
-0.019
(0.058)

-0.019
(0.047)
-0.126**
(0.056)
-0.088
(0.062)
-0.083
(0.077)
-0.052
(0.085)

-0.053
(0.067)
-0.131*
(0.073)
-0.190**
(0.085)
-0.124
(0.097)
-0.151
(0.109)

-0.072
(0.089)
-0.182*
(0.095)
-0.197*
(0.103)
-0.251**
(0.121)
-0.289**
(0.129)

3,893
1,353

2,821
1,116

Observations
8,774
6,917
5,236
Number of Families
2,241
1,972
1,647
Note: Robust standard errors are presented in parentheses (*** p<0.01, ** p<0.05, * p<0.1).

53

Table 1.C9 The Effects of Husband’s Job Displacement on Sub-components of Family Income in
2013$ Terms
(1)

(2)

(3)

(4)

(5)

(6)

Husband's
Earnings

Wife's
Earnings

Public
Transfer

Private
Transfer

Other
Family Income

Total
Family Income

2 - 2.5 years before
(t-2)

-13,340***
(3,962)

-590
(1,799)

-1
(414)

-707
(627)

-2,322
(1,592)

-16,959***
(4,886)

Current year 0.5 year before (t)

-17,385***
(6,024)

-1,695
(1,989)

2,099***
(574)

-535
(918)

-681
(1,691)

-18,197***
(6,931)

1.5 - 2 years after
(t+2)

-29,475***
(5,786)

-3,027
(1,929)

797
(518)

-593
(789)

-2,899*
(1,594)

-35,198***
(6,630)

3.5 - 4 years after
(t+4)

-27,761***
(6,885)

-2,992
(2,399)

1,066**
(540)

-935
(884)

-2,250
(1,694)

-32,872***
(7,967)

5.5+ years after
(t+6 plus)

-28,863***
(9,591)

-4,582
(2,909)

1,238**
(588)

-1,087
(1,071)

-1,812
(2,129)

-35,105***
(10,561)

83,777
(146,838)

31,584
(36,065)

1,260
(4,517)

2,894
(15,738)

9,629
(40,404)

129,144
(170,737)

9,924
2,245

9,924
2,245

9,924
2,245

9,924
2,245

9,924
2,245

9,924
2,245

Mean
(SD)

Observations
Number of Families

Note: Robust standard errors are presented in parentheses (*** p<0.01, ** p<0.05, * p<0.1).

54

Table 1.C10 The Effects of the Great Recession and Multiple Job Losses on Total Family
Expenditures of Displaced Husbands
(1)
Log Total
Expenditures

(2)
Log Total
Expenditures

(3)
Log Total
Expenditures

(4)
Log Total
Expenditures

0.014
(0.022)

0.013
(0.022)

0.012
(0.022)

0.012
(0.022)

Current year of 0.5 year
before 1st JD (t)

-0.023
(0.028)

-0.023
(0.028)

-0.025
(0.028)

-0.025
(0.028)

1.5 - 2 years after 1st JD
(t+2)

-0.049*
(0.028)

-0.061*
(0.031)

-0.043
(0.028)

-0.056*
(0.031)

3.5 - 4 years after 1st JD
(t+4)

-0.087**
(0.034)

-0.066*
(0.036)

-0.075**
(0.035)

-0.054
(0.037)

5.5+ years after 1st JD
(t+6 plus)

-0.089**
(0.036)

-0.086**
(0.038)

-0.067*
(0.038)

-0.066*
(0.040)

Dummy Variables
1st Job Displacement
2 - 2.5 years before 1st JD
(t-2)

The Great Recession (GR)
GR in 1.5 - 2 years after 1st JD

0.033
(0.037)

0.037
(0.038)

GR in 3.5 - 4 years after 1st JD

-0.071
(0.049)

-0.070
(0.048)

GR in 5.5 + years after 1st JD

-0.010
(0.025)

-0.005
(0.024)

Multiple Job Losses
Current year of - 0.5 year
before 2nd/3rd JD

-0.063**
(0.030)

-0.064**
(0.030)

1.5 - 2 years after 2nd/3rd JD

-0.046
(0.045)

-0.048
(0.044)

3.5 - 4 years after 2nd/3rd JD

-0.039
(0.038)

-0.040
(0.038)

5.5+ years after 2nd/3rd JD

-0.097*
(0.054)

-0.095*
(0.054)

9,924
2,245

9,924
2,245

Observations
9,924
9,924
Number of Families
2,245
2,245
Note: Robust standard errors are presented in parentheses (*** p<0.01, ** p<0.05, * p<0.1).

55

APPENDIX D

ADDITIONS FOR CHAPTER 1

56

Constructing the Variables for Annual Family Expenditures
Most of the questions related to annual family income in the PSID have referred to the
previous year of each survey year. But the time frames of questions related to expenditures are
varied across expenditure categories; for some expenditures such as for health care and education,
for example, the PSID has asked about the amount of expenditures during the two years prior to
the interview year, while for many other expenditures related to food, housing, etc, it has referred
the previous month or just an unspecified recent period around the interview month, which was
mostly between March and May for each interview year. There are also some expenditure
categories, such as for car-purchase and lease, which refer to the whole period from two years
before the interview year through the interview month. Thus, in general, the time frame of
annualized expenditures in the PSID would match that of the annual income, with some errors.
More specific methods that I use to construct annualized expenditure amounts for each
expenditure category are as follows:
Food Expenditures: The annual food expenditures in this study include food-at-home, food
delivery, food-away-from-home, and food stamps. Specifically, I use the following questions in
the 1999-2013 PSID to calculate annualized family expenditures for food-at-home, food delivery,
and food-away-from-home: “How much do you (and everyone else in your family) spend on
food that you use at home in an average week?”, “How much do you spend on that food
delivered to the door?”, “How much do you (and everyone else in your family) spend eating
out?”. As for food stamps, there are two different questions in the 1999-2013 PSID, “How much
did you receive in food stamp benefits in the previous year?” and “How much did you receive in
food stamp benefits last month?” It turns out both measures for food stamps generate comparable
regression results in this paper, although I choose to use the former one because my primary

57

purpose is to proxy the annual expenditures for the year prior to each interview year as much as
possible to more correctly match the time frame of annual income information in the PSID.
Housing Expenditures: The annual housing expenditures consist of rent/mortgage payments,
home utility expenses for electricity and heating, home insurance premiums, and property tax.
Specifically, annualized payments for rent and mortgage are calculated based on the following
questions, “About how much rent do you pay a month?” and “How much are your monthly
mortgage payments?”, respectively. Home utility expenditures for electricity and heating refer to
the questions, “How much do you (and your family living there) usually pay for electricity per
month on average?” and “how much do you (and your family living there) usually pay for gas or
other types of heating fuel per month on average?” Lastly, the home insurance premium was
calculated from the question “How much is your total yearly homeowner's insurance premium?”,
and property tax from the question “About how much are your total yearly property taxes,
including city, county, and school taxes?”
Transportation Expenditures: The annual transportation expenditures encompasses the
expenses for purchasing/leasing car(s), car-operation costs such as car insurance, car repair,
fueling, and parking/pulling, and the expenses for using public transportation such as buses,
trains, taxi.
Specifically, the 1999-2013 PSID has asked about detailed expenditures for up to three cars
which had been newly purchased or leased since the January two years before each interview
year. For example, the questions for newly-purchased car(s) cover total purchase price, trade-in
dollar amounts if any, and down/loan payments if any. Using those questions, I calculate average
annual expenses for each car newly purchased since January two years before the interview year.
Similarly, the questions for newly-leased car(s) cover the dollar amount of initial outlay and

58

lease payments; using those questions, I also calculate average annual expenses for each car
newly leased since January two years before the interview year. The 1999-2013 PSID also asked
about additional purchase/lease expenditures for additional cars which are not covered by the
questions above. Combining all this information, I proxy average annual dollar amounts that a
family has paid for buying/leasing car(s) for the last two years of the interview year.
I obtain the information on annual car insurance premium from the following question: “How
much do you (and your family living there) pay for car insurance for (all of) your vehicle(s) per
year?” In turn, to calculate other annualized car-operation costs such as for car
repair/gasoline/car parking and pooling, I refer to the following question: “In [LAST MONTH],
how much did (you/your family living there) pay for each of these transportation related
expenses [a. Car repairs or maintenance b. Gasoline, c. Parking and car pooling]?” Similarly, to
construct annualized public transportation costs for bus, train, taxi, etc, I refer to the following
question: “In [LAST MONTH], how much did (you/your family living there) pay for each of
these transportation related expenses [a. Bus fares and train fares b.Taxicabs c.Other
transportation costs]?”
Health Care Expenditures: The annual family expenditures for health care include the
expenses for doctor appointments, surgery, dental treatment, nursing care, hospitalization,
prescription medicine and health insurance. Each component of them is calculated based on the
following questions respectively: “About how much did you pay out-of-pocket for doctors,
outpatient surgery, dental bills in the last two years combined?”; “About how much did you pay
out-of-pocket for nursing home and hospital bills in the last two years combined?”; “About how
much did you pay out-of-pocket for prescriptions, in-home medical care, special facilities, and
other services in 1997 and 1998 combined?”; “Altogether, how much did [you/your family] pay

59

for health insurance premiums, in the last two years combined, for (all of) the health insurance or
health care coverage(s) you just mentioned? Please include amounts that you had automatically
deducted from your pay, as well as amounts you paid directly.”
Education Expenditures: The annual family expenditures for education consist of schoolrelated costs and child care costs. Each component of them refers to the following questions,
respectively: “In the last year, how much in total were these expenses such as [a. Purchase or
rental of books, supplies, uniforms, or equipment including computers and software; b. Tuition
or tutoring (not including any amounts already mentioned for day care or nursery school); c.
Room and board for a family member who is away at school]?”; “In the last year, were there any
other school-related expenses not already covered in the previous question?”; “How much did
you (and your family living there) pay for child care last year?”

60

REFERENCES

61

REFERENCES

Blundell, Richard, Luigi Pistaferri, and Ian Preston. 2008. “Consumption Inequality and Partial
Insurance.” The American Economic Review 98(5): 1887–1921.
Blundell, Richard, Luigi Pistaferri, and Itay Saporta-Eksten. 2016. “Consumption Inequality and
Family Labor Supply.” American Economic Review 106(2): 387–435.
Boisjoly, Johanne, Greg J. Duncan, and Timothy Smeeding. 1998. “The Shifting Incidence of
Involuntary Job Losses from 1968 to 1992.” Industrial Relations 37(2): 207–31.
Browning, Martin, and M. Dolores Collado. 2001. “The Response of Expenditures to
Anticipated Income Changes: Panel Data Estimates.” American Economic Review 91(3):
681–92.
Browning, Martin, and Thomas F. Crossley. 2001. “The Life-Cycle Model of Consumption and
Saving.” Journal of Economic Perspectives 15(3): 3–22.
Browning, Martin, and Thomas F. Crossley. 2008. “The Long-Run Cost of Job Loss as
Measured by Consumption Changes.” Journal of Econometrics 145(1–2): 109–20.
Couch, Kenneth A., and Dana W. Placzek. 2010. “Earnings Losses of Displaced Workers
Revisited.” American Economic Review 100(1): 572–89.
Deaton, Angus, and John Muellbauer. 1980. “An Almost Ideal Demand System.” The American
economic review 70(3): 312–26.
Dynarski, Susan, and Jonathan Gruber. 1997. “Can Families Smooth Variable Earnings ?”
Brookings Papers on Economic Activity 1997(1): 229–303.
Elsby, Michael W. L., Bart Hobijn, and Ayşegül Şahin. 2011. “The Labor Market in the Great
Recession—An Update to September 2011 [with Comment and Discussion].” Brookings
Papers on Economic Activity (2011): 353–84.
Gruber, Jonathan. 1997. “The Consumption Smoothing Benefits of Unemployment Insurance.”
American Economic Review 87(1): 192–205.
Hsieh, Chang-Tai. 2003. “Do Consumers React to Anticipated Income Changes? Evidence from
the Alaska Permanent Fund.” American Economic Review 93(1): 397–405.
Jacobson, Louis S., Robert J. Lalonde, and Daniel G. Sullivan. 1993. “Earnings Losses of
Displaced Workers.” American Economic Review 83(4): 685–709.
Kletzer, Lori G. 1998. “Job Displacement.” Journal of Economic Perspectives 12(1): 115–36.

62

Lee, Jin Young. 2014. “The Plateau in U . S . Women ’ S Labor Force Participation : A Cohort
Analysis.” 53(1): 46–71.
Mincer, Jacob. 1974. Schooling, Experience, and Earnings. Human Behavior & Social
Institutions No. 2. National Bureau of Economic Research.
Parker, Jonathan A. 1999. “The Reaction of Household Consumption to Predictable Changes in
Payroll Tax Rates.” American Economic Review 89(4): 959–73.
Ruhm, Christopher J. 1991. “Are Workers Permanently Scarred by Job Displacements?”
American Economic Review 81(1): 319–24.
Saporta-Eksten, Itay. 2014. “Job Loss, Consumption and Unemployment Insurance.”
Unpublished manuscript (October): 1–50.
Souleles, Nicholas S. 1999. “The Response of Household Consumption to Income Tax Refunds.”
American Economic Review 89(4): 947–58.
Stephens, Melvin. 2001. “The Long-Run Consumption Effects of Earnings Shocks.” Review of
Economics and Statistics 83(1): 28–36.
———. 2002. “Worker Displacement and the Added Worker Effect.” Journal of Labor
Economics 20(3): 504–37.
Stevens, Ann Huff. 1997. “Persistent Effects of Job Displacement: The Importance of Multiple
Job Losses.” Journal of Labor Economics 15(1): 165–88.
Topel, Robert. 1990. “Specific Capital and Unemployment: Measuring the Costs and
Consequences of Job Loss.” Carnegie-Rochester Conference Series on Public Policy 33:
181–214.
von Wachter, Till, Jae Song, and Joyce Manchester. 2007. “Long-Term Earnings Losses due to
Job Separation During the 1982 Recession.” Columbia University Department of
Economics Discussion Paper Series (708–16).
Zeldes, Stephen P. 1989. “Consumption and Liquidity Constraints: An Empirical Investigation.”
Journal of Political Economy 97(2): 305–46.

63

CHAPTER 2
The Evolution of Earnings Volatility
During and After the Great Recession1
2.1 Introduction
In a world with incomplete financial markets, securing a stable income path is important for
individuals’ welfare. Accordingly, the topic of changes in earnings volatility has received
continual attention from various researchers since the seminal work by Gottschalk and Moffitt
(1994).
A great deal of literature has documented changes in men’s earnings volatility from the
1970s through the mid-2000s. The consensus is that men’s earnings volatility increased
substantially over the 1970s, reached its peak in the early 1980s, and decreased somewhat but
still remained at a higher level than in the 1970s through the mid-2000s. It is still a matter of
controversy whether the upward trend resumed around 2000 (Cameron and Tracy 1998; Celik et
al. 2012; Congressional Budget Office 2007; Dynan, Elmendorf, and Sichel 2012; Haider 2001;
Moffitt and Gottschalk 2002, 2012; Shin and Solon 2011). A smaller literature on women’s
earnings volatility has found trends entirely different from men’s. Women’s earnings volatility is
greater than men’s, but it trended steadily downward from the 1970s until prior to the Great
Recession (Dynan, Elmendorf, and Sichel 2012; Ziliak, Hardy, and Bollinger 2011). 2

1
2

This chapter was published in Industrial Relations in 2016
See Dynan et al. (2012) for a more detailed summary of the literature on earnings volatility.

64

This paper explores the evolution of earnings volatility during and after the Great Recession,
which has been described as “the deepest downturn in the postwar era” (Elsby et al. 2011). Using
longitudinally matched Current Population Survey data for 1979-2012, I update the evidence on
men's and women's earnings volatility to include the Great Recession and its aftermath, 3 and I
then explore the main factors underlying the identified patterns. Although previous studies of
earlier periods have documented counter-cyclicality in earnings volatility (Dynan, Elmendorf,
and Sichel 2012; Shin and Solon 2011; Ziliak, Hardy, and Bollinger 2011), the present study
goes further in examining the sources of that counter-cyclicality.
The first main finding of this paper is that men’s earnings volatility increased significantly
during the Great Recession, to such an extent that the rise was comparable to the one that
occurred during the severe recession of the early 1980s. Women’s earnings volatility also
increased after the Great Recession began, even though it had shown a strong downward trend
prior to that time. These patterns are evident across most subgroups defined by age or education,
for both men and women.
I go on to show that the counter-cyclicality of earnings volatility is due mainly to countercyclicality in the volatility of annual work hours, rather than hourly wages. Moreover, the
counter-cyclical volatility in work hours appears mainly among those workers who experienced
unemployment. Thus, the greater volatility in earnings paths during recessions stems mainly
from the annual hours reductions associated with increased unemployment.
The analysis begins in the next section with an explanation of the data source, sample
selection criteria, and measures used for earnings volatility. The empirical results and conclusion
will follow in the later sections.
3

Dynan et al. (2012) and Ziliak et al. (2011) cover the periods up to 2008, and Celik et al. (2012) and Hardy and
Ziliak (2014) until 2009. To my knowledge, this is the first paper documenting the evolution of earnings volatility in
the United States during and after the Great Recession.

65

2.2 Data
This study uses matched March Current Population Survey (CPS) data from 1979–2012. In
the following section, the appropriateness of the data set is briefly discussed. I then describe the
sample selection criteria and three measures for earnings volatility that this study mainly uses.

2.2.1 Matched CPS
From the March CPS data, I construct thirty one 2-year short panel data sets. 4 March CPS
provides the most recent annual earnings information of all datasets available at this point. Hence,
it is appropriate for tracking the recent evolution of earnings volatility since the start of the Great
Recession. In addition, it provides a large sample size and diverse demographic information.
Therefore, it enables us to obtain more reliable estimates on earnings volatility for various
subgroups as well as for the whole sample.
A well-known problem associated with using the matched March CPS data is that a portion
of the sample is typically lost from the matching process from year to year. This is because of the
particular survey design of CPS: its sampling is based on housing units, not on households.
Therefore, it does not track interviewees who relocate between interview periods, and thus a
certain portion of individuals in the March CPS fails to be matched each time.
The main focus of this study, however, is looking at the evolution of earnings volatility over
time, not the level of that volatility. Therefore, if no systematic error occurs in matching March
CPS across the sample years, the matched March CPS can be expected to provide reliable
information with respect to earnings volatility dynamics. In the literature estimating the
4

Matching March CPS over 1984-1985 and 1994-1995 is not possible because of changes in the survey design. I
followed Madrian and Lefgren (2000)’s guidelines for matching the March CPS. The detailed procedure for
matching is presented in Appendix 1.

66

evolution of earnings volatility from the matched March CPS, no specific evidence of systematic
errors in matching March CPS has been reported for the periods prior to the Great Recession
(Cameron and Tracy 1998; Celik et al. 2012; Ziliak, Hardy, and Bollinger 2011). Even during
and after the Great Recession, the matching process does not seem to have been affected
significantly by the shock from the Great Recession. 5 Thus, the matched March CPS is
considered a legitimate dataset to explore the evolution in earnings volatility for a long period
including the Great Recession.

