Children of the Great Recession
Irwin Garfinkel, Sara McLanahan, and Christopher Wimer
EDITORS
Children of the
Great Recession
Children of the
Great Recession
G G G
Irwin Garfinkel,
Sara McLanahan,
Christopher Wimer,
Editors
Russell Sage Foundation
New York
The Russell Sage Foundation
The Russell Sage Foundation, one of the oldest of America’s general purpose foundations, was established in 1907 by Mrs. Margaret Olivia Sage for “the improvement of
social and living conditions in the United States.” The foundation seeks to fulfill this
mandate by fostering the development and dissemination of knowledge about the country’s political, social, and economic problems. While the foundation endeavors to assure
the accuracy and objectivity of each book it publishes, the conclusions and interpretations in Russell Sage Foundation publications are those of the authors and not of the
foundation, its trustees, or its staff. Publication by Russell Sage, therefore, does not imply
foundation endorsement.
BOARD OF TRUSTEES
Sara S. McLanahan, Chair
Larry M. Bartels
Karen S. Cook
W. Bowman Cutter III
Sheldon Danziger
Kathryn Edin
Lawrence F. Katz
David Laibson
Nicholas Lemann
Martha Minow
Peter R. Orszag
Claude M. Steele
Shelley E. Taylor
Richard H. Thaler
Hirokazu Yoshikawa
Library of Congress Cataloging-in-Publication Data
Names: Garfinkel, Irwin, editor. | McLanahan, Sara, editor. | Wimer,
Christopher, editor.
Title: Children of the great recession / Irwin Garfinkel, Sara McLanahan, and
Christopher Wimer, editors.
Description: New York : Russell Sage Foundation, 2016. | Includes
bibliographical references and index.
Identifiers: LCCN 2016002195 | ISBN 9781610448598 (ebook)
Subjects: LCSH: Recessions—United States. | Global Financial Crisis,
2008-2009. | Child welfare—United States. | Families—United States. |
Parenting—United States.
Classification: LCC HB3743 .C624 2016 | DDC 330.973/0931—dc23 LC record available
at http://cp.mcafee.com/d/FZsS738Acy1J5xYsUOYUYMrKrjKqenPhOCYYCqejqtPhOqekTzhOyMrjKqenPhOCYYOyrhhKUqen6m7AjqKNJVZgl6CAvU02rzjifY01dSh
gIzXz_nVZAQsLCzBVzHTbFIFIsztMQszDT7eEyCJtdmWb7axVZicHs3jq9J4T
vAn3hOYyyODtUTsS03fJq77RJN6FD4XEKnKCyqenPtGlr1hZrshGpdILIz
zoA6xoQg8rfjh0Xm9Ewd78VV5N6X33PVEVudFK6Rn1P
Copyright © 2016 by Russell Sage Foundation. All rights reserved. Printed in the United
States of America. No part of this publication may be reproduced, stored in a retrieval
system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, without the prior written permission of the publisher.
Reproduction by the United States Government in whole or in part is permitted for any
purpose.
The paper used in this publication meets the minimum requirements of American
National Standard for Information Sciences—Permanence of Paper for Printed Library
Materials. ANSI Z39.48-1992.
Text design by Suzanne Nichols.
RUSSELL SAGE FOUNDATION
112 East 64th Street, New York, New York 10065
10 9 8 7 6 5 4 3 2 1
Contents
List of Illustrations
vii
Contributors
xi
Chapter 1
Introduction
Irwin Garfinkel, Sara McLanahan,
and Christopher Wimer
1
Chapter 2
Economic Well-Being
Irwin Garfinkel and Natasha Pilkauskas
31
Chapter 3
Public and Private Transfers
Natasha Pilkauskas and Irwin Garfinkel
58
Chapter 4
Mothers’ and Fathers’ Health
Janet Currie and Valentina Duque
88
Chapter 5
Parents’ Relationships
Daniel Schneider, Sara McLanahan,
and Kristen Harknett
118
Chapter 6
Nonresident Father Involvement
Ronald B. Mincy and Elia De la Cruz Toledo
149
Chapter 7
Mothers’ and Fathers’ Parenting
William Schneider, Jane Waldfogel,
and Jeanne Brooks-Gunn
173
Chapter 8
Child Well-Being
William Schneider, Jane Waldfogel