2.2.2 Sample Selection
I use two types of samples to explore the change in earnings volatility after the Great
Recession. The first one, Sample A, consists of civilian workers ages 25 to 59 6 who worked for
at least 1 week with positive earnings in the year previous to the survey year. I exclude students
and the self-employed from the sample. Thus, I focus on workers with strong labor attachment
and little influence on decisions regarding their own wages. To prevent outliers from generating
misleading results, I exclude those who had the top and bottom 1% of annual earnings. Along the
same lines, individuals with imputed earnings data are excluded. 7 These restrictions are almost
the same as those imposed by Shin and Solon (2011) and Celik et al. (2012). Therefore, results
from Sample A will be comparable to theirs.

5

There was no serious change in the percentage of movers among March CPS interviewees since the Great
Recession. See Appendix 1 for a detailed discussion.
6
The primary purpose of this age restriction is to make the sample comparable to those used by previous papers.
When I redid my analyses for a sample ages 30 to 54, the results were qualitatively similar to those reported here.
7
The portion of imputed data due to non-reporting has trended upward in the CPS. Hence dropping imputed data
may cause sample composition bias over the sample period. Cameron and Tracy (1998) address this issue in their
appendix (see A-5). They suggest some evidence that people missing earnings reports tend to have larger earnings
volatility than those who do not. Accordingly, there is a possibility that earnings volatility would be especially
underestimated for the recent periods. In that case, the increased earnings volatility that I measure during and after
the Great Recession may understate the increase that actually occurred.

67

One caution about using Sample A, however, is that I drop all workers whose annual
earnings were zero (zero-earners). 8 This may generate seriously misleading interpretations of the
results, especially during the Great Recession, which was characterized by a remarkably high
unemployment rate that persisted for a long time (Elsby et al. 2011).
Accordingly, I have created a second sample, Sample B, which is the same as Sample A
except that it includes all zero-earners and other outliers, i.e., the top and bottom 1% of all
earners. 9 Figure 2.A1 presents the men’s and women’s proportion of zero-earners for each year
in Sample B. It shows that the men’s fraction of zero-earners displays a relatively large upsurge
from 2009 to 2011, even relative to the continuous upward trend seen from 1979 on. Also, we
observe that a sudden upsurge also occurs in the women’s zero-earner fraction from 2009 to
2011, which is distinct from its relatively flat pattern in the preceding period since 2003. Hence,
it seems valuable to consider Sample B together with Sample A. By looking at both together, we
should be able to understand more accurately the overall change in men’s and women’s earnings
volatility since the beginning of the Great Recession.
I use a total of thirty one 2-year short panel data sets from 1979-2012. For Sample B,
observations of men and women total about 240,000 and 300,000, respectively, which means
about 7700 and 9700 observations, respectively, per each panel dataset.

8

Note that zero-earners include individuals who are not in the labor force.
The March CPS has maintained the same top-coding policy since 1995. Prior to 1995, however, there had been
several changes in the policy. Larrimore et al. (2008) got access to internal earnings data corresponding to each topcoded earner in the March CPS before 1995, and generated new top-coded earnings series following the post-1995
top-coding policy from 1976 onward. I incorporate the work of Larrimore et al. (2008) to correct for the
inconsistency of top-coding policy over the sample period.
9

68

2.2.3 Measuring Earnings Volatility
I use three main approaches for measuring earnings volatility. The first is the standard
deviation of the difference in log real earnings residuals over two adjacent years. This has been
used as a standard measure for earnings volatility by various researchers since Dynarski and
Gruber (1997) (henceforth denoted as DG). The following is the specific formula for DG’s
measurement:
�Var�log(𝑌𝚤𝚤 ) �
− log(𝑌𝚤𝚤−1 )�

𝑌𝑖𝑖 indicates real earnings for individual 𝑖 at year 𝑡. �log(𝑌𝚤𝚤 ) �
− log(𝑌𝚤𝚤−1 )� is the residual

obtained by regression of (log(𝑌𝑖𝑖 ) − log(𝑌𝑖𝑖−1 )) on quadratics of age for year 𝑡 − 1, which is
expected to eliminate the life-cycle effect, cohort effect and year effect from individual log
earnings. Shin and Solon (2011) show that this measure not only captures well the size of
transitory and persistent earnings shock, it also is much less affected by the evolution of the
return to individuals’ time-invariant characteristics. It is also relatively transparent and less
dependent on parametric assumptions because it is not derived from an arbitrary complicated
earnings model. 10
One weakness of DG, however, is that it does not account for zero-earners when estimating
earnings volatility because it is based on log earnings. Thus, it cannot be applied to Sample B. As
discussed in the previous section, zero-earners seem to increase substantially during the Great
Recession, so I need to employ another measure, which can be applied to Sample B that
encompasses all zero-earners.
As an alternative measure for both Samples A and B, I employ the standard deviation of the
arc percentage change in real earnings, suggested by Dynan, Elmendorf, and Sichel (2012)
10

See Baker and Solon (2003) and Shin and Solon (2011) for related discussions.

69

(hereafter DES) for estimating earnings volatility. The specific formula that DES introduced is as
follows:
�Var �100 ∗

(𝑌𝚤𝚤 �
− 𝑌𝚤𝚤−1 )
ďż˝
𝑌�𝚤

𝑌𝑖𝑖 denotes real earnings (not log real earnings) for individual 𝑖 in year 𝑡, and 𝑌�𝑖 denotes the

average of 𝑌𝑖𝑖 and 𝑌𝑖𝑖−1. 11

(𝑌𝚤𝚤�
−𝑌𝚤𝚤−1 )
𝑌�𝚤

refers to the residual from the regression of

(𝑌𝑖𝑖 −𝑌𝑖𝑖−1 )
𝑌�𝑖

on

quadratics of age for year 𝑡 − 1, which is expected to remove the life-cycle effect, cohort effect,
and year effect. The advantage of using the DES measurement is that its interpretation is easier

and more intuitive than DG, and it can also deal with zero earners. Furthermore, its construction
causes it to be bounded below and above by -200 and +200 respectively. 12
I use one further alternative method, beyond the two measures just described: I construct
quantile distributions for the tenth, twenty-fifth, fiftieth, seventy-fifth, and ninetieth sample
percentiles of the age-adjusted arc percent change in real earnings for each sample year. This
measure can be applied to both Samples A and B because it uses arc percent change in real
earnings instead of log real earnings. This method’s main rationale for estimation of earnings
volatility was suggested by Shin and Solon (2011) (hereafter SS). 13 This rationale is to take a
more complete look at the distribution of earnings changes, in order to capture details that might
escape an examination of second-moment statistics such as in DG or DES. For example, I can
consider the difference between the 90th and 10th percentiles (say, p90-p10) as something of a
counterpart to the variance of changes in earnings, and decompose it into p90-p50 and p50-p10

11
12

If both of 𝑌𝑖𝑖 and 𝑌𝑖𝑖−1 are zero, I let

(𝑌𝑖𝑖 −𝑌𝑖𝑖−1 )
𝑌�𝑖

equal zero.

In other words, no matter how much one’s earnings change, the measured earnings volatility for the individual
by DES cannot exceed 200 in terms of absolute magnitude. This might alleviate potential problems from outliers.
13
Shin and Solon (2011) looked at the quantile distributions for change in log real earnings and a rescaled
difference in real earnings, instead of the age-adjusted arc percent change in real earnings.

70

to check the respective contributions of changes in p90-p50 and p50-p10 to the change in p90p10. 14
The three measures explained above complement each other, and thus help to generate more
robust empirical results. One caution is that, as other researchers have pointed out in the
literature, the measures cannot be taken directly as the amount of risk faced by individual
workers. This is because those measures themselves do not tell how much earnings volatility has
been expected and insured by individuals before it is realized. If people do well in dealing in
advance with increased earnings volatility, then that volatility will not be so serious a problem
for them. Hence, one must be cautious about deriving welfare implications from the estimated
change in earnings volatility.

2.3 Results
This section presents how earnings volatility evolved during and after the Great Recession by
gender, age, and education level in the United States. I then discuss how unemployment shocks
during recessions have related to the counter-cyclical behaviors of earnings volatility.

2.3.1 Men’s and Women’s Earnings Volatility during and after the Great
Recession

14

Admittedly, p90-p10 is not exactly comparable with variance measures such as DG and DES, because p90-p10
is a simple difference between arbitrarily selected percentiles, while DG and DES are a type of synthetic statistic
summarizing how much the distribution of earnings changes spreads, although their standard weighting schemes
might also be regarded as arbitrary. Accordingly, we cannot say which measure is superior, so looking at all of them
together and finding common patterns across various measures will be a reasonable approach to determining the
evolution of earnings volatility.

71

Figure 2.A2 shows the evolution of men’s earnings volatility over 1979–2011, measured by
DG and DES. 15,16 Each point in the graphs indicates the earnings volatility between the
corresponding year in the horizontal axis and the previous year.
According to the results from DG for Sample A (the graph with triangle points), men’s
earnings volatility over 2008-2009 was about 0.45. This represented an increase of about 18%
over the figure of 0.38 reported for the previous period, 2007-2008. This seems to have been an
extraordinary upsurge, because earnings volatility appeared quite stable from 1993-1994 to
2007-2008, ranging only from 0.36 to 0.39. Earnings volatilities over 2009-2010 and 2010-2011
decreased to about 0.42 and 0.41, respectively, but they were still at higher levels than in the
preceding stable period, 1993-1994 to 2007-2008.
One interesting benchmark period would be the early 1980s, when there was another severe
recession. A large upswing also occurred in men’s earnings volatility for 1981-1982, followed by
even higher earnings volatility for 1982-1983, which was the peak of the early 1980s recession
period. That was almost the same level as the peak in earnings volatility during the Great
Recession. After 1984, men’s earnings volatility started to diminish. Except for the early 1990s,
it tended to stabilize at around 0.38 to 0.40, from the mid-1990s up to 2007-2008. 17 Thus, men’s

15

CPS oversampled for certain states, in order to obtain reliable statistics by state, and generated individual
weights to adjust for oversampling. One main purpose of this paper is to estimate population values for earnings
volatility, so applying weights to estimation seems to be a reasonable procedure to obtain estimates close to the
population value (Solon, Haider, and Wooldridge 2013). I use basic weights rather than March Supplement weights,
because March Supplement weights are known to account for the newly added sample since 2002, which cannot be
matched. To make the basic weights applicable to the matched March CPS across two adjacent years, I make a new
weight for each observation by simply averaging the CPS individual basic weights over 2 years. Note that I use the
same weights for the rest of the analysis in this paper. I also do all analysis without using the weights, although the
results are not displayed in the paper. Those results are almost the same qualitatively as those estimated with the
weights.
16
Due to the large sample size, all the estimates have relatively small standard errors. Specific estimates
corresponding to Figure 2 are presented in Table 2.C1 with asymptotic standard errors.
17
The overall evolution pattern of men’s earnings volatility before the Great Recession appears consistent with
Celik et al. (2012), whose work suggests the most comparable results. The magnitudes of my estimates are
somewhat lower than theirs, because my sample selection differs slightly. However, they show almost the same
pattern in terms of how men’s earnings volatility changed from 1979 to 2009.

72

earnings volatility during the Great Recession returned to the previous historic high observed a
quarter-century earlier.
The graph with circle points in Figure 2.A2 is the estimation result by DES for the same
sample, Sample A. As we see, the evolution patterns measured by DG and DES for Sample A are
fairly close to each other, although their scales differ. Therefore, it can be said that the two
different measures, DG and DES, tell almost the same story for Sample A about the growth of
men’s earnings volatility in the Great Recession.
On the other hand, the graph with square points is the result measured by DES for Sample B,
which adds zero-earners and other outliers to Sample A. Compared with the circle points
estimated by DES for Sample A, it shows a substantial difference. First of all, it is much higher
overall, by about 37% to 72% across the sample years. Furthermore, earnings volatility for
Sample B shows an upward trend, while that for Sample A does not. As a result of this upward
trend, the highest level of earnings volatility in Sample B after the Great Recession is much
higher than that experienced during the recession of the early 1980s. This implies that adding the
1% top and bottom earners (outliers) back to Sample A, and additionally including zero-earners,
leads to a significant change in the estimated evolution of men’s earnings volatility. It actually
turns out that the zero-earners in Sample B are the main factor driving the changes: if I apply
DES to Sample B excluding the 1% top-earners, I obtain almost the same pattern but just a bit
lower one in terms of magnitude than that with the square points in Figure 2.
The stark difference between the results measured by DES for Samples A and B should be
interpreted cautiously, because it seems closely related to particular characteristics of the DES
measure and the fact that the zero-earners’ fraction increased continuously from 1979 to 2011, as
displayed in Figure 1. Specifically, in two cases, earnings volatility measured by DES increases

73

with respect to zero-earners: zero-earnings in the first year but positive earnings in the following
year, and vice versa. For both, the magnitude of arc percent change in real earnings measured by
DES is always 200, the maximum absolute value that DES can generate for any individual. 18 In
other words, DES for Sample B puts the largest weight on zero-earners’ earnings change, which
is surely an arbitrary way to account for zero-earners. Hence, the overall much higher level of
earnings volatility measured by DES for Sample B compared to that for Sample A should be due
largely to DES’s unique approach. In addition, this approach should be determinative in
generating the upward trend of earnings volatility for Sample B, because the proportion of male
zero-earners rose continuously over 1979-2011, as seen in Figure 1.
Thus, it might not be fair to take literally the difference in magnitude between DES for
Samples A and B as shown in Figure 2. However, for studying Sample B’s own evolution of
earnings volatility, DES still has some validity, if the matching process for the matched March
CPS has no systemic errors, as discussed in Section 2.2.1. Therefore, I can summarize the results
from Figure 2.A2 as follows: After the Great Recession, men’s earnings volatility increased
substantially for both Samples A and B, and the magnitude in the increase was comparable to
that during the severe recession of the early 1980s. Particularly for Sample B, the increase during
the Great Recession was so intense that the level of earnings volatility became even higher than
that during any other period from 1979 to 2008. 19

18

A hypothetical extreme example would be that of a worker who earned $0 in one year and $1 the next year. In
that case his arc percent change in annual earnings would be calculated at 200, the maximum value that can be
generated by DES. Thus, it can be controversial whether it is justifiable to give the largest weight to the earnings
change for every zero-earner. DG has a potential problem similar to DES (even though it does not deal with zeroearners): it gives the same weight to an individual whose annual earnings changed from $1 to $2 as to an individual
whose annual earnings changed from $10,000 to $20,000. One difference here is that DG is applied to a restricted
sample, Sample A, which excludes zero-earners and 1 percent top- and bottom-earners to mitigate such problems.
19
The remaining persistently high level of earnings volatility for Sample B after the initial rise at 2008-2009,
which seems different from what the earnings volatility for Sample A shows, implies that in a significant number of
cases, positive-earners became zero-earners and vice versa during and after the Great Recession.

74

For the next step, I apply the third approach, which I have labeled SS, to both Samples A and
B. As I explained in Section 2.2.3, the key idea of SS is to see the whole major quantile
distribution of the age-adjusted arc percent change in real earnings along the sample years
underlying the variance-term measure, DES. By doing this, a more detailed picture can be drawn
by capturing distributional transformation of annual change in earnings over the entire sample
years.
Figure 2.A3 shows the results of applying SS to Samples A and B. The upper panels in
Figure 2.A3 display the evolution of the 10th, 25th, 50th, 75th, and 90th percentiles of the ageadjusted arc percent change in real earnings from 1979-2011, while the lower panels exhibit the
difference between the 90th and 10th percentiles (p90-p10), the 90th and 50th percentiles (p90-p50),
and the 50th and 10th percentiles (p50-p10). These lower panels provide, in effect, summary
results for the upper panels.
The overall evolution patterns for the percentiles displayed in the upper panels look similar
across Samples A and B although the magnitudes of changes differ across the samples. In the
beginning of the Great Recession, i.e., the period from 2008-2009, the 10th percentile decreased
substantially for both samples, but more drastically in Sample B. This implies that the increase in
men’s earnings volatility after the start of the Great Recession was driven mainly by the more
frequent incidence of extreme negative change in real earnings for both Samples A and B. The
more severe decrease in the 10th percentile in Sample B than in Sample A indicates a relatively
large number of male workers became zero-earners during the Great Recession. It is noteworthy
that during the early 1980s, another severe recession period, the 10th percentile also decreased
significantly, but not so much as during the Great Recession.

75

How the substantial decrease in the 10th percentile affected earnings volatility as a whole
during the Great Recession can be observed more clearly in the lower panels. For instance, a
large increase in p50-p10 occurred for both samples during the Great Recession, and this was the
main driver of the increase in p90-p10 in that period. If we think of p90-p10 as a statistic
summarizing the dispersion of change in earnings, corresponding to variance measures such as
DG and DES, it can be argued that a large portion of men became at risk of experiencing severe
loss of earnings after the Great Recession began, and that this contributed importantly to the
overall increase in earnings volatility during that recession.
It is interesting to note that we can find similar patterns in other recession periods. A
relatively strong counter-cyclicality of p50-p10 and a relatively weak pro-cyclicality of p90-p50
are found in both samples and, as a result, overall we observe counter-cyclicality of p90-p10
throughout the whole sample period. Note that the stronger counter-cyclicality of p50-p10 is
observed in both severe recession periods, the early 1980s and the Great Recession, and this led
to a relatively large increase in p90-p10 for both samples. Especially for Sample B, the increase
in p50-p10 in the Great Recession was historic, resulting in a noteworthy upsurge in p90-p10
during that period.
To sum up, we observe that in the Great Recession, men’s earnings volatility increased to a
large extent even without the inclusion of zero-earners in the sample, and the extent of this
increase was at least as large as the one that occurred during the early 1980s recession, which
witnessed the largest increase in men’s earnings volatility prior to 2008. If we take into account
the large increase in zero-earners since 2008 in the analysis, the level of men’s earnings volatility
after the Great Recession seems to have reached its highest level since the 1970s. Additionally,
through a quantile-distribution analysis of the age-adjusted arc percent change in real earnings,

76

we observe that a larger portion of male workers had to suffer extreme reduction in their earnings
during the Great Recession, and this seems to have played a crucial role in drastically boosting
men’s earnings volatility after 2008. The increase in frequency of severe earnings loss appeared
to be a common factor driving the rise in men’s earnings volatility during other recessions as
well as the Great Recession.
Figure 2.A4 documents the evolution of women’s earnings volatility as measured by DG and
DES for Sample A and by DES for Sample B. 20 The main distinctive feature of this graph is that
women’s earnings volatility showed a continuous downward secular trend over the 1979-2011
period. This differs completely from what men’s earnings volatility showed in Figure 2, although
the level of women’s earnings volatility had been always higher than that of men, even with the
continuous downward trend. Hence, we should consider this downward trend when evaluating
the extent of the change in women’s earnings volatility during the Great Recession. Another
relevant aspect to consider is that the female zero-earners’ fraction has typically been higher than
the men’s. In Figure 2.A1 we find that the women’s zero-earners fraction ranged from 26% to 40%
while the men’s ranged from 7% to 18% over the period 1979-2008. In this respect, Sample B
plays a more crucial role in analyzing women’s earnings volatility than in analyzing men’s.
As in the analysis for men, in Figure 4, DG and DES both give us the same story for
women’s earnings volatility in Sample A. We observed that men’s earnings volatility for Sample
A went up substantially at the beginning of the Great Recession, i.e. in 2008-2009, but by
contrast, women’s earning volatility for Sample A decreased over that same period. However,
when viewed in light of Sample B, women’s earnings volatility started to track upwards slightly
from 2008 to 2009, and the magnitude of the increase should be regarded as more than first

20

Due to the large sample size, all the estimates have relatively small standard errors. Specific estimates
corresponding to Figure 2.A4 are presented in Table 2.C4 with asymptotic standard errors.

77

appears if we consider the continuous downward trend in women’s earnings volatility prior to the
Great Recession. As discussed above, bearing in mind that Sample B better represents the whole
female population than does Sample A, it seems fair to argue that women also had to go through
a significant increase in earnings volatility during the Great Recession.
Figure 2.A5 shows the results from applying SS to women in Samples A and B. In that graph
we see in more detail the distributional change in arc percent change in females’ earning over the
sample years. The main features of Figure 2.A5 may be summarized as follows: first, we observe
a large distributional change from Sample A to Sample B, even compared with the one evident
for men (see Figure 3). This implies that zero-earners have a more important role for women’s
earnings volatility than for men’s, as previously anticipated. To be more specific, the gap
between p90 and p10 becomes much bigger in Sample B than in Sample A; this is well
summarized in the evolution of p90-p10 in the lower panels. 21
The second noteworthy feature in Figure 2.A5 is that strong counter-cyclicality for p50-p10
and weak pro-cyclicality for p90-p50 are observed, as they were in the analysis for men (see
Figure 3); this appears more clearly in Sample B. As a result, a relatively clear countercyclicality for p90-p10 is found in Sample B, even considering its downward trend. We note that
p90-p10 for Sample B rose after 2008-2009, when the Great Recession started; this result is
consistent with what DES shows us about women’s earnings volatility in Figure 4.
To summarize, if it is considered that Sample B is more appropriate for the analysis of
women’s earnings volatility, it can be concluded that women’s earnings volatility also increased
21

A notable aspect is that the gap between p75 and p25 for females in Sample B had been relatively small over
the whole period, compared to that for males in the same sample (see Figure 3); this means that a relatively large
portion of women experienced little earnings change over 1979-2011, compared to men. Thus, we see that the
continuous reduction in women’s earnings volatility over the sample period had been driven mostly by the decrease
in the probability that women would experience severe earnings change; this is likely to be associated closely with
the fact that the zero-earners’ fraction for women decreased considerably during the same period (see Figure 1).

78

during the Great Recession, even though it displayed a continuous downward trend prior to that
time. The main factor behind the rise seems to be the strong counter-cyclicality of p50-p10 for
arc percent change in real earnings, as observed in the analysis for men. This means that during
the Great Recession, women also became more likely to experience severe real earnings loss, and
that was the main reason why women’s earnings path also became more volatile during the Great
Recession.

2.3.2 Subgroup Analysis by Education and Age
This section presents the results when earnings volatility evolution is disaggregated by
education and age subgroups. I have used all three measures—DG, DES, and SS—to estimate
earnings volatility for each subgroup in both Samples A and B, but for brevity I mainly display
results measured by DES for subgroups in sample B, which contains the zero-earners group.
Other subgroup analysis results appear in the appendix. 22
The upper panel of Figure 2.A6 shows the change in men’s earnings volatility by three
education subgroups: those with high school diplomas or less, with some college, and with
college degrees. On top of that, the lower panel of Figure 2.A6 shows the analysis for male age
subgroups: ages 25-34, 35-49, and 50-59. 23,24
Within each subgroup analysis for education and age, we find some level-differences across
subgroups. For instance, in the education subgroup analysis, we observe that the overall level of
earnings volatility is inversely related to education level: the less-educated had overall higher
earnings volatility throughout the sample period than did those with more education. On the

22

Refer to Figure 2.B2-Figure 2.B5.
Specific estimates corresponding to Figure 2.A6 are presented in Table 2.C5 with asymptotic standard errors.
24
I also have performed additional subgroup analyses for four education- and seven age-subgroups, and the
results are almost the same qualitatively. Please refer to Figure 2.B6-Figure 2.B11
23

79

other hand, in the age subgroup analysis, we see that the overall level of earnings volatility is
higher for the younger (25-34) and older (50-59) men than the middle-aged (35-49) ones. 25
However, in terms of how earnings volatility evolved during and after the Great Recession,
all subgroups show a story similar to what we previously saw for the aggregated sample. Across
all education or age subgroups, men’s earnings volatility increased substantially after the Great
Recession began, at which time it reached a historic high over the entire sample period.
Furthermore, this increased level of earnings volatility was sustained across all the subgroups
throughout the Great Recession, as was also observed in the aggregated sample.
It is noteworthy that the main factor driving the subgroup patterns above also turns out to be
quite similar to that for the aggregated sample. In an additional analysis where I apply SS to all
subgroups, 26 we observe a substantial increase in frequency of severe earnings loss for every
male subgroup in the Great Recession, and this was true for other previous recessions, too. This
implies that intensive negative change in earnings should be considered a universal factor
affecting all male subgroups by education or age, which plays a crucial role in generating the
counter-cyclical behavior of men’s earnings volatility.
Figure 2.A7 displays the results for female subgroup analyses by education (the upper panel)
and age (the lower panel). 27,28 We observe a similar pattern with men’s education subgroups that
overall earnings volatility increases with a decrease in education. But as for women’s age
subgroups, we find a bit different patterns from men’s. Note that in the case of men, the middle-

25

These level differences seem to be closely related to the differences in working-hours volatility across the male
subgroups, considering that working-hours volatility considerably accounts for overall earnings volatility, as we will
see in Section 2.3.3. For example, less skilled younger or older male workers are likely to have less job security
compared to high skilled and middle-aged ones. Also, general differences in the level of earnings across the male
subgroups would be another factor causing such level differences in earnings volatility.
26
See Figure 2.B2.
27
Specific estimates corresponding to Figure 2.A6 are presented in Table 2.C6 with asymptotic standard errors.
28
I also have performed additional subgroup analyses for four education- and seven age-subgroups, and the
results are almost the same qualitatively. Please refer to Figure 2.B12-Figure 2.B17

80

aged men show relatively low earnings volatility compared to the younger and older. In the case
of women, however, the overall level of women’s earnings volatility tends to decrease with age.
It might reflect generally different labor participation patterns between men and women. For
example, women tend to enter and leave the labor market more frequently than men over the
whole life cycle, and women’s entrances and exits also tend to be frequent even in middle age;
this differs from typical men’s labor-market behavior. 29
Even with some level differences of earnings volatility across the female subgroups by
education and age, they generally show a consistent evolution pattern of earnings volatility,
which is similar to that for the female aggregated sample. Except for the female college graduate
group and the 25-34 age group, all other female education and age subgroups display an upsurge
in earnings volatility after the Great Recession began. The main factor driving the rise was the
increase in severe negative change in earnings, 30 as we saw previously in the analysis for the
female aggregated sample. It also turns out to be a main factor raising earnings volatility during
earlier recessions. This implies, as in the men’s subgroup analyses, that severe earnings loss was
a prevalent and important factor underlying the counter-cyclical behavior of women’s earnings
volatility across most female subgroups by education or age.
To sum up, the increasing patterns of earnings volatility we see in the whole aggregated
sample during the Great Recession also appear generally in various male and female subgroups.
Furthermore, the increased incidence of severe negative change in earnings appears to be a
common factor producing the counter-cyclical volatility in earnings across most male and female
subgroups, as in the analysis for the aggregated sample. Particularly during the Great Recession,
we observe a historic upswing in the frequency of large earnings loss for most male and female
29

To be more specific, women’s work hours tend to become more volatile than men’s depending on changes in
family compositions, such as marriage, pregnancy, and child care.
30
See Figure 2.B3.

81

subgroups, so it led to a historic increase in men’s and women’s earnings volatility for the
aggregated group as well as for each subgroup.

2.3.3 Unemployment, Work Hours, and Earnings Volatility in the Great
Recession.
So far, we have found that male and female workers experienced a substantial upsurge in
earnings volatility during the Great Recession, even though the increase for women was not so
great as that for men. A historic increase in the frequency of severe earnings loss was found to be
a significant factor in the rise in men’s and women’s earnings volatility during the Great
Recession. Moreover, the increased tendency to experience large earnings loss seems to largely
account for the rises in earnings volatility during previous recessions. Through the subgroup
analysis in the previous section, we saw that these patterns were also generally observed across
most subgroups by education and age for both men and women.
Then, a natural question arises: what is the fundamental driving force that caused the
substantial increases in the frequency of severe loss in earnings during recessions including the
Great Recession? When we consider that the Great Recession was characterized mainly by an
extraordinarily high unemployment rate that persisted for a relatively long time, studying the
relationship between unemployment and earnings volatility appears to be a sensible starting point
to deal with this question.
In the March CPS, a survey question asks whether an interviewee looked for a job or was laid
off during nonworking weeks in the year prior to the survey. By using the answers to this
question, I calculate the yearly portion of individuals who had worked for less than or equal to 49
weeks during the preceding year and who had experienced unemployed status in their non-

82

working periods. Figure 2.A8 shows the results for men and women from 1979 to 2011. In that
graph, the counter-cyclicality of the men’s and women’s unemployment rate is easily
recognizable for both Samples A and B. We observe especially that the percentage of male
workers who experienced unemployment went up more substantially after the Great Recession
than in any other recession periods. It suggests that a high positive correlation might exist
between unemployment and earnings volatility.
To look at how closely the experience of unemployment relates to men’s and women’s
earnings volatility during the Great Recession, I re-estimated men’s and women’s earnings
volatility after excluding from the matched samples individuals who had experienced
unemployment. Figure 2.A9 summarizes the results; the upper panel displays results for men, the
lower panel for women. For both upper and lower panels, the line with X-shaped points denotes
earnings volatility measured by DES for Sample A, and the line with triangle points stands for
earnings volatility measured by DES for Sample A excluding workers who experienced
unemployment during the 2 years in the reporting period 31 (for convenience, I call this “Sample
A without the unemployed”). Similarly, for both the upper and lower panels, the line with circle
points denotes earnings volatility measured by DES for Sample B, and the line with square
points stands excludes the unemployed.
A noteworthy feature in the upper panel of Figure 2.A9 is that a large portion of the increase
in men’s earnings volatility after the Great Recession in the original samples largely disappears
in the samples without the unemployed. Moreover, while a significant counter-cyclicality of
men’s earnings volatility over the whole sample period is observed in the original samples, this
diminishes substantially for samples that exclude the unemployed.
31

According to the discussion above, “the individuals having experienced unemployment during the consecutive
2 years” are defined as those who worked less than or equal to 49 weeks, and were laid off or looked for jobs in the
nonworking period in either the first year or second year.

83

In the lower panel of Figure 2.A9, we observe similar patterns for women. If I remove the
unemployed from the original Sample B, the increase in women’s earnings volatility during the
Great Recession disappears, and the continuing downward trend in volatility becomes more
evident. As for women in Sample A, similarly we see that excluding the unemployed makes the
downward trend of earnings volatility more obvious during the recession periods. That pattern is
especially evident in the Great Recession.
The observations above suggest that the intense increase in earnings volatility after the Great
Recession likely reflected the severity of the unemployment shock during the Great Recession.
In addition, the earnings shock from unemployment seems to have been a common influence
driving the rise in earnings volatility during preceding recessions.
Considering that earnings data are a mixture of market-wage and hours-worked information,
the above analysis implies that a large source of the increase in earnings volatility during the
Great Recession might be hours-worked volatility, rather than market-wage volatility. To
explicitly investigate this point, I decompose earnings volatility into the variance of change in
log annual worked hours, the variance of change in log hourly wage, and the covariance of the
two terms. The specific decomposition formula is as follows:
Var(∆ log(𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒)) = Var(∆ log(ℎ𝑜𝑜𝑜𝑜)) + Var(Δ log(𝑤𝑤𝑤𝑤))
+2Cov(∆ log(ℎ𝑜𝑜𝑜𝑜) , Δ log(𝑤𝑤𝑤𝑤))

Here earnings means annual real earnings, hours indicates total annual working hours, and
wage implies hourly wage, calculated from the March CPS by dividing earnings by hours.
∆ log(∙) refers to year-to-year change in log terms, and ∆ log(𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒), ∆ log(ℎ𝑜𝑜𝑜𝑜),

Δ log(𝑤𝑤𝑤𝑤) were all age-adjusted for the analysis. Note that Var(∆ log(𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒)) is the same
measure as DG, except that it is expressed in terms of the variance rather than the standard

84

deviation. Because it uses logarithms, it can be applied to only Sample A, whose earnings are all
positive.
Figure 2.A10 presents results from this decomposition analysis for men (upper panel) and
women (lower panel) in Sample A. For both the upper and lower panels, the upper line with
square points represents the variance version of earnings volatility previously measured by DG
for sample A. Lines below with circle points, diamond points, and x-shaped points are graphs for
Var(∆ log(ℎ𝑜𝑜𝑜𝑜)), Var(Δ log(𝑤𝑤𝑤𝑤)), and 2Cov(∆ log(ℎ𝑜𝑜𝑜𝑜) , Δ log(𝑤𝑤𝑤𝑤)), respectively.
If we add up all of these, the upper graph with square points will be obtained. Note that the

consistent negativity of the covariance terms is due largely to the innate negative correlation
between total annual working hours and calculated hourly wage (= annual earnings / total annual
working hours).
Noteworthy in the upper panel of Figure 2.A10 is that, among men, the counter-cyclicality of
Var(∆ log(ℎ𝑜𝑜𝑜𝑜)) has been more intense and evident than that of Var(Δ log(𝑤𝑤𝑤𝑤)). Thus, the
counter-cyclicality of men’s earnings volatility observed for all sample years seems to have
come mainly from counter-cyclical annual work hours volatility. Particularly for the Great
Recession period, the upsurge in work hours volatility (Var(∆ log(ℎ𝑜𝑜𝑜𝑜))) was more severe

than in any prior recession period. Therefore, it can be argued that the drastic increase in work
hours volatility had a more important role in raising men’s earnings volatility in Sample A
during the Great Recession than it did in any other recession.
The lower panel of Figure 2.A10 depicts the decomposition results for women in Sample A.
As seen previously, women’s earnings volatility continuously decreased, in contrast to men’s.
An interesting aspect in the lower panel of Figure 2.A10 is that the downward trend of women’s
earnings volatility has been driven largely by a decrease in women’s work-hours volatility.

85

Moreover, the strong counter-cyclicality of women’s earnings volatility during the early 1980s
and 1990s and the relatively weak counter-cyclicality during the early 2000s and the Great
Recession also seem mainly dependent on how counter-cyclical women’s work-hours volatility
was for each corresponding time.
Overall, Figure 2.A10 tells us that the evolution of earnings volatility can be accounted for
largely by work-hours volatility (Var(∆ log(ℎ𝑜𝑜𝑜𝑜)) rather than hourly wage volatility

(Var(∆ log(𝑤𝑤𝑤𝑤))). In particular, a large percentage of men experienced significant changes in
their work hours in the Great Recession, compared to previous years, and this led them to
experience large variations in their annual earnings during this period.
Connecting the results above with the previous unemployment analysis in Figure 2.A9, we
can come up with a more complete explanation for the counter-cyclicality of earnings volatility
over the sample period, as well as the substantial increase in earnings volatility during the Great
Recession. The increase in earnings volatility during recessions seems driven mainly by an
increase in work-hours volatility, and that increase appears to come largely from unemployment
shocks. Figure 2.A11 corroborates this link: I decompose the variance of change in log worked
hours (work-hours volatility) into two components, one from individuals who experienced
unemployment, the other from those who did not. 32 As is seen in the picture, the countercyclicality of the variance of the change in log worked hours is largely due to the group that
experienced unemployment. Especially in the Great Recession, a historic unemployment shock
occurred—in other words, there was an unprecedented chance to experience reduced working
hours—and this facilitated the momentous upsurge in men’s earnings volatility. It also
substantially slowed the downward trend of women’s earnings volatility.

32

See Appendix 2 for the detailed decomposition method.

86

2.4 Conclusion
Men’s and women’s earnings volatility increased substantially during the Great Recession.
Especially for men, the increase was at least as great as it had been during the last severe
recession, in the early 1980s. The rise in earnings volatility was largely due to a historic increase
in the probability of a reduction in annual earnings. Furthermore, the increase in the frequency of
severe earnings loss was mainly due to intensive decreases in hours worked among the workers
experiencing unemployment during the Great Recession. This empirical pattern is also found to
be a common main factor driving the increase in earnings volatility during previous recessions.
Accordingly, it seems that unemployment shocks have played a central role in generating the
strong counter-cyclical behavior of earnings volatility over the last three decades.
As discussed in Section 2.2.3, the increase in earnings volatility cannot be equated directly
with increased risk or uncertainty if individuals insure against earnings fluctations in advance.
For the substantial rise in earnings volatility during recessions including the Great Recession,
however, an unwanted decrease in hours worked due to unemployment appears to have played a
crucial part. This implies that individuals were likely to go through an unexpected decrease in
their earnings during recessions. Particularly for the Great Recession, there was a historic
increase in unemployment, and it persisted for a long time. Moreover, the unemployment shock
hit more severely the less-skilled and younger male workers (Elsby et al. 2011), who were less
likely to have enough savings or credit for a buffer against earnings shocks. Therefore, the
findings of this paper indicate a possibility that the substantial rise in earnings volatility during
the Great Recession would have caused a severe loss in social welfare.
Additional studies, however, are still required to clarify the welfare implications of the
substantial rise in earnings volatility during the Great Recession. For instance, we need to

87

investigate how the increase in individual earnings volatility affected family income volatility.
This paper shows that there was a historic upsurge in men’s earnings volatility and a significant
rise in women’s during the Great Recession. As earnings account for a large share of total family
income generally (Dynan, Elmendorf, and Sichel 2012), we can be confident that the increased
earnings volatility for both men and women translated into increased family income volatility as
well.
However, a remaining empirical question is the extent to which the increase in earnings
volatility was alleviated through various family income adjustments during the Great Recession.
For example, a negative earnings shock to an individual could be mitigated by family labor
supply adjustment, reductions in tax rates, governmental income transfer programs, and private
transfers from relatives or friends (Dynarski and Gruber. 1997; Stephens, 2002).
Another relevant question is how the increase in individual earnings volatility affected family
consumption in the Great Recession. Family consumption responses to earnings changes depend
on saving and borrowing behavior as well as the family income adjustments already discussed.
Hence, studying the consumption consequences of earnings changes will require data on
consumption as well as family income (Blundell, Pistaferri, and Preston 2008; Dynarski and
Gruber. 1997; Gorbachev 2014).
Therefore, investigating the dynamics of both family income changes and family
consumption responses in the Great Recession is a natural extension of the research reported here.
Implementing that extension is likely to require an alternative data set with richer information on
family income and consumption, probably the Panel Study of Income Dynamics. I intend to
pursue this line of inquiry in my own future research.

88

APPENDICES

89

APPENDIX A

FIGURES FOR CHAPTER 2

90

.3
.25
.2

Year
Male

Female

1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011

.3

30

35

.35

.4

40

45

.45

DG

DES

.5

50

55

.55

.6

60

65

.65

Figure 2.A2 Men’s Earnings Volatility (DG and DES; Ages 25-59)

Year
DG for sample A; male

DES for sample A; male

DES for sample B; male

Notes: DG refers to the standard deviation of difference in log real earnings residuals over two adjacent years (see Section 2.2.3).
DES refers to the standard deviation of age-adjusted arc percent change in real earnings (see Section 2.2.3). Sample A
consists of civilian workers, ages 25-59, having worked at least a week with positive earnings in the previous year of the
survey year, not self-employed, not students, and with no imputed earnings data. Furthermore, Sample A excludes the
top and bottom 1 percent earners among individuals who satisfy the prelisted criteria. Sample B includes zero-earners
and the top and bottom 1percent earners in addition to those in Sample A.

91

2011

2010

2009

2008

2007

2006

2005

2004

2003

2002

2001

2000

1999

1998

1997

1996

1995

1994

1993

1992

1991

1990

1989

1988

1987

1986

1985

1984

1983

1982

1981

1980

1979

.05

.1

.15

Zero-earners' Proportion

.35

.4

Figure 2.A1 Fraction of Zero-earners in Matched March CPS

Figure 2.A3 Men’s Earnings Volatility (SS; Ages 25-59)
Sample A; male
p25

Sample B; male

p50

p75

p90

p10

p25

p50

p75

p90

p90-p50

2011

2007

2009

2003

2005

2001

1997

1999

1993

1995

1991

1987

1989

1983
1983

1985

1981
1981

p50-p10

p90-p10

p90-p50

2011

2009

2007

2005

2003

2001

1999

1997

1995

1993

1991

1989

1987

1985

1979

2011

2009

2007

2005

2003

2001

1999

1997

1995

1993

1991

20
p90-p10

1989

1987

1985

1983

1981

1979

20

40

40

60

60

80

80

100

1979

2009

2011

2005

2007

2003

1999

2001

1995

1997

1993

1989

1991

1985

1987

1981

1983

100

1979

-60

-60

-40

-40

-20

-20

0

0

20

20

40

40

p10

p50-p10

Notes: SS refers to the method of looking at the evolution of major quantiles for age-adjusted arc percent change in real earnings (see
Section 2.2.3). Sample A consists of civilian workers, ages 25-59, having worked at least a week with positive earnings in the
previous year of the survey year, not self-employed, not students, and with no imputed earnings data. Furthermore, Sample A
excludes the top and bottom 1percent earners among individuals who satisfy the prelisted criteria. Sample B includes zeroearners and the top and bottom 1percent earners in addition to those in Sample A.

1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011

35

40

45

50

55
DES

60

65

70

.35 .4 .45 .5 .55 .6DG.65 .7 .75 .8 .85 .9

75

Figure 2.A4 Women’s Earnings Volatility (DES & DG; Ages 25-59)

Year
DG for sample A; female

DES for sample A; female

DES for sample B; female

Notes: DG refers to the standard deviation of difference in log real earnings residuals over two adjacent years (see Section 2.2.3).
DES refers to the standard deviation of age-adjusted arc percent change in real earnings (see Section 2.2.3). Sample A
consists of civilian workers, ages 25-59, having worked at least a week with positive earnings in the previous year of the
survey year, not self-employed, not students, and with no imputed earnings data. Furthermore, Sample A excludes the
top and bottom 1 percent earners among individuals who satisfy the prelisted criteria. Sample B includes zero-earners
and the top and bottom 1percent earners in addition to those in Sample A.

92

Figure 2.A5 Women’s Earnings Volatility (SS; Ages 25-59)
Sample A; female
p25

Sample B; female

p50

p75

p10

p90

p25

p50

p75

p90

1997

1999

2001

2003

2005

2007

2009

2011

1999

2001

2003

2005

2007

2009

2011

1995

1997

1993

1991

2011

1989

2009

2011

1987

2007

2009

1985

2005

2007

1983

2003

2005

1981

2001

2003

1979

1999

2001

p90-p10

p90-p50

p50-p10

p90-p10

1995

1993

1991

1989

1987

1985

1983

1981

1979

1995

1993

1991

1989

1987

1985

1983

40

60

80

100

120

1997

1999

1995

1997

1993

1991

1989

1987

1985

1983

1981
1981

1979

40

60

80

100

120

1979

-50

-50

0

0

50

50

100

100

p10

p90-p50

p50-p10

Notes: SS refers to the method of looking at the evolution of major quantiles for age-adjusted arc percent change in real earnings
(see Section 2.2.3). Sample A consists of civilian workers, ages 25-59, having worked at least a week with positive
earnings in the previous year of the survey year, not self-employed, not students, and with no imputed earnings data.
Furthermore, Sample A excludes the top and bottom 1percent earners among individuals who satisfy the prelisted
criteria. Sample B includes zero-earners and the top and bottom 1percent earners in addition to those in Sample A.

93

60
50
40

2011

2010

2009

2008

2007

2006

2005

2004

2003

2002

2001

2000

1999

1998

1997

1996

Some College

College Grad.

60
50

Aged 25-34

Aged 35-49

2011

2010

2009

2008

2007

2006

2005

2004

2003

2002

2001

2000

1999

1998

1997

1996

1995

1994

1993

1992

1991

1990

1989

1988

1987

1986

1985

1984

1983

1982

1981

1980

1979

40

Earnings Volatility (DES)

70

HS Grad. or Less

1995

1994

1993

1992

1991

1990

1989

1988

1987

1986

1985

1984

1983

1982

1981

1980

1979

30

Earnings Volatility (DES)

70

Figure 2.A6 Men’s Earnings Volatility by Education and Age (DES for Sample B)

Aged 50-59

Notes: DES refers to the standard deviation of the age-adjusted arc percent change in real earnings (see Section 2.2.3). Sample B includes
zero-earners and the top and bottom 1percent earners in addition to those in Sample A; Sample A consists of civilian workers,
ages 25-59, having worked at least a week with positive earnings in the previous year of the survey year, not self-employed, not
students, and with no imputed earnings data. Furthermore, Sample A excludes the top and bottom 1percent earners among
individuals who satisfy the prelisted criteria.

94

70
65

2011

2010

2009

2008

2007

2006

2005

2004

2003

2002

2000

1999

1998

1997

1996

2001

College Grad.

Some College

80
70
60

Aged 25-34

Aged 35-49

2011

2010

2009

2008

2007

2006

2005

2004

2003

2002

2001

2000

1999

1998

1997

1996

1995

1994

1993

1992

1991

1990

1989

1988

1987

1986

1985

1984

1983

1982

1981

1980

1979

50

Earnings Volatility (DES)

90

HS Grad. or Less

1995

1994

1993

1992

1991

1990

1989

1988

1987

1986

1985

1984

1983

1982

1981

1980

1979

55

60

Earnings Volatility (DES)

75

80

Figure 2.A7 Women’s Earnings Volatility by Education and Age (DES for Sample B)

Aged 50-59

Notes: DES refers to the standard deviation of the age-adjusted arc percent change in real earnings (see Section 2.2.3). Sample B includes
zero-earners and the top and bottom 1percent earners in addition to those in Sample A; Sample A consists of civilian workers,
ages 25-59, having worked at least a week with positive earnings in the previous year of the survey year, not self-employed, not
students, and with no imputed earnings data. Furthermore, Sample A excludes the top and bottom 1percent earners among
individuals who satisfy the prelisted criteria.

95

Figure 2.A8 The Yearly Fraction of Male and Female Workers Having Experienced
Unemployment

2004

2005

2006

2007

2008

2009

2010

2011

2005

2006

2007

2008

2009

2010

2011

2003

2004

2002

2001

2000

1999

1998

1997

1996

1995

1994

1993

1992

1991

1990

1989

1988

1987

1986

1985

1984

1983

1982

1981

1980

1979

.05

.1

.15

.2

.25

Male

Unemployment-experience rate for sample A; male
Unemployment-experience rate for sample B; male

2003

2002

2001

2000

1999

1998

1997

1996

1995

1994

1993

1992

1991

1990

1989

1988

1987

1986

1985

1984

1983

1982

1981

1980

1979

.05

.1

.15

.2

.25

Female

Unemployment-experience rate for sample A; female
Unemployment-experience rate for sample B; female

96

50
40

2010

DES for sample A; male

DES for sample B w/o unemp.-experienced; male

DES for sample B; male

2011

2008

2009

2006

2007

2005

2003

2004

2001

2002

1999

2000

1997

1998

1995

1996

1993

1994

1991

1992

1989

1990

1987

1988

1985

1986

1983

1984

1981

1979

1982

DES for sample A w/o unemp.-experienced; male

60
50

DES for sample A w/o unemp.-experienced; female

DES for sample A; female

DES for sample B w/o unemp.-experienced; female

DES for sample B; female

2011

2010

2008

2009

2007

2005

2006

2003

2004

2001

2002

2000

1999

1998

1997

1995

1996

1993

1994

1991

1992

1990

1989

1987

1988

1985

1986

1983

1984

1982

1981

1980

1979

30

40

Earnings Volatility (DES)

70

80

1980

20

30

Earnings Volatility (DES)

60

Figure 2.A9 The Effect of Unemployment on Men’s and Women’s Earnings Volatility

Notes: DES refers to the standard deviation of the age-adjusted arc percent change in real earnings (see Section 2.2.3). Sample A consists
of civilian workers, ages 25-59, having worked at least a week with positive earnings in the previous year of the survey year, not
self-employed, not students, and with no imputed earnings data. Furthermore, Sample A excludes the top and bottom 1percent
earners among individuals who satisfy the prelisted criteria. Sample B includes zero-earners and the top and bottom 1percent
earners in addition to those in Sample A.

97

Figure 2.A10 Decomposition of DG for Men and Women in Sample A

1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011

-.1

0

.1

.2

Male

Var(Δlog(hours))

Var(Δlog(wage))

2Cov(Δlog(hours),Δlog(wage))

DG for Sample A (variance version)

1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011

-.2

0

.2

.4

.6

Female

Var(Δlog(hours))

Var(Δlog(wage))

2Cov(Δlog(hours),Δlog(wage))

DG for Sample A (varinace version)

Notes: DG means the standard deviation of difference in log real earnings residuals over two adjacent years (see
Section 2.2.3.) Sample A consists of civilian workers, ages 25-59, having worked at least a week with positive
earnings in the previous year of the survey year, not self-employed, not students, and with no imputed earnings
data. Furthermore, Sample A excludes the top and bottom 1percent earners among individuals who satisfy the
prelisted criteria.

98

Figure 2.A11 Decomposition of Var(Δlog(hours)) for Men and Women in Sample A

1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011

0

.05

.1

.15

.2

Male

Year
Var(Δlog(hours))

unemp.-experienced

non-unemp.-experienced

1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011

0

.1

.2

.3

.4

Female

Year
Var(Δlog(hours))

unemp.-experienced

non-unemp.-experienced

Note : Sample A consists of civilian workers, ages 25-59, having worked at least a week with positive
earnings in the previous year of the survey year, not self-employed, not students, and with no imputed
earnings data. Furthermore, Sample A excludes the top and bottom 1 percent earners among
individuals who satisfy the prelisted criteria.

99

APPENDIX B

SUPPLEMENTAL FIGURES FOR CHAPTER 2

100

Male

101

Year
Female

2011

2010

2009

2008

2007

2006

2005

2004

2003

2002

2001

2000

1999

1998

1997

1996

1995

1994

1993

1992

1991

1990

1989

1988

1987

1986

1985

1984

1983

1982

1981

1980

1979

.3

.4

.5

Match Rate
.6
.7

.8

.9

Figure 2.B1 Yearly Match Rate for Matched March CPS

p90-p10

p90-p50

p50-p10

p90-p10

102

p90-p50

2011

p90-p50
2011

2009

p90-p50

2009
2011

2009

p90-p50

2007

2005

2003

2001

2011

2009

2007

2005

2003

2001

1999

p90-p50

2007

2005

2003

2001

1999

1997

2011

2009

2007

2005

2003

2001

1999

1997

1995

1993

1991

1989

1987

1985

1983

1981

1979

20 40 60 80 100

20 40 60 80 100

p90-p50

2007

2005

2003

2001

p90-p10
1999

p90-p10

1999

p90-p10

1997

1995

80

p90-p10

1997

1995

60

2011

2009

2007

2005

2003

2001

1999

1997

1995

1993

1991

1989

1987

1985

1983

1981

1979

p90-p10

1997

Sample A
1995

Sample A

1995

Sample A
1993

1991

1989

40

80

Sample A

1993

1991

1989

1987

20

60

Sample A

1993

1991

1989

1987

1985

1983

1981

1979

2011

40

2011

2009

2007

2005

2003

2001

1999

1997

1995

1993

1991

1989

1987

1985

1983

1981

1979

2011

2009

2007

2005

2003

2001

1999

1997

1995

1993

1991

1989

1987

1985

1983

1981

1979

20 40 60 80100120

20 40 60 80100120

Sample A

1993

1991

1989

p50-p10
1987

p50-p10
1985

1983

1981

1979

20 40 60 80100120

p50-p10

1985

1983

1981

1979

20 40 60 80 100

2011

2009

20

p50-p10

1987

20 40 60 80 100

2011

2009

2007

2005

2003

2001

1999

1997

1995

1993

1991

1989

1987

1985

1983

1981

1979

p50-p10

1985

1983

1981

1979

2011

p90-p50
2009

p90-p50

2009

p90-p50

2007

2005

2003

2001

20 40 60 80100120

p90-p50

2007

2005

2003

2001

1999

1997

1995

1993

1991

1989

1987

1985

1983

1981

1979

p90-p50

2007

2005

2003

p90-p10

2001

p90-p10

1999

20 40 60 80 100

p90-p10

1997

1995

1993

1991

1989

1987

1985

1983

1981

1979

p90-p10

1999

20 40 60 80 100

p90-p10

1997

1995

1993

1991

1989

1987

1985

1983

1981

1979

Figure 2.B2 Men’s Earnings Volatility by Education and Age (SS for A &B)
HS Grad. or Less
Sample B

Some College

p50-p10

Sample B

p50-p10

College Grad.
Sample B

Aged 25-34
p50-p10

Sample B

Aged 35-49
p50-p10

Sample B

p50-p10

Aged 50-59

Sample B

p50-p10

p90-p10

p90-p50

p50-p10

p90-p10

103

1991

p90-p50

2001

p90-p50

2011

2011

p90-p50

2007

2007

2011

2009

p90-p50

2009

2005

2003

2001

1999

1995

2011

2009

2007

2005

2003

2001

1999

1997

p90-p50

2009

2005

2003

2001

1995

1995

2011

2009

2007

2005

2003

2001

1999

1997

p90-p50

2007

2005

2003

p90-p10
1999

p90-p10
1997

p90-p10

1997

1995

p90-p10

1999

Sample A
1993

1991

1989

1987

1985

1983

1981

1979

40 60 80100120140

p90-p10

1997

Sample A

1995

Sample A
1993

1991

1989

Sample A

1993

1991

1989

1987

1985

1983

1981

1979

20 40 60 80 100120

2009
2011

40 60 80100120140

Sample A

1993

1991

1989

1985
1987

2009
2011

2007

2005

2003

2001

1999

1997

1995

1993

1991

1989

1987

1985

1983

1981

1979

2011

2009

2007

2005

2003

2001

1999

1997

1995

1993

1991

1989

1987

1985

1983

1981

1979

2011

2009

2007

2005

2003

2001

1999

1997

1995

1993

1991

1989

1987

1985

1983

1981

1979

40 60 80 100120

40 60 80 100120

Sample A

1993

p50-p10

1989

p50-p10

1987

1983

1981

1979

50 100 150 200

p50-p10

1985

20 40 60 80 100120

2009
2011

20 40 60 80 100120

p50-p10

1987

1983

1981

1979

2011

2009

2007

2005

2003

2001

1999

1997

1995

1993

1991

1989

1987

1985

1983

1981

1979

p50-p10

1985

1983

1981

1979

20 40 60 80 100

p90-p50

2009

p90-p50
2007

2005

2003

p90-p50

2011

1999
2001

50 100 150 200

p90-p50

2007

2005

2003

2001

1997

1995

1993

1991

1989

1987

1985

1983

1981

1979

p90-p50

2007

2005

2003

p90-p10
1999

p90-p10

1999

20 40 60 80 100120

p90-p10

1997

1995

1993

1991

1989

1987

1985

1983

1981

1979

p90-p10

2001

20 40 60 80 100

p90-p10

1997

1995

1993

1991

1989

1987

1985

1983

1981

1979

Figure 2.B3 Women’s Earnings Volatility by Education and Age (SS for A &B)
HS Grad. or Less
Sample B

Some College

p50-p10

Sample B

College Grad.
p50-p10

Sample B

p50-p10

Aged 25-34
Sample B

Aged 35-49
p50-p10

Sample B

Aged 50-59
p50-p10

Sample B

p50-p10

Aged 25-34

104
Aged 35-49
Aged 50-59
2011

2010

2009

2008

2007

35

40

Some College

2006

2005

2004

2003

2002

2001

2000

1999

1998

1997

1996

30

Earnings Volatility (DES)

HS Grad. or Less

1995

1994

1993

1992

1991

1990

1989

1988

1987

1986

1985

1984

1983

1982

1981

1980

1979

25

2011

2010

2009

2008

2007

2006

2005

2004

2003

2002

2001

2000

1999

1998

1997

1996

1995

1994

1993

1992

1991

1990

1989

1988

1987

1986

1985

1984

1983

1982

1981

1980

1979

25
35

40

Earnings Volatility (DES)
30
45

Figure 2.B4 Men’s Earnings Volatility by Education and Age (DES for A)

College Grad.

Aged 25-34

105
Aged 35-49
Aged 50-59
2011

2010

2009

2008

2007

45

50

Earnings Volatility (DES)
55

60

Some College

2006

2005

2004

2003

2002

2001

2000

1999

1998

1997

1996

40

HS Grad. or Less

1995

1994

1993

1992

1991

1990

1989

1988

1987

1986

1985

1984

1983

1982

1981

1980

1979

35

2011

2010

2009

2008

2007

2006

2005

2004

2003

2002

2001

2000

1999

1998

1997

1996

1995

1994

1993

1992

1991

1990

1989

1988

1987

1986

1985

1984

1983

1982

1981

1980

1979

35
45

50

Earnings Volatility (DES)
40

55

Figure 2.B5 Women’s Earnings Volatility by Education and Age (DES for A)

College Grad.

No HS Graduation

HS Graduate

106

Some College

Year

College Graduate

2011

2010

2009

2008

2007

60

70

80

90

Some College

2006

2005

2004

2003

2002

2001

2000

1999

50

Earnings Volatility (DES)
40

HS Graduate

1998

1997

1996

1995

1994

1993

1992

30

No HS Graduation

1991

1990

1989

1988

1987

1986

1985

1984

1983

1982

1981

1980

1979

20

2011

2010

2009

2008

2007

2006

2005

2004

2003

2002

2001

2000

1999

1998

1997

1996

1995

1994

1993

1992

1991

1990

1989

1988

1987

1986

1985

1984

1983

1982

1981

1980

1979

20

30

40

50

60

70

Earnings Volatility (DES)
80

90

Figure 2.B6 Men’s Earnings Volatility by Four Education Subgroups (DES for A)
Sample A; male

Year
College Graduate

Figure 2.B7 Men’s Earnings Volatility by Four Education Subgroups (DES for B)
Sample B; male

p90-p10

p90-p50

p50-p10

107

p90-p10

p90-p50

2011

p90-p50

2009

2011

2009

p90-p50

2007

2005

2003

2001

1999

2011

2009

2007

2005

2003

2001

1999

1997

1995

20 40 60 80 100120

p90-p50

2007

2005

2003

2001

p90-p10
1997

1995

p90-p10

1999

Sample A
1993

1991

1989

1987

1985

1983

1981

1979

2011

p90-p10

1997

20 40 60 80 100

Sample A

1993

1991

1989

1987

1985

1983

1981

1979

2011

20 40 60 80 100120

Sample A

1995

80

2009

2007

2005

2003

2001

1999

1997

1995

1993

1991

1989

1987

1985

1983

1981

1979

2011

2009

2007

2005

2003

2001

1999

1997

1995

1993

1991

1989

1987

1985

1983

1981

1979

2011

2009

2007

2005

2003

2001

1999

1997

1995

1993

1991

1989

1987

1985

1983

1981

1979

0

0

50

50

100

100

150

150

Sample A

1993

1991

60

p50-p10

1989

40

p50-p10

1987

20

2009

20 40 60 80 100

p50-p10

1985

1983

1981

1979

2011

p90-p50
2007

2005

2003

2001

1999

1997

1995

1993

1991

1989

1987

1985

1983

1981

1979

p90-p50

2009

80

p90-p50

2007

2005

2003

60

p90-p10

2001

40

p90-p10

1999

20

p90-p10

1997

1995

1993

1991

1989

1987

1985

1983

1981

1979

Figure 2.B8 Men’s Earnings Volatility by Four Education Subgroups (SS for A & B)

No High School Graduation
Sample B

p50-p10

High School Graduate
Sample B

p50-p10

Some College
Sample B

p50-p10

College Graduate
Sample B

p50-p10

Year

108
2011

2010

2009

2008

2007

2006

2005

2004

2003

2002

2001

2000

1999

1998

1997

1996

1995

1994

1993

1992

1991

1990

1989

1988

1987

1986

1985

1984

1983

1982

1981

1980

1979

30

35

40

45

Earnings Volatility (DES)

25

2011

2010

2009

2008

2007

2006

2005

2004

2003

2002

2001

2000

1999

1998

1997

1996

1995

1994

1993

1992

1991

1990

1989

1988

1987

1986

1985

1984

1983

1982

1981

1980

1979

30

35

40

45

Earnings Volatility (DES)

25

2011

2010

2009

2008

2007

2006

2005

2004

2003

2002

2001

2000

1999

1998

1997

1996

1995

1994

1993

1992

1991

1990

1989

1988

1987

1986

1985

1984

1983

1982

1981

1980

1979

40

Aged 25-29

35

Aged 30-34

30

45

Earnings Volatility (DES)

25

Figure 2.B9 Men’s Earnings Volatility by Seven Age Subgroups (DES for A)
Sample A; male

Aged 35-39

Year

Aged 40-44

Aged 45-49

Year

Aged 50-54

Aged 55-59

Year

109
2011

2010

2009

2008

2007

2006

2005

2004

2003

2002

2001

2000

1999

1998

1997

1996

1995

1994

1993

1992

1991

1990

1989

1988

1987

1986

1985

1984

1983

1982

1981

1980

1979

50

60

70

Earnings Volatility (DES)

40
80

2011

2010

2009

2008

2007

2006

2005

2004

2003

2002

2001

2000

1999

1998

1997

1996

1995

1994

1993

1992

1991

1990

1989

1988

1987

1986

1985

1984

1983

1982

1981

1980

1979

50
60

70

Earnings Volatility (DES)

40
80
1979

2011

2010

2009

2008

2007

2006

2005

2004

2003

2002

2001

2000

1999

1998

1997

1996

1995

1994

1993

1992

1991

1990

1989

1988

1987

1986

1985

1984

1983

1982

1981

1980

60

Aged 30-34

50

70

Earnings Volatility (DES)

40
80

Figure 2.B10 Men’s Earnings Volatility by Seven Age Subgroups (DES for B)
Sample B; male

Aged 25-29

Aged 35-39

Year

Aged 40-44

Aged 45-49

Year

Aged 50-54

Aged 55-59

p90-p10

p90-p50

p50-p10

110

p90-p10

p90-p50

p90-p50

2011

p90-p50

2011

2011

p90-p50

2011

2009

2007

2005

2003

2001

p90-p50

2009

2007

2005

2003

2001

1999

2011

2009

2007

2005

2003

2001

1999

1997

60

80

80

100

100

p90-p50

2009

2007

2005

2003

2001

1999

1997

1995

40

60

2011

2009

2007

2005

2003

2001

1999

1997

1995

1993

1991

1989

1987

1985

1983

1981

1979

2011

2009

2007

2005

2003

2001

1999

1997

1995

1993

1991

1989

1987

1985

1983

1981

1979

20

20

40

40

60

60

80

80

100

100

p90-p50

2009

2007

2005

2003

p90-p10

2001

p90-p10

1999

p90-p10

1997

p90-p10

1997

1995

1993

1991

1989

1987

1985

1983

1981

1979

20

40

p90-p10

1999

Sample A

1993

1991

1989

1987

1985

1983

1981

1979

20 40 60 80 100 120

20

p90-p10

1997

Sample A

1995

Sample A

1995

80 100

Sample A

1995

60

2011

2009

2007

2005

2003

2001

1999

1997

1995

1993

1991

1989

1987

1985

1983

1981

1979

Sample A

1993

1991

40

20 40 60 80 100 120

Sample A

1993

1991

1989

1987

1985

1983

1981

1979

20

2011

2009

2007

2005

2003

2001

1999

1997

1995

1993

1991

1989

1987

1985

1983

1981

1979

2011

2009

2007

2005

2003

2001

1999

1997

1995

1993

1991

1989

1987

1985

1983

1981

1979

2011

2009

2007

2005

2003

2001

1999

1997

1995

1993

1991

1989

1987

1985

1983

1981

1979

0

0

50

50

100

100

150

150

Sample A

1993

1991

p50-p10

1989

100

p50-p10

1989

80

p50-p10

1987

60

80 100

p50-p10

1987

40

60

p50-p10

1985

1983

1981

1979

20

2011

40

p50-p10

1985

1983

1981

p90-p50

1979

20 40 60 80 100 120

p90-p50

2011

20

p90-p50

2009

2007

2005

2003

2001

1999

1997

1995

1993

1991

1989

1987

1985

1983

1981

1979

p90-p50

2011

100

p90-p50

2009

2007

2005

2003

80

p90-p50

2009

2007

2005

2003

p90-p10

2001

60

p90-p10

2001

40

p90-p10

1999

20

p90-p10

1997

1995

1993

1991

1989

1987

1985

1983

1981

1979

p90-p10

1999

20 40 60 80 100 120

p90-p10

1997

1995

1993

1991

1989

1987

1985

1983

1981

1979

Figure 2.B11 Men’s Earnings Volatility by Seven Age Subgroups (SS for A & B)

Aged 25-29
Sample B

Aged 30-34

p50-p10

Sample B

Aged 35-39

p50-p10

Sample B

Aged 40-44
p50-p10

Sample B

Aged 45-49
p50-p10

Sample B

Aged 50-54
p50-p10

Sample B

Aged 55-59
p50-p10

Sample B

p50-p10

No HS Graduation

1992

HS Graduate

111

2000

Some College

2011

2010

2009

2008

2007

60

70

80

90

100

Some College

2006

2005

2004

2003

2002

2001

50

HS Graduate

1999

1998

1997

1996

1995

1994

1993

40

No HS Graduation

1991

1990

1989

1988

1987

1986

1985

1984

1983

1982

1981

1980

1979

Earnings Volatility (DES)

2011

2010

2009

2008

2007

2006

2005

2004

2003

2002

2001

2000

1999

1998

1997

1996

1995

1994

1993

1992

1991

1990

1989

1988

1987

1986

1985

1984

1983

1982

1981

1980

1979

40

50

60

70

80

Earnings Volatility (DES)
90

100

Figure 2.B12 Women’s Earnings Volatility by Four Education Subgroups (DES for A)

Sample A; female

Year
College Graduate

Figure 2.B13 Women’s Earnings Volatility by Four Education Subgroups (DES for B)
Sample B; female

Year

College Graduate

p90-p10

p90-p50

p50-p10

112

p90-p10

p90-p50

2011

p90-p50

2009
2011

2009

p90-p50

2007

2005

2003

2001

1999

2011

2009

2007

2005

2003

2001

1999

1997

1995

p90-p50

2007

2005

2003

2001

p90-p10

1999

p90-p10

1997

1995

40 60 80 100120

p90-p10

1997

Sample A

1995

Sample A
1993

1991

1989

1987

1985

1983

1981

1979

2011

Sample A

1993

1991

1989

1987

40 60 80 100120140

2009

2007

40 60 80 100120

2011

2009

2007

2005

2003

2001

1999

1997

1995

1993

1991

1989

1987

1985

1983

1981

1979

2011

2009

2007

2005

2003

2001

1999

1997

1995

1993

1991

1989

1987

1985

1983

1981

1979

0

0

50

50

100 150

100 150

Sample A

1993

1991

1989

p50-p10
1985

1983

1981

1979

2011

p50-p10

1987

20 40 60 80 100120

2009

2005

2003

2001

1999

1997

1995

1993

1991

1989

1987

1985

1983

1981

1979

p50-p10

1985

1983

1981

1979

2011

p90-p50
2007

p90-p50

2009

2005

2003

2001

1999

40 60 80 100120140

p90-p50

2007

2005

2003

2001

p90-p10
1997

1995

1993

1991

1989

1987

1985

1983

1981

1979

p90-p10

1999

20 40 60 80 100120

p90-p10

1997

1995

1993

1991

1989

1987

1985

1983

1981

1979

Figure 2.B14 Women’s Earnings Volatility by Four Education Subgroups (SS for A & B)

No High School Graduation
Sample B

p50-p10

High School Graduate
Sample B

p50-p10

Some College
Sample B

p50-p10

College Graduate
Sample B

p50-p10

Year

113
2011

2010

2009

2008

2007

2006

2005

2004

2003

2002

2001

2000

1999

1998

1997

1996

1995

1994

1993

1992

1991

1990

1989

1988

1987

1986

1985

1984

1983

1982

1981

1980

1979

40

50

60

Earnings Volatility (DES)

30

2011

2010

2009

2008

2007

2006

2005

2004

2003

2002

2001

2000

1999

1998

1997

1996

1995

1994

1993

1992

1991

1990

1989

1988

1987

1986

1985

1984

1983

1982

1981

1980

1979

40

50

60

Earnings Volatility (DES)

30

2011

2010

2009

2008

2007

2006

2005

2004

2003

2002

2001

2000

1999

1998

1997

1996

1995

1994

1993

1992

1991

1990

1989

1988

1987

1986

1985

1984

1983

1982

1981

1980

1979

40

50

60

Earnings Volatility (DES)

30

Figure 2.B15 Women’s Earnings Volatility by Seven Age Subgroups (DES for A)
Sample A; female

Aged 25-29

Aged 30-34

Aged 35-39

Year

Aged 40-44

Aged 45-49

Year

Aged 50-54

Aged 55-59

Year

114
2011

2010

2009

2008

2007

2006

2005

2004

2003

2002

2001

2000

1999

1998

1997

1996

1995

1994

1993

1992

1991

1990

1989

1988

1987

1986

1985

1984

1983

1982

1981

1980

1979

60

70

80
90

Earnings Volatility (DES)

50

2011

2010

2009

2008

2007

2006

2005

2004

2003

2002

2001

2000

1999

1998

1997

1996

1995

1994

1993

1992

1991

1990

1989

1988

1987

1986

1985

1984

1983

1982

1981

1980

1979

60

70

80

90

Earnings Volatility (DES)

50

2011

2010

2009

2008

2007

2006

2005

2004

2003

2002

2001

2000

1999

1998

1997

1996

1995

1994

1993

1992

1991

1990

1989

1988

1987

1986

1985

1984

1983

1982

1981

1980

1979

80

Aged 25-29

70

Aged 30-34

60

90

Earnings Volatility (DES)

50

Figure 2.B16 Women’s Earnings Volatility by Seven Age Subgroups (DES for B)
Sample B; female

Aged 35-39

Year

Aged 40-44

Aged 45-49

Year

Aged 50-54

Aged 55-59

p90-p10

p90-p50

p50-p10

115

p90-p10

p90-p50

p90-p50

2011

p90-p50

2011

2011

p90-p50

2011

2009

2007

2005

2003

2001

p90-p50

2009

2007

2005

2003

2001

1999

2011

2009

2007

2005

2003

2001

1999

1997

p90-p50

2009

2007

2005

2003

2001

1999

1997

1995

1993

1991

1989

1987

1985

1983

1981

1979

40 60 80 100 120 140

2011

2009

2007

2005

2003

2001

1999

1997

1995

1993

1991

1989

1987

1985

1983

1981

1979

2011

2009

2007

2005

2003

2001

1999

1997

1995

1993

1991

1989

1987

1985

1983

1981

1979

0

0

50

50

100 150 200

100 150 200

p90-p50

2009

2007

2005

2003

p90-p10

2001

p90-p10

1999

p90-p10

1997

1995

p90-p10

1999

Sample A

1993

1991

1989

1987

1985

1983

1981

1979

20 40 60 80 100 120

40 60 80 100 120 140

p90-p10

1997

100

2011

2009

2007

2005

2003

2001

1999

1997

1995

1993

1991

1989

1987

1985

1983

1981

1979

p90-p10

1997

Sample A

1995

80

Sample A

1995

60

20 40 60 80 100 120

Sample A

1995

40

2011

2009

2007

2005

2003

2001

1999

1997

1995

1993

1991

1989

1987

1985

1983

1981

1979

Sample A

1993

1991

1989

1987

1985

1983

1981

1979

20

100

Sample A

1993

1991

100

80

2011

2009

2007

2005

2003

2001

1999

1997

1995

1993

1991

1989

1987

1985

1983

1981

1979

2011

2009

2007

2005

2003

2001

1999

1997

1995

1993

1991

1989

1987

1985

1983

1981

1979

50 100 150 200 250

50 100 150 200 250

Sample A

1993

1991

p50-p10

1989

80

p50-p10

1989

60

60

p50-p10

1987

40

40

p50-p10

1985

1983

1981

1979

20

20

p50-p10

1987

80 100

2011

2009

2007

2005

2003

2001

1999

1997

1995

1993

1991

1989

1987

1985

1983

1981

1979

p50-p10

1985

1983

60

p90-p50

1981

40

100

p90-p50

1979

20

80

p90-p50

2011

60

p90-p50

2011

40

p90-p50

2009

2007

2005

2003

2001

1999

20

p90-p50

2009

2007

p90-p10

1997

1995

1993

1991

1989

1987

1985

1983

1981

1979

p90-p10

2005

80 100

p90-p10

2003

60

p90-p10

2001

40

p90-p10

1999

20

p90-p10

1997

1995

1993

1991

1989

1987

1985

1983

1981

1979

Figure 2.B17 Women’s Earnings Volatility by Seven Age Subgroups (SS for A & B)
Aged 25-29
Sample B

Aged 30-34

p50-p10

Sample B

Aged 35-39

p50-p10

Sample B

Aged 40-44
p50-p10

Sample B

Aged 45-49
p50-p10

Sample B

Aged 50-54
p50-p10

Sample B

Aged 55-59
p50-p10

Sample B

p50-p10

APPENDIX C

SUPPLEMENTAL TABLES FOR CHAPTER 2

116

Table 2.C1 Match Rate
Male
Female
Survey
Possible to be
Matched
Match Rate
Possible to be
Matched
Match Rate
Year
matched
Number
matched
Number
1979
11,055
7,814
71%
14,048
10,255
73%
1980
13,203
9,648
73%
16,665
12,622
76%
1981
12,992
8,474
65%
16,314
10,983
67%
1982
12,180
8,732
72%
15,628
11,524
74%
1983
12,334
8,543
69%
15,527
11,203
72%
1984
12,155
8,286
68%
15,534
10,900
70%
1985
N/A
N/A
N/A
N/A
N/A
N/A
1986
12,389
8,273
67%
15,352
10,625
69%
1987
12,111
8,332
69%
15,186
10,713
71%
1988
12,345
7,872
64%
14,855
9,683
65%
1989
11,194
7,823
70%
13,863
9,910
71%
1990
12,475
8,496
68%
15,207
10,726
71%
1991
12,141
8,296
68%
14,605
10,340
71%
1992
12,453
8,515
68%
14,977
10,495
70%
1993
11,996
6,258
52%
14,862
7,868
53%
1994
10,561
5,520
52%
12,839
6,856
53%
1995
N/A
N/A
N/A
N/A
N/A
N/A
1996
9,816
6,907
70%
11,955
8,555
72%
1997
10,198
7,049
69%
12,271
8,714
71%
1998
10,289
7,182
70%
12,414
8,896
72%
1999
9,894
6,954
70%
11,877
8,460
71%
2000
9,216
7,106
77%
10,759
8,442
78%
2001
9,527
6,544
69%
11,426
8,202
72%
2002
15,473
7,963
51%
18,706
9,555
51%
2003
15,152
8,019
53%
18,629
9,934
53%
2004
14,833
7,064
48%
18,239
8,549
47%
2005
14,774
7,305
49%
18,192
9,122
50%
2006
15,591
7,823
50%
19,022
9,578
50%
2007
15,162
7,644
50%
18,471
9,417
51%
2008
15,135
7,820
52%
18,334
9,539
52%
2009
15,530
8,129
52%
18,737
9,645
51%
2010
15,349
7,766
51%
18,665
9,259
50%
2011
14,751
7,447
50%
17,710
8,878
50%
Note: “Possible to be matched” means the total number of men at risk of being matched and satisfying criteria for the
main sample suggested in Section 2.2.2, i.e., whose annual earnings last year belonged to 1%-99% among male
earners 25-59, not self-employed, not a student, with positive annual earnings and positive weeks worked.

117

Table 2.C2 Summary Statistics for Matched Men
Possible to be matched
Matched
Mean Real
College
Mean Real
College
Survey
Mean
Mean
Annual
Graduates
Unemployed
Annual
Graduates
Unemployed
Year
Age
Age
Earnings
%
Earnings
%
1979
46,356
39.6
22%
11%
48,873
40.8
22%
10%
1980
46,981
39.5
22%
12%
49,453
40.5
23%
9%
1981
45,946
39.2
23%
14%
49,165
40.3
24%
12%
1982
44,686
39.3
23%
16%
47,269
40.4
24%
14%
1983
42,247
39.0
24%
20%
45,179
40.2
24%
18%
1984
41,984
39.0
24%
18%
45,249
40.2
25%
15%
1985
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
1986
43,620
39.0
24%
14%
46,387
40.1
25%
12%
1987
44,117
39.2
25%
14%
47,487
40.4
26%
12%
1988
44,174
39.0
25%
13%
47,031
40.1
26%
11%
1989
44,816
39.1
26%
11%
47,979
40.2
27%
9%
1990
44,296
39.2
25%
11%
47,812
40.4
27%
9%
1991
43,256
39.2
25%
13%
46,821
40.4
27%
11%
1992
41,504
39.3
28%
15%
44,798
40.4
29%
13%
1993
41,513
39.5
29%
14%
44,835
40.5
31%
12%
1994
41,829
39.7
30%
13%
45,640
40.7
33%
12%
1995
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
1996
44,172
40.0
29%
11%
47,640
41.1
30%
10%
1997
44,644
40.1
30%
10%
47,640
41.3
31%
9%
1998
46,685
40.3
30%
9%
50,636
41.5
32%
8%
1999
46,796
40.6
31%
8%
50,254
41.9
34%
7%
2000
48,368
41.0
32%
7%
50,796
42.0
33%
6%
2001
49,194
41.1
31%
7%
53,853
42.4
34%
6%
2002
49,939
40.9
32%
9%
53,548
42.4
35%
8%
2003
49,033
41.3
32%
10%
51,678
42.8
35%
9%
2004
48,677
41.4
33%
10%
51,840
42.7
36%
8%
2005
50,093
41.4
33%
8%
52,134
42.9
35%
8%
2006
51,446
41.7
33%
7%
55,104
43.1
35%
7%
2007
52,681
41.7
34%
7%
55,011
43.3
36%
7%
2008
51,955
41.7
35%
8%
55,017
43.1
37%
7%
2009
50,584
41.8
34%
11%
52,690
43.2
35%
10%
2010
47,106
41.8
35%
15%
49,734
43.2
37%
13%
2011
45,799
41.9
35%
14%
48,269
43.2
37%
13%
Note 1: “Possible to be matched” means the total number of men at risk of being matched and satisfying criteria for the main
sample suggested in Section 2.2.2, i.e., whose annual earnings last year belonged to 1%-99% among male earners 25-59,
not self-employed, not a student, with positive annual earnings and positive weeks worked.
Note 2: “Unemployed” refers to individuals who worked less than equal to 49 weeks, and were laid off or looked for jobs in the
non-working period for each year.

118

Table 2.C3 Summary Statistics for Matched Women
Possible to be matched
Matched
Mean Real
College
Mean Real
College
Survey
Mean
Mean
Annual
Graduates
Unemployed
Annual
Graduates
Unemployed
Year
Age
Age
Earnings
%
Earnings
%
1979
13,644
40.0
14%
9%
13,621
41.1
14%
8%
1980
15,200
39.9
15%
9%
15,078
40.8
16%
8%
1981
15,177
39.8
15%
10%
15,110
40.8
15%
9%
1982
14,919
39.7
16%
11%
15,092
40.7
16%
10%
1983
15,194
39.5
17%
12%
15,306
40.6
17%
11%
1984
15,884
39.6
18%
11%
16,346
40.7
18%
9%
1985
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
1986
17,332
39.2
18%
10%
17,705
40.2
19%
9%
1987
17,975
39.4
19%
10%
18,373
40.4
19%
9%
1988
19,243
39.3
20%
9%
19,601
40.3
20%
7%
1989
19,427
39.2
20%
9%
20,062
40.3
21%
7%
1990
20,143
39.4
21%
8%
20,837
40.4
22%
7%
1991
20,365
39.4
21%
9%
21,127
40.3
22%
9%
1992
20,550
39.4
24%
10%
21,472
40.4
25%
8%
1993
20,817
39.7
25%
9%
21,760
40.6
26%
8%
1994
21,508
40.0
27%
10%
22,548
40.9
29%
9%
1995
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
1996
22,079
40.0
28%
9%
23,064
41.0
30%
8%
1997
22,504
40.5
29%
8%
23,843
41.5
31%
7%
1998
22,983
40.5
29%
7%
24,222
41.5
31%
7%
1999
23,841
40.9
30%
7%
25,402
42.0
32%
6%
2000
25,353
41.3
32%
6%
25,956
42.2
33%
5%
2001
25,159
41.2
31%
6%
26,405
42.3
33%
5%
2002
25,349
40.8
33%
7%
27,484
42.4
34%
6%
2003
24,872
41.0
33%
7%
26,908
42.7
35%
7%
2004
25,512
41.1
35%
7%
28,370
42.7
38%
6%
2005
25,203
41.2
34%
7%
26,991
42.8
36%
6%
2006
26,241
41.6
35%
6%
28,559
43.3
38%
5%
2007
26,762
41.6
37%
6%
28,752
43.3
38%
5%
2008
27,343
41.7
38%
6%
29,352
43.2
40%
5%
2009
27,311
41.7
38%
8%
29,126
43.4
40%
7%
2010
26,342
41.4
39%
10%
28,469
42.9
42%
9%
2011
26,096
41.5
40%
10%
28,242
43.0
42%
9%
Note 1: “Possible to be matched” means the total number of women at risk of being matched and satisfying criteria for the main
sample suggested in Section 2.2.2, i.e., whose annual earnings last year belonged to 1%~99% among female earners 2559, not self-employed, not a student, with positive annual earnings and positive weeks worked.
Note 2: “Unemployed” refers to individuals who worked less than equal to 49 weeks, and were laid off or looked for jobs in the
non-working period for each year.

119

Table 2.C4 Earnings Volatility for Men and Women

1978-1979

DG
for Sample A
0.36 (.009)

Men
DES
for Sample A
31.3 (.52)

DES
for Sample B
42.9 (.79)

DG
for Sample A
0.68 (.013)

Women
DES
for Sample A
52.9 (.53)

DES
for Sample B
74.6 (.55)

1979-1980

0.37 (.007)

32.4 (.46)

46.7 (.70)

0.65 (.011)

50.8 (.49)

74.7 (.50)

1980-1981

0.35 (.008)

31.0 (.47)

46.3 (.76)

0.61 (.011)

48.6 (.51)

73.1 (.55)

1981-1982

0.43 (.008)

36.7 (.48)

53.0 (.69)

0.65 (.013)

50.0 (.53)

73.3 (.54)

1982-1983

0.46 (.010)

37.9 (.55)

55.6 (.73)

0.66 (.014)

49.9 (.55)

71.7 (.56)

1983-1984

0.42 (.009)

35.7 (.54)

53.4 (.77)

0.63 (.013)

48.4 (.57)

72.0 (.58)

1984-1985

N/A

N/A

N/A

N/A

N/A

N/A

1985-1986

0.41 (.009)

34.9 (.52)

51.8 (.77)

0.61 (.012)

48.2 (.54)

71.5 (.58)

1986-1987

0.38 (.010)

32.8 (.51)

49.4 (.77)

0.60 (.012)

47.4 (.51)

71.8 (.57)

1987-1988

0.38 (.009)

32.5 (.54)

47.4 (.81)

0.56 (.011)

45.6 (.52)

70.8 (.61)

1988-1989

0.38 (.009)

32.6 (.54)

46.2 (.80)

0.56 (.011)

44.9 (.53)

70.0 (.61)

1989-1990

0.37 (.008)

32.5 (.48)

46.2 (.76)

0.53 (.010)

43.7 (.50)

66.3 (.62)

1990-1991

0.41 (.009)

34.9 (.52)

50.4 (.74)

0.53 (.011)

43.3 (.54)

66.6 (.63)

1991-1992

0.40 (.009)

34.6 (.52)

53.7 (.76)

0.54 (.011)

43.4 (.54)

67.0 (.63)

1992-1993

0.41 (.011)

35.1 (.64)

50.9 (.91)

0.57 (.013)

45.0 (.64)

66.5 (.73)

1993-1994

0.39 (.010)

34.5 (.59)

51.7 (.96)

0.54 (.015)

43.6 (.69)

67.4 (.80)

1994-1995

N/A

N/A

N/A

N/A

N/A

N/A

1995-1996

0.39 (.009)

33.8 (.53)

49.4 (.84)

0.54 (.012)

43.8 (.57)

65.6 (.70)

1996-1997

0.38 (.008)

33.6 (.50)

52.7 (.83)

0.50 (.012)

41.2 (.58)

65.6 (.71)

1997-1998

0.36 (.008)

32.4 (.48)

49.5 (.84)

0.52 (.013)

42.1 (.57)

68.0 (.69)

1998-1999

0.37 (.008)

32.9 (.50)

49.4 (.85)

0.47 (.010)

40.0 (.55)

66.5 (.72)

1999-2000

0.37 (.008)

33.5 (.47)

49.8 (.85)

0.50 (.012)

41.5 (.58)

64.6 (.74)

2000-2001

0.37 (.010)

32.8 (.51)

50.2 (.90)

0.51 (.013)

41.8 (.59)

65.1 (.75)

2001-2002

0.39 (.008)

34.4 (.49)

53.4 (.80)

0.51 (.011)

41.7 (.58)

68.0 (.67)

2002-2003

0.39 (.008)

34.8 (.47)

56.3 (.79)

0.52 (.012)

42.4 (.57)

65.2 (.68)

2003-2004

0.38 (.009)

33.7 (.50)

53.1 (.86)

0.48 (.011)

40.1 (.59)

66.3 (.74)

2004-2005

0.38 (.010)

33.5 (.49)

52.7 (.85)

0.50 (.011)

41.5 (.54)

64.1 (.72)

2005-2006

0.36 (.007)

32.9 (.43)

52.0 (.82)

0.48 (.010)

40.2 (.54)

64.9 (.71)

2006-2007

0.37 (.007)

33.6 (.45)

52.3 (.82)

0.47 (.010)

40.1 (.50)

64.1 (.72)

2007-2008

0.38 (.008)

34.3 (.45)

55.0 (.79)

0.50 (.011)

41.1 (.57)

65.0 (.70)

2008-2009

0.45 (.009)

38.5 (.51)

61.9 (.71)

0.46 (.010)

39.2 (.53)

65.5 (.70)

2009-2010

0.42 (.009)

36.6 (.52)

63.0 (.79)

0.47 (.011)

39.7 (.57)

66.6 (.72)

0.46 (.011)

39.3 (.57)

64.4 (.75)

Year

2010-2011
0.42 (.011)
35.9 (.56)
61.6 (.83)
Note: The numbers in parentheses indicate asymptotic standard errors.

120

Table 2.C5 Men’s Earnings Volatility by Education and Age (DES for Sample B)
Men's Education Subgroups

Men's Age Subgroups

Year

HS Grad. or Less

Some College

College Grad.

Age 25-34

Age 35-49

Age 50-59

1978-1979

46.83 (1.01)

40.49 (1.81)

31.97 (1.65)

39.96 (1.17)

41.21 (1.30)

48.67 (1.65)

1979-1980

50.83 (0.87)

44.36 (1.71)

34.53 (1.50)

49.87 (1.10)

41.12 (1.12)

49.85 (1.48)

1980-1981

50.96 (0.95)

44.81 (1.88)

32.92 (1.60)

49.15 (1.17)

41.83 (1.28)

48.57 (1.55)

1981-1982

57.16 (0.85)

50.21 (1.66)

42.39 (1.59)

57.09 (1.09)

47.85 (1.11)

54.39 (1.48)

1982-1983

61.47 (0.88)

50.95 (1.83)

41.13 (1.59)

59.73 (1.15)

49.83 (1.17)

57.92 (1.50)

1983-1984

59.20 (0.96)

50.40 (1.86)

39.88 (1.59)

54.94 (1.19)

48.41 (1.22)

59.14 (1.63)

1984-1985

N/A

N/A

N/A

N/A

N/A

N/A

1985-1986

57.53 (0.97)

49.77 (1.80)

36.49 (1.56)

49.21 (1.24)

49.21 (1.19)

60.01 (1.61)

1986-1987

54.51 (1.02)

46.98 (1.75)

38.57 (1.51)

51.27 (1.29)

43.77 (1.19)

56.48 (1.59)

1987-1988

52.31 (1.04)

45.55 (1.87)

35.51 (1.60)

49.26 (1.34)

42.33 (1.20)

53.86 (1.76)

1988-1989

51.46 (1.06)

42.34 (1.78)

36.95 (1.56)

45.38 (1.25)

44.25 (1.23)

51.39 (1.78)

1989-1990

50.60 (1.03)

44.45 (1.73)

37.31 (1.43)

46.35 (1.28)

44.18 (1.13)

50.23 (1.75)

1990-1991

55.64 (0.98)

45.90 (1.71)

42.02 (1.42)

51.21 (1.29)

45.10 (1.09)

60.10 (1.55)

1991-1992

60.52 (1.04)

53.39 (1.56)

39.81 (1.43)

53.42 (1.33)

49.30 (1.11)

63.52 (1.59)

1992-1993

57.81 (1.24)

47.70 (1.95)

40.90 (1.70)

53.88 (1.52)

46.70 (1.33)

55.93 (2.04)

1993-1994

58.76 (1.38)

52.57 (2.00)

39.29 (1.62)

52.94 (1.67)

49.37 (1.39)

55.39 (2.17)

1994-1995

N/A

N/A

N/A

N/A

N/A

N/A

1995-1996

53.70 (1.25)

48.58 (1.77)

42.64 (1.42)

51.94 (1.53)

44.11 (1.17)

57.23 (1.87)

1996-1997

58.98 (1.18)

48.06 (1.80)

45.10 (1.45)

53.08 (1.61)

51.27 (1.17)

55.53 (1.74)

1997-1998

53.55 (1.28)

48.31 (1.75)

43.57 (1.38)

52.29 (1.58)

46.82 (1.19)

51.70 (1.81)

1998-1999

55.22 (1.25)

50.70 (1.81)

39.31 (1.41)

50.54 (1.59)

46.52 (1.24)

53.78 (1.69)

1999-2000

55.30 (1.30)

48.85 (1.78)

42.07 (1.35)

48.63 (1.58)

47.72 (1.22)

54.93 (1.71)

2000-2001

52.61 (1.38)

49.51 (1.80)

47.22 (1.52)

49.55 (1.77)

47.89 (1.27)

54.87 (1.76)

2001-2002

56.31 (1.24)

51.24 (1.65)

50.88 (1.34)

56.14 (1.57)

49.24 (1.13)

57.97 (1.57)

2002-2003

60.76 (1.18)

56.16 (1.58)

50.28 (1.35)

59.88 (1.60)

52.46 (1.13)

59.58 (1.48)

2003-2004

55.39 (1.35)

58.40 (1.76)

46.27 (1.41)

56.03 (1.63)

48.87 (1.27)

57.05 (1.64)

2004-2005

56.22 (1.28)

52.67 (1.79)

48.16 (1.43)

55.12 (1.62)

51.23 (1.22)

52.88 (1.65)

2005-2006

54.64 (1.25)

52.68 (1.62)

47.97 (1.41)

56.32 (1.66)

47.09 (1.19)

55.43 (1.49)

2006-2007

58.23 (1.27)

48.69 (1.75)

46.56 (1.32)

57.64 (1.62)

47.29 (1.21)

54.71 (1.51)

2007-2008

59.31 (1.24)

57.42 (1.67)

47.70 (1.27)

61.54 (1.60)

50.02 (1.16)

56.33 (1.44)

2008-2009

67.22 (0.98)

61.40 (1.47)

54.45 (1.31)

66.40 (1.34)

60.17 (1.02)

60.39 (1.37)

2009-2010

69.79 (1.13)

63.93 (1.67)

52.98 (1.36)

65.02 (1.57)

61.21 (1.19)

63.78 (1.39)

68.25 (1.52)

58.41 (1.30)

60.24 (1.53)

2010-2011
67.62 (1.21)
62.90 (1.73)
52.95 (1.44)
Note: The numbers in parentheses indicate asymptotic standard errors.

121

Table 2.C6 Women’s Earnings Volatility by Education and Age (DES for Sample B)
Women's Education Subgroups

Women's Age Subgroups

Year

HS Grad. or Less

Some College

College Grad.

Age 25-34

Age 35-49

Age 50-59

1978-1979

74.82 (0.67)

76.57 (1.32)

71.27 (1.44)

87.02 (0.52)

71.25 (0.96)

61.66 (1.35)

1979-1980

75.10 (0.60)

76.64 (1.18)

70.44 (1.32)

85.21 (0.55)

73.07 (0.83)

60.89 (1.26)

1980-1981

73.40 (0.66)

77.30 (1.20)

66.69 (1.53)

83.18 (0.65)

69.69 (0.95)

62.64 (1.30)

1981-1982

74.71 (0.64)

72.59 (1.31)

67.27 (1.46)

82.57 (0.66)

72.91 (0.85)

59.31 (1.36)

1982-1983

72.43 (0.70)

75.28 (1.20)

64.44 (1.43)

83.22 (0.66)

68.78 (0.91)

57.92 (1.43)

1983-1984

72.59 (0.72)

73.75 (1.35)

68.11 (1.40)

80.16 (0.76)

71.54 (0.93)

59.65 (1.40)

1984-1985

N/A

N/A

N/A

N/A

N/A

N/A

1985-1986

73.20 (0.71)

71.86 (1.27)

64.94 (1.45)

80.73 (0.74)

68.45 (0.93)

61.47 (1.46)

1986-1987

74.55 (0.69)

71.67 (1.25)

61.99 (1.46)

79.70 (0.78)

69.72 (0.91)

62.95 (1.36)

1987-1988

74.26 (0.74)

68.08 (1.44)

61.51 (1.50)

76.23 (0.90)

67.26 (0.96)

68.39 (1.45)

1988-1989

72.75 (0.76)

70.96 (1.38)

60.92 (1.44)

80.12 (0.79)

63.92 (0.99)

65.01 (1.51)

1989-1990

68.38 (0.79)

66.72 (1.35)

60.01 (1.45)

73.00 (0.93)

63.49 (0.95)

60.73 (1.50)

1990-1991

70.02 (0.78)

66.54 (1.38)

56.57 (1.49)

73.79 (0.95)

64.92 (0.95)

57.45 (1.53)

1991-1992

71.30 (0.82)

66.40 (1.35)

57.52 (1.37)

75.39 (0.93)

63.49 (0.97)

60.40 (1.54)

1992-1993

71.11 (0.98)

66.82 (1.44)

56.28 (1.57)

77.82 (1.05)

61.29 (1.11)

59.69 (1.77)

1993-1994

71.28 (1.10)

69.61 (1.59)

57.83 (1.59)

73.84 (1.28)

65.76 (1.17)

61.04 (1.95)

1994-1995

N/A

N/A

N/A

N/A

N/A

N/A

1995-1996

70.43 (0.96)

65.64 (1.44)

56.82 (1.38)

75.06 (1.08)

61.01 (1.05)

62.20 (1.61)

1996-1997

70.80 (0.97)

62.36 (1.53)

59.14 (1.37)

72.05 (1.23)

64.88 (1.02)

58.12 (1.58)

1997-1998

73.54 (0.95)

67.19 (1.38)

59.34 (1.38)

76.24 (1.10)

65.34 (1.03)

63.32 (1.55)

1998-1999

72.26 (1.01)

64.02 (1.45)

59.20 (1.39)

73.66 (1.19)

63.68 (1.07)

63.87 (1.56)

1999-2000

68.99 (1.08)

64.92 (1.50)

57.93 (1.36)

73.05 (1.28)

62.29 (1.09)

59.82 (1.55)

2000-2001

70.40 (1.08)

63.67 (1.54)

58.40 (1.35)

75.01 (1.24)

63.21 (1.08)

58.44 (1.63)

2001-2002

71.20 (1.01)

69.57 (1.30)

62.36 (1.20)

79.87 (1.03)

64.83 (1.00)

61.20 (1.41)

2002-2003

69.15 (1.02)

65.77 (1.39)

59.83 (1.20)

75.44 (1.15)

62.18 (1.01)

60.32 (1.39)

2003-2004

68.98 (1.15)

68.47 (1.51)

61.90 (1.26)

76.88 (1.24)

64.60 (1.09)

58.20 (1.53)

2004-2005

65.84 (1.17)

67.95 (1.35)

59.41 (1.22)

73.80 (1.29)

63.41 (1.05)

56.12 (1.43)

2005-2006

70.01 (1.07)

65.60 (1.44)

58.58 (1.21)

76.52 (1.21)

63.69 (1.03)

56.37 (1.42)

2006-2007

68.99 (1.13)

65.23 (1.43)

57.68 (1.21)

73.03 (1.28)

63.74 (1.06)

56.53 (1.41)

2007-2008

68.22 (1.12)

65.18 (1.44)

61.55 (1.14)

73.12 (1.28)

64.68 (1.04)

58.40 (1.34)

2008-2009

70.43 (1.06)

66.26 (1.42)

59.48 (1.18)

71.68 (1.28)

65.53 (1.04)

60.45 (1.32)

2009-2010

70.51 (1.13)

69.15 (1.51)

61.37 (1.16)

71.30 (1.33)

66.63 (1.06)

62.49 (1.37)

71.39 (1.33)

64.10 (1.17)

58.29 (1.41)

2010-2011
67.97 (1.23)
65.16 (1.55)
60.52 (1.20)
Note: The numbers in parentheses indicate asymptotic standard errors.

122

APPENDIX D

ADDITIONS FOR CHAPTER 2

123

Matching March CPS
In the Current Population Survey (CPS), an interviewee is interviewed for 4 consecutive
months, once he or she enters the sample, and then has an 8-month interim without interviews.
After the interim, he or she is interviewed again for 4 consecutive months, then leaves the sample.
By exploiting this rotating property of CPS, two-year short panel data sets are constructed. The
household identification number, house number, and line number in CPS are used mainly for
matching individuals across the two adjacent years. This “naïve” matching method usually
generates a number of incorrect matches, and Madrian & Lefgren (2000) suggest several
methods to reduce false matches by using additional variables. I follow their recommendations
and drop matched individuals if there is an inconsistency in gender, race, or age.
Theoretically, about 50% of the individuals in March Supplement of CPS for each year
should be matched with data from the following year. However, as previously explained, the
survey does not track interviewees who moved elsewhere, so the real matching rate tends to be
lower than the full matching rate.
Table 2.C1 lists the match rate for each survey year. That rate is calculated by dividing the
number of matched individuals by the total number of individuals who theoretically can be
matched among those satisfying the sample selection criteria of this study. 33 From 1979 to 2001,
for both male and female interviewees, the match rate ranges from 64% to 78%, usually staying
around 70%, except in 1993 and 1994. 34 Since 2002, the overall match rate falls again and
remains around 50% for both men and women, varying from 47% to 53% through 2011. 35
33

More specifically, it satisfies the criteria for the sample B. See Section 2.2.2.
The match rate went down to 52-53% in 1993 and 1994 because the March CPS was redesigned substantially
since 1994 and the sampling method for 1995 March CPS was significantly different from those used in other years.
35
The decrease in the match rate since 2002 was also due to the redesign of the sampling for March CPS at that
time: a considerable number of additional samples were added to more precisely estimate statistics for minority
groups, but these additional samples did not have panel property, so they were not able to be matched across years.
See Appendix in Celik et al. (2012).
34

124

Accordingly, it is expected that the matched sample contains a higher portion of individuals
who are relatively unlikely to relocate. This will cause the matched sample to be somewhat
different from the general population. In Table 2.C2 and Table 2.C3 we can see that the
individuals in the matched sample tend to have higher earnings, older ages, a larger probability
of being college graduates, and fewer chances to experience unemployment than those in the
entire sample, which includes all individuals who could potentially be matched. Consequently, it
is hard to exclude the possibility that the estimated earnings volatility might be biased in the
matched March CPS sample. 36
As discussed in Section 2.2.1, however, this study tracks the dynamics of earnings volatility.
It does not focus on estimating the exact level of earnings volatility each year. Thus, if there was
no systemic bias in the matching process, the matched March CPS still can be considered a
legitimate data set for exploring the evolution of earnings volatility. As for the sample periods
prior to the Great Recession, 1979-2007, a number of researchers argued that no serious systemic
bias was found in the matching process for the March CPS. (Cameron and Tracy 1998; Celik et
al. 2012; Ziliak, Hardy, and Bollinger 2011)
One relevant concern which still needs to be considered, however, is whether any serious
shock influenced the matching process since 2008, due to the Great Recession. For instance, if
shock from the Great Recession significantly affected the proportion of movers among the
March CPS interviewees, we cannot regard the estimated earnings volatility after the Great
Recession as comparable with that during preceding periods.
To check this possibility, I drew Figure 2.B1, plotting the match rate in Table 2.C1. Figure
2.B1 shows that, for both men and women, the match rate stays rather stable even after 2008-

36

Considering that there is a positive correlation between the unemployment rate and earnings volatility, as
discussed in Section 2.3.3, earnings volatility for the matched sample from CPS seems likely underestimated.

125

2009. Thus, it does not seem that the Great Recession caused serious change in the percentage of
movers among March CPS interviewees.

Decomposition of Var(Δlog(hours))
Var(∆log(ℎ𝑜𝑜𝑜𝑠𝑖𝑖 )) = E((∆ log(ℎ𝑜𝑜𝑜𝑠𝑖𝑖 ))2 ) − {E(∆ log(ℎ𝑜𝑜𝑜𝑠𝑖𝑖 ))}2
= 𝑆unemp,t E((∆ log(ℎ𝑜𝑜𝑜𝑠𝑖𝑖 ))2 |𝑖 ∈ unemp)

+𝑆nounemp,t E((∆ log(ℎ𝑜𝑜𝑜𝑠𝑖𝑖 ))2 |𝑖 ∈ nounemp)

−{E(∆ log(ℎ𝑜𝑜𝑜𝑠𝑖𝑖 ))}2

where ℎ𝑜𝑜𝑜𝑠𝑖𝑖 indicates total annual working hours of the individual i at year t; 𝑆j,t stands for

the share of group j out of sample A at year t; “unemp” means individuals who had worked for

less than or equal to 49 weeks, and also had experienced unemployed status in their non-working
periods in the year t-1 or t; “nounemp” refers to other individuals in sample A who do not belong
to “unemp” group.
Note that E(∆ log(ℎ𝑜𝑜𝑜𝑠𝑖𝑖 )) is also practically zero because ∆ log(ℎ𝑜𝑜𝑜𝑠𝑖𝑖 ) is basically the

age-adjusted residual for each year t. Thus, the formula above is rendered to:

Var(∆ log(ℎ𝑜𝑜𝑜𝑠𝑖𝑖 )) ≈ 𝑆unemp,t E((∆ log(ℎ𝑜𝑜𝑜𝑠𝑖𝑖 ))2 |𝑖 ∈ unemp)
+𝑆nounemp,t E((∆ log(ℎ𝑜𝑜𝑜𝑠𝑖𝑖 ))2 |𝑖 ∈ nounemp)

126

REFERENCES

127

REFERENCES

Baker, Michael, and Gary Solon. 2003. “Earnings Dynamics and Inequality among Canadian
Men, 1976-1992: Evidence from Longitudinal Income Tax Records.” Journal of Labor
Economics 21(2): 267–88.
Blundell, Richard, Luigi Pistaferri, and Ian Preston. 2008. “Consumption Inequality and Partial
Insurance.” American Economic Review 98(5): 1887–1921.
Cameron, Stephen, and Joseph Tracy. 1998. “Earnings Variability in the United States: An
Examination Using Matched-CPS Data.” Columbia University and Federal Reserve Bank of
New York.
Celik, Sule, Chinhui Juhn, Kristin McCue, and Jesse Thompson. 2012. “Recent Trends in
Earnings Volatility: Evidence from Survey and Administrative Data.” The B.E. Journal of
Economic Analysis & Policy 12(2).
Congressional Budget Office. 2007. “Trends in Earnings Variability over the Past 20 Years.”
Dynan, Karen, Douglas Elmendorf, and Daniel Sichel. 2012. “The Evolution of Household
Income Volatility.” The B.E. Journal of Economic Analysis and Policy 12(2): 1–42.
Dynarski, Susan, Jonathan Gruber. 1997. “Can Families Smooth Variable Earnings ?” Brookings
Papers on Economic Activity 1997(1): 229–303.
Elsby, Michael W. L., Bart Hobijn, and Ayşegül Şahin. 2011. “The Labor Market in the Great
Recession—An Update to September 2011 [with Comment and Discussion].” Brookings
Papers on Economic Activity (2011): 353–84.
Gorbachev, Olga. 2014. “Did Household Consumption Become More Volatile?” American
Economic Review 101(5): 2248–70.
Gottschalk, Peter, and Robert Moffitt. 1994. “The Growth of Earnings Instability in the U. S.
Labor Market.” Brookings Papers on Economic Activity 1994(2): 217–72.
Haider, Steven J. 2001. “Earnings Instability and Earnings Inequality of Males in the United
States: 1967 – 1991.” Journal of Labor Economics 19(4): 799–836.
Hardy, Bradley, and James P. Ziliak. 2014. “Decomposing Trends in Income Volatility: The
‘Wild Ride’ at the Top and bottomDecomposing Trends In Income Volatility: The ‘Wild
Ride’ At The Top And Bottom.” Economic Inquiry 52(1): 459–76.

128

Larrimore, Jeff, Richard V. Burkhauser, Shuaizhang Feng, and Laura Zayatz. 2008. “Consistent
Cell Means for Topcoded Incomes in the Public Use March CPS ( 1976 - 2007 ).” Journal
of Economic and Social Measurement 33(2): 89–128.
Madrian, Brigitte C., and Lars John Lefgren. 2000. “An Approach to Longitudinally Matching
Current Population Survey ( CPS ) Respondents.” Journal of Economic and Social
Measurement 26(1): 31–62.
Moffitt, Robert A., and Peter Gottschalk. 2002. “Trends in the Transitory Variance of Earnings
in the United States.” Economic Journal 112(478): C68–73.
———. 2012. “Trends in the Transitory Variance of Male Earnings: Methods and Evidence.”
Journal of Human Resources 47(1): 204–36.
Shin, Donggyun, and Gary Solon. 2011. “Trends in Men’s Earnings Volatility: What Does the
Panel Study of Income Dynamics Show?” Journal of Public Economics 95(7): 973–82.
Stephens, Melvin. 2002. “Worker Displacement and the Added Worker Effect.” Journal of Labor
Economics 20(3): 504–37.
Ziliak, James P, Bradley Hardy, and Christopher Bollinger. 2011. “Earnings Volatility in
America: Evidence from Matched CPS.” Labour Economics 18(6): 742–54.

129

CHAPTER 3
How Did Men’s Earnings Volatility Change
during the Great Recession? A Comparison
of Evidence between the PSID and CPS
3.1 Introduction
This paper explores the evolution of men’s earnings volatility in the United States during the
Great Recession by using the Panel Study of Income Dynamics (PSID). I further compare the
findings with those of Koo (2016), who examines the same topic with the matched March
Current Population Survey (CPS).1
The major motivations of this study are twofold. First, given the paucity of studies on men’s
earnings volatility during the Great Recession, which was the severest downturn in the post-war
era (Elsby et al. 2011), I document the relevant facts that are robustly supported by the datasets
PSID and the CPS, both of which have been extensively used in the literature to explore men’s
earnings volatility in the United States. Second, according to previous studies, the PSID
generates a divergent trend in men’s earnings volatility from the CPS, particularly for the period
of the early 2000s (Shin and Solon 2011; Celik et al. 2012; Dynan et al. 2012). There have been
no further explorations, however, of the factors driving the inconsistent results between the two
1

Notwithstanding an extensive body of literature documenting changes in men’s earnings volatility until the time
prior to the Great Recession (Cameron and Tracy 1998; Celik et al. 2012; Office 2007; Dynan et al. 2012; Haider
2001; Moffitt and Gottschalk 2002; Moffitt and Gottschalk 2012; Shin and Solon 2011), there are only a limited
number of studies on men’s earnings volatility during the Great Recession. To my knowledge, Koo (2016) is the
first study focusing on the evolution of earnings volatility during the Great Recession in the United States by using
the matched March Current Population Survey (CPS).

130

datasets. This study not only looks at whether such inconsistency between the two datasets has
persisted through the Great Recession, but also sheds light on the source of that inconsistency, by
comparing the results from multiple samples across the two datasets, the PSID and the CPS.
The two main findings of this study are as follows. First, the PSID results are fairly
comparable to the CPS for the evolution of men’s earnings volatility during the Great Recession.
Specifically, the PSID shows that there was a substantial increase in men’s earnings volatility
during and after the Great Recession. Furthermore, the PSID indicates that the substantial
increase in men’s earnings volatility during the Great Recession as well as other previous
recessions was mainly driven by the increase in earnings volatility of unemployed workers.
These findings are largely consistent with those of Koo (2016) from the matched March CPS.
Second, in the sample excluding zero-earners and the top and bottom 1 % of positive-earners,
the PSID generates fairly comparable results with the CPS for the evolution of men’s earnings
volatility from 1980 to 2012. However, in the sample including those zero-earners and the
outliers, the PSID generates much higher estimates for men’s earnings volatility compared to
what the CPS generates, particularly for the period 1993-2004. This partly explains why previous
studies using the PSID report different trends from those of other studies using different datasets,
such as the CPS, for the same period. Specifically the former studies document high and upwardtrended men’s earnings volatility between the late 1990s and mid 2000s, whereas the latter ones
find relatively low and flat trends in men’s earnings volatility for the same period (Shin and
Solon 2011; Celik et al. 2012; Dynan et al. 2012).
In section 3.2, I explain the dataset, sample selection criteria, and measures for earnings
volatility. Section 3.3 presents the main results, and the conclusion is provided in Section 3.4.

131

3.2 Data and Measurement
The PSID is one of the longest panel datasets which has collected various family income data
and other extensive socioeconomic and demographic information on national representative
samples in the U.S. Among a number of family income sources contained in the PSID, annual
earnings from wage and salaries for each family head is the one which has been recorded the
most consistently since 1970. Thus, I mainly use this variable to calculate men’s earnings
volatility over the last four decades.
One point to be considered is that the PSID has started to conduct its surveys biennially since
1997, while it had collected information annually from 1968 to 1996. Because of this
inconsistency in survey frequency, I use a two-year difference in each individual annual earnings
instead of a one-year difference in order to consistently track changes in earnings volatility, as
Shin and Solon (2011) do. Specifically, I measure the earnings volatility for the period 19691971, by using the information on annual earnings contained in the 1970 and 1972 waves of the
PSID. Similarly, the earnings volatility for the period 1970-1972 is measured by using the 1971
and 1973 waves of the PSID. I keep doing this by using every currently available wave of the
PSID. Consequently, I use all usable two-year differences in individual annual earnings as key
variables for measuring earnings volatility from 26 overlapping periods , (1969-1971, 1970-1972,
1971-1973, … ,1993-1995, and 1994-1996), and 8 non-overlapping periods, (1996-1998, 19982000, … , and 2010-2012) from the 1970-2013 waves of the PSID.I employ the same sample
selection criteria as Koo’s (2016) to maximize the comparability of this study to his. Specifically,
the basic sample (hereafter, Sample A) consists of only male heads whose age ranged from 25 to
59 with positive annual earnings from wage and salaries for each year, and whose main job was
neither that of a student nor self-employed. To reduce potential measurement errors, Sample A

132

does not include ones with imputed earnings, and it further excludes the ones whose annual
earnings belonged to the top and bottom 1 % of positive earnings in each year.
One primary purpose of this paper is to update changes in men’s earnings volatility in the
Great Recession from the PSID. Thus, excluding all zero- and low-earners from the sample
would be too restrictive to fully capture the impact of the Great Recession on men’s earnings
variability, as Koo (2016) points out. Hence, as Koo (2016) does, I also include in analyses
another sample encompassing zero-earners and the top and bottom 1 % of positive earners for
each year in addition to Sample A (hereafter, Sample B).
Accordingly, I have a total of 48,800 and 56,000 observations for Sample A and B,
respectively, over 40 years. In terms of average sample size per year on a two-year difference, I
have 1,440 and 1,560 observations for Sample A and B, respectively. More detailed sample size
and summary statistics for Sample A and B by year are presented in Table 3.B1.
I use three different approaches to robustly measure changes in earnings volatility over the
last four decades, including the Great Recession. Those approaches are also employed by Koo
(2016) so that I may directly compare my results with his.
The first measure I use is the standard deviation of the two-year difference in log real
earnings residuals, whose usage was initiated by Dynarski and Gruber (1997) (henceforth
denoted as DG). The specific formula of DG is:
�Var�log(𝑌𝚤𝚤 ) �
− log(𝑌𝚤𝚤−2 )�

, where 𝑌𝑖𝑖 stands for real earnings for individual 𝑖 at year 𝑡; �log(𝑌𝚤𝚤 ) �
− log(𝑌𝚤𝚤−2 )� indicates

the residual from a regression of (log(𝑌𝑖𝑖 ) − log(𝑌𝑖𝑖−2 )) on quadratics of age for individual 𝑖 at

year 𝑡. Hence, this residual refers to two-year difference in log real earnings which is
unexplained by cohort effects, life-cycle effects, and aggregate time effects.

133

Shin and Solon (2011) show that this measure DG captures well the variances of transitory
and permanent earnings shocks while being little affected by the growth of the return to an
individual’s time-invariant characteristics. Moreover, it does not depend on any complicated
earnings model with a number of assumptions. Thus, it is a relatively simple and transparent
measure for earnings volatility.
One weakness of DG, however, is that it cannot be applicable to Sample B, including zeroearners, because it is calculated based on log real earnings. Hence, I employ another measure
suggested by Dynan, Elmendorf, and Sichel (2012) (hereafter DES), which is the standard
deviation of the age-adjusted arc percentage change in real earnings. The specific formula of
DES is:
�Var �100 ∗

(𝑌𝚤𝚤 �
− 𝑌𝚤𝚤−2 )
ďż˝
𝑌�𝚤

,where 𝑌𝑖𝑖 indicates real earnings (not log real earnings) for individual 𝑖 in year 𝑡; 𝑌�𝑖 refers to

the average of 𝑌𝑖𝑖 and 𝑌𝑖𝑖−2 ;

(𝑌𝚤𝚤�
−𝑌𝚤𝚤−2 )
𝑌�𝚤

means the age-adjusted arc percentage change in real

earnings, which is residual from the regression of
Thus,

(𝑌𝚤𝚤�
−𝑌𝚤𝚤−2 )
𝑌�𝚤

(𝑌𝑖𝑖 −𝑌𝑖𝑖−2 )
𝑌�𝑖

on quadratics of age for year 𝑡 − 2.

refers to arc percentage change in real earnings which is unexplained by cohort

effects, life cycle effects, and aggregate time effects for individual 𝑖 in year 𝑡.

It is notable that DES has upper and lower bounds (200 and -200, respectively) by its

construction, which would alleviate potential problems from outliers showing extreme changes
in earnings. Compared to DG, in addition, DES has the merit of enabling easier and more
intuitive interpretation, as well as being able to encompass zero-earners in analyses.

134

I also employ another measure applicable to both Sample A and B, which is motivated by
Shin and Solon (2011) (hereafter SS): this measure constructs quantile distributions for the tenth,
twenty-fifth, fiftieth, seventy-fifth, and ninetieth sample percentiles of the age-adjusted arc
percentage change in real earnings,

(𝑌𝚤𝚤�
−𝑌𝚤𝚤−2 )
𝑌�𝚤

, for each sample year. Compared to the previous

measures, DG and DES, that generate one summary statistics for the dispersion of changes in
earnings for each year, SS has the advantage of letting us know the whole distributional changes
in earnings differences over the sample years. Moreover, it also produces alternative measures
summarizing the dispersion of change in earnings for each year, such as the difference between
the ninetieth and tenth percentiles, so that we may compare its evolution with the trends in other
measures such as DG and DES.

3.3 Results
Figure 3.A1 shows how, in Sample A, men’s earnings volatility evolved from 1971 to 2012
based on the PSID. For comparison, I put together Koo's (2016) corresponding results from the
matched March CPS. The first noteworthy feature in Figure 3.A1 is that both measures, DG and
DES, generate similar dynamic patterns of men’s earnings volatility, even with the difference in
their scales, for both the PSID and CPS. The second feature is that both the PSID and CPS
datasets generate a similar evolution of men’s earnings volatility during the period 1980-2012 in
terms of its magnitudes as well as dynamic patterns, although the former is calculated based on a
two-year difference in (log) real earnings while the latter is based on a one-year difference in
them.

135

Specifically, men’s earnings volatility for Sample A calculated from the PSID shows a
dramatic increase during 1970s and hits its peak during the recession in the early 1980s. Then, in
general, it stabilized at lower levels afterwards until the time prior to the Great Recession,
although it showed temporary, relatively mild upsurges during two other recessions, in the early
1990s and 2000. After the Great Recession began in 2008, men’s earnings volatility showed
another historic upswing, and remained at relatively high levels until the early 2010s. This
pattern confirms the well-documented counter-cyclical property of earnings volatility, which is
consistent with Koo's (2016) findings from the CPS during the same period since 1980, as seen
in the same figure.
The similar results for Sample A from both the PSID and CPS are a novel finding, because
Celik et al. (2012) show that, for a comparable sample, the matched March CPS generates a
different trend in men’s earnings volatility during the early 2000s compared to what Shin and
Solon (2011) find with the PSID. Specifically, while Shin and Solon (2011) find a resuming
upward trend during the early 2000s, Celik et al. (2012) find no such evidence from the matched
March CPS as well as other datasets such as SIPP (Survey of Income and Program Participation)
and LEHD (Longitudinal Employer-Household Dynamics).
One main reason why I find more comparable results between the PSID and CPS than Celik
et al. (2012) do is that my findings with the PSID somewhat differ from that of Shin and Solon
(2011) with the same data; I find a more moderate increase in men’s earnings volatility during
the early 2000s’ which also ended earlier than what Shin and Solon (2011) find over the same
period. This difference is partly because I use a different sample selection from theirs; unlike
their main sample, Sample A here excludes students and self-employed people, although Sample
A is still largely comparable to Shin and Solon’s sample. Moreover, I use sample weights when

136

estimating earnings volatility in order to closely follow Koo’s (2016) study for comparison,
whereas Shin and Solon (2011) do not.
Another important reason for the different comparison results between this study and Celik et
al. (2012) is that I use a different rule of time matching from that of Celik et al. (2012), to
compare the results across the PSID and CPS. For example, Celik et al. (2012) compare the
earnings volatility for 1978-1979 from the CPS with that for 1976-1978 from the PSID, that for
1979-1980 from the CPS with that for 1977-1979 from the PSID, and so on for the subsequent
periods. Unlike them, I compare the earnings volatility for 1978-1979 from the CPS with that for
1977-1979 from the PSID, that for 1979-1980 from the CPS with that for 1978-1980 from the
PSID, and so on for the subsequent periods. 2 The former way of time matching used by Celik et
al. (2012) turns out to make the discrepancy of the results between the PSID and CPS look more
salient compared to that of this study.
Figure 3.A2, in turn, shows men’s earnings volatility for Sample B from PSID (circle-shaped
points) and compares it with the previous result for Sample A (square-shaped points) and the
corresponding results of Koo (2016) from the CPS (triangle- and diamond- shaped points).
Sample B, as explained earlier, encompasses zero-earners and the 1 % top and bottom positiveearners in addition to Sample A. In general, Figure 3.A2 shows that the PSID also generates a
comparable trend in men’s earnings volatility for Sample B compared to the counterpart from the
CPS. To be more specific, I find much higher earnings volatility with a more salient upward
trend for Sample B compared to that for Sample A from the PSID over all of the sample years.
Moreover, the PSID shows a substantial upsurge of men’s earnings volatility during the Great
Recession for Sample B, as it does for Sample A. All these features in men’s earnings volatility

2

See Figure 1 in Celik et al. (2012).

137

for Sample B are comparably found in the results from the CPS reported by Koo (2016) put in
the same figure.
In terms of how close the estimates are to each other, in comparing the PSID and CPS for
each sample, however, larger disparities are observed in Sample B than Sample A. For instance,
the earnings volatility for Sample B calculated from the PSID shows overall a higher level
during the period prior to 2005 compared to that calculated from the CPS, whereas we see fairly
comparable results for Sample A from both datasets over the same period. In particular, men’s
earnings volatility for Sample B in the PSID appears to be much higher from 1993 to 2004, than
that from the CPS. Accordingly, for Sample B, the PSID shows a higher level of men’s earnings
volatility during the mild recession in the early 2000s even compared to that during the Great
Recession. This finding differs from Koo's (2016) based on the CPS; he finds an increase in
men’s earnings volatility for Sample B during the early 2000’s, but a much moderate one
compared to that during the Great Recession, which is also consistent with the results from
Sample A that he and I find using the CPS and the PSID, respectively.
I employ another measure SS in Figure 3.A3 to see if all the findings in Figure 3.A2 are
consistently found even with an alternative measure. 3 Specifically, I identify the tenth, twentyfifth, fiftieth , seventy-fifth, and ninetieth percentiles of arc-percent change in men’s real earnings
for each year for both Sample A and B from the PSID, and then look at how those percentiles
change over the last four decades for both samples. For the sake of brevity, Figure 3.A3 shows
only some summary results—the differences in ninetieth and tenth (p90-p10), fiftieth and tenth
(p50-p10), and ninetieth and fiftieth (p90-p50) percentiles for both samples—rather than

As discussed earlier, SS tracks the whole distributional changes in earnings differences, rather than
generates one statistics summarizing the dispersion of change in earnings for each sample year as DG and
DES do.
3

138

displaying the full trajectories of percentile changes. 4 I also put together Koo's (2016)
corresponding results from the CPS for comparison in Figure 3.A3, as I do in the previous
figures.
Based on the evolution of p90-p10 in the top panels of Figure 3.A3, the PSID shows that
there was an increase in men’s earnings volatility during the Great Recession in Sample A, and
the increase stands out more clearly in Sample B. In addition, the evolution of p90-p10 shows
counter-cyclical movements along the sample period for both Sample A and B. All these
observations are comparable to the findings from the PSID based on DES in Figure 3.A2.
In the lower panels of Figure 3.A3, the PSID further shows that the counter-cyclical property
of men’s earnings volatility was mainly driven by the counter-cyclicality of p50-p10 for both
samples. In other words, men tended to go through severe earnings losses during recessions, and
that was the main reason that made men’s earnings paths more unstable in the Great Recession as
well as in other recessions compared to other non-recession periods.
All the time trends in p90-p10, p50-p10, and p90-p50 from the PSID in Figure 3.A3
described above are largely consistent with what Koo (2016) correspondingly finds from the
CPS for both Sample A and B (see the graphs with circle-shaped points in the same figure).
However, it is also apparent that the discrepancy of the results between the PSID and CPS
becomes more prominent for Sample B than Sample A, as seen in Figure 3.A2. For Sample B,
for example, the PSID reports a larger upsurge in p50-p10 during the period 1993-2004 than the
CPS does, while for Sample A, both datasets show fairly comparable results during the same
period (compare panel e and f in Figure 3.A3). Accordingly, for Sample B, the PSID shows a
more substantial increase in p90-p10 during the recession in the early 2000s than the CPS does.

4

The figure for the whole change of the percentiles from 1971 to 2012 is available on request.

139

The fact that the PSID and the CPS shows more divergent trends in men’s earnings volatility
in Sample B than in Sample A is notable in two respects. First, this result is consistent with the
findings of previous studies using the PSID to explore men’s earnings volatility until the time
prior to the Great Recession: The inclusion of zero-earners and outliers to the analysis in the
PSID makes the increasing trend in men’s earnings volatility more salient during the early 2000s
(Shin and Solon 2011; Dynan et al. 2012). Second, this observation implies that the zero-earners
and outliers added to Sample A from the PSID would have different characteristics related to
higher earning volatility than the counterpart from the CPS, particularly for the 1994-2004 period.
The conjecture above is well corroborated by the results in Figure 3.A4, where I decompose
men’s earnings volatility for each year into that of the unemployed workers and that of notunemployed ones. 5 Figure 3.A4 first shows that for both Sample A and B from the PSID, the
counter-cyclical behavior of men’s earnings volatility from the early 1980s through the Great
Recession largely stemmed from the considerable increases in earnings volatility among
unemployed male workers rather than not-unemployed ones. This finding is fairly consistent
with what Koo (2016) finds with the CPS, as seen in the same figure. The most salient exception
observed in Figure 3.A4, however, is that the substantial upswing in men’s earnings volatility for
Sample B during the recession in the early 2000s from the PSID was largely driven by that of
non-unemployed workers. More specifically, during the period 1994-2004, the trend in “not
unemployed” men’s earnings volatility for Sample B from the PSID becomes much higher than
the counterpart from the CPS (see panel f in Figure 3.A4). As a result, the PSID shows a historic
increase in men’s earnings volatility for Sample B during the early 2000s, which is the most
different result of the PSID compared to those from the CPS.

Unemployed workers are defined here as ones having experienced unemployment in either of two years,
while not-unemployed ones indicate workers having experienced no unemployment in both years.
5

140

It is not clear why including zero-earners and outliers in addition to Sample A with the PSID
leads to higher men’s earnings volatility during the 1993-2004 period compared to that from the
CPS. One possibility is that the zero-earners and outliers during the period 1993-2004 in the
PSID were likely to contain a relatively large numbers of errors in terms of their earnings
compared to those in the CPS; another possibility is that this result relates to the fact that the
PSID began a severe reform in its data collection process in the mid 1990s. Although I do not
pursue directly where such difference came from for that period between the PSID and CPS in
this paper, it seems requisite, at least, to interpret with caution the evolution of men’s earnings
volatility for Sample B when using the PSID, particularly for the period 1993-2004.

3.4 Conclusion
I explore the evolution of men’s earnings volatility over the last four decades including the
period of the Great Recession by using the PSID. I show men’s earnings volatility went up
substantially after the Great Recession began, which confirms the well-documented countercyclicality of earnings volatility. I further find the counter-cyclicality of earnings volatility was
mainly experienced by the unemployed workers. All these findings turn out to be largely
comparable with what Koo (2016) finds from the CPS, although there are some discrepancies
observed in the analyses for Sample B including zero-earners and outliers for the period 19932004. The discrepancies between the PSID and CPS imply that the zero-earners and outliers
added to Sample B for the period 1993-2004 in the PSID are likely to contain a relatively large
number of errors in their reported earnings. This issue needs to be more rigorously explored in

141

future research, given that the PSID has been one of most extensively used panel datasets in the
literature for numerous social/economic topics as well as the topic of earnings volatility.

142

APPENDICES

143

APPENDIX A

FIGURES FOR CHAPTER 3

144

1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012

25

.2

30

.25

35

.3

DES

DG

40

.35

45

.4

50

.45

Figure 3.A1 Men’s Earnings Volatility for Sample A (DG and DES)

DG for Sample A from PSID
DG for Sample A from CPS

DES for Sample A from PSID
DES for Sample A from CPS

145

1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012

25

30

35

40

DES
45
50

55

60

65

Figure 3.A2 Earnings Volatility for Sample A and B (DES)

DES for Sample A from PSID
DES for Sample A from CPS

DES for Sample B from PSID
DES for Sample B from CPS

146

20

40

40

60

60

80

80

100

40

60

60

80

80

100

PSID

PSID

100

c. p90-p50

e. p50-p10
1971
1973
1975
1977
1979
1981
1983
1985
1987
1989
1991
1993
1995
1997
1999
2001
2003
2005
2007
2009
2011

20

40

PSID

100

1971
1973
1975
1977
1979
1981
1983
1985
1987
1989
1991
1993
1995
1997
1999
2001
2003
2005
2007
2009
2011

20

1971
1973
1975
1977
1979
1981
1983
1985
1987
1989
1991
1993
1995
1997
1999
2001
2003
2005
2007
2009
2011

1971
1973
1975
1977
1979
1981
1983
1985
1987
1989
1991
1993
1995
1997
1999
2001
2003
2005
2007
2009
2011
20

20

40

40

60

60

80

80

100

100

a. p90-p10

1971
1973
1975
1977
1979
1981
1983
1985
1987
1989
1991
1993
1995
1997
1999
2001
2003
2005
2007
2009
2011

1971
1973
1975
1977
1979
1981
1983
1985
1987
1989
1991
1993
1995
1997
1999
2001
2003
2005
2007
2009
2011

20

Figure 3.A3 Earnings Volatility for Sample A and B (SS)
Sample A
Sample B

CPS

CPS

CPS

147

b. p90-p10

PSID

PSID

PSID

CPS

d. p90-p50

CPS

f. p50-p10

CPS

DES^2
500 1000 1500

PSID

PSID
0 10002000 30004000

c. Unemployed

e. Not unemployed
1971
1973
1975
1977
1979
1981
1983
1985
1987
1989
1991
1993
1995
1997
1999
2001
2003
2005
2007
2009
2011

DES^2
500 1000 1500

PSID

0 1000 20003000 4000

1971
1973
1975
1977
1979
1981
1983
1985
1987
1989
1991
1993
1995
1997
1999
2001
2003
2005
2007
2009
2011

0

1971
1973
1975
1977
1979
1981
1983
1985
1987
1989
1991
1993
1995
1997
1999
2001
2003
2005
2007
2009
2011

1971
1973
1975
1977
1979
1981
1983
1985
1987
1989
1991
1993
1995
1997
1999
2001
2003
2005
2007
2009
2011

DES^2
500 1000 1500

0 10002000 30004000

0

a. All

1971
1973
1975
1977
1979
1981
1983
1985
1987
1989
1991
1993
1995
1997
1999
2001
2003
2005
2007
2009
2011

1971
1973
1975
1977
1979
1981
1983
1985
1987
1989
1991
1993
1995
1997
1999
2001
2003
2005
2007
2009
2011

0

Figure 3.A4 Decomposition of Earnings Volatility: Unemployed vs. Not unemployed
Sample A
Sample B

CPS

CPS

CPS

148

b. All

PSID

PSID

PSID

CPS

d. Unemployed

CPS

f. Not Unemployed

CPS

APPENDIX B

SUPPLEMENTAL TABLE FOR CHAPTER 3

149

Table 3.B1 Descriptive Statistics for Sample A and B
Year

Number

1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1998
2000
2002
2004
2006
2008
2010
2012

1,005
1,008
1,044
1,118
1,136
1,168
1,203
1,217
1,267
1,296
1,329
1,333
1,324
1,344
1,359
1,377
1,405
1,414
1,437
1,429
1,455
1,397
1,382
1,408
1,593
1,595
1,628
1,891
1,912
1,885
1,954
1,932
1,832
1,756

Sample A
Unemp. Age
21%
22%
19%
22%
20%
23%
20%
18%
17%
20%
20%
26%
24%
24%
22%
19%
18%
17%
17%
16%
19%
17%
17%
15%
14%
13%
12%
13%
13%
11%
10%
9%
12%
12%

42.0
41.9
41.7
41.5
41.5
41.2
41.2
41.1
41.2
41.1
41.0
40.8
40.7
40.5
40.7
40.5
40.4
40.4
40.3
40.5
40.5
40.9
41.2
41.1
41.6
41.7
42.3
42.6
43.1
43.1
43.6
43.8
44.1
43.9

Earnings
(2013$)
63,360
64,267
62,267
60,295
60,466
62,268
62,450
61,120
58,041
57,333
57,931
58,800
58,801
61,551
62,287
63,113
62,565
61,877
60,982
60,230
61,162
61,329
63,415
63,391
62,084
63,122
66,929
66,047
64,041
63,667
67,248
68,772
66,681
67,670

Number

Unemp.

1,056
1,062
1,108
1,178
1,208
1,242
1,278
1,316
1,358
1,389
1,428
1,437
1,430
1,449
1,452
1,476
1,502
1,514
1,537
1,535
1,562
1,499
1,502
1,546
1,748
1,769
1,792
2,068
2,142
2,111
2,137
2,107
2,069
2,006

21%
22%
19%
22%
20%
24%
21%
19%
17%
21%
20%
26%
25%
25%
23%
20%
18%
18%
17%
16%
19%
17%
19%
16%
15%
14%
13%
13%
14%
12%
10%
10%
15%
16%

150

Sample B
Age
Earnings
(2013$)
42.3
60,754
42.1
61,818
42.0
59,252
41.8
57,643
41.7
57,260
41.6
58,905
41.4
59,347
41.4
57,345
41.5
55,055
41.4
54,135
41.3
54,224
41.2
54,792
41.0
54,442
41.0
57,281
41.0
58,560
40.8
59,039
40.8
58,760
40.7
57,924
40.7
56,986
40.8
56,224
41.0
57,131
41.2
57,308
41.5
58,607
41.4
57,978
42.0
57,107
42.1
57,685
42.7
61,639
43.0
60,570
43.5
57,465
43.7
59,408
43.9
61,396
44.2
62,838
44.4
58,737
44.4
58,834

Fraction of
Zero-earners
3.3%
3.0%
4.4%
3.6%
4.0%
4.4%
3.7%
4.4%
4.2%
4.6%
5.0%
5.2%
5.7%
5.0%
4.3%
4.9%
4.7%
4.9%
5.0%
5.4%
5.4%
5.7%
6.3%
6.9%
6.4%
7.5%
6.3%
6.7%
9.2%
7.4%
7.7%
7.1%
10.8%
10.5%

REFERENCES

151

REFERENCES

Cameron, Stephen, and Joseph Tracy. 1998. “Earnings Variability in the United States: An
Examination Using Matched-CPS Data.” Columbia University and Federal Reserve Bank of
New York.
Celik, Sule, Chinhui Juhn, Kristin McCue, and Jesse Thompson. 2012. “Recent Trends in
Earnings Volatility: Evidence from Survey and Administrative Data.” The B.E. Journal of
Economic Analysis & Policy 12(2).
Congressional Budget Office. 2007. “Trends in Earnings Variability over the Past 20 Years.”
Dynan, Karen, Douglas Elmendorf, and Daniel Sichel. 2012. “The Evolution of Household
Income Volatility.” The B.E. Journal of Economic Analysis and Policy 12(2): 1–42.
Dynarski, Susan, Jonathan Gruber. 1997. “Can Families Smooth Variable Earnings ?” Brookings
Papers on Economic Activity 1997(1): 229–303.
Elsby, Michael W. L., Bart Hobijn, and Ayşegül Şahin. 2011. “The Labor Market in the Great
Recession—An Update to September 2011 [with Comment and Discussion].” Brookings
Papers on Economic Activity (2011): 353–84.
Haider, Steven J. 2001. “Earnings Instability and Earnings Inequality of Males in the United
States: 1967 – 1991.” Journal of Labor Economics 19(4): 799–836.
Koo, Kyong Hyun. 2016. "The Evolution of Earnings Volatility During and After the Great
Recession." Industrial Relations, 55(4), pp.705–732.
Moffitt, Robert A., and Peter Gottschalk. 2002. “Trends in the Transitory Variance of Earnings
in the United States.” Economic Journal 112(478): C68–73.
———. 2012. “Trends in the Transitory Variance of Male Earnings: Methods and Evidence.”
Journal of Human Resources 47(1): 204–36.
Shin, Donggyun, and Gary Solon. 2011. “Trends in Men’s Earnings Volatility: What Does the
Panel Study of Income Dynamics Show?” Journal of Public Economics 95(7): 973–82.

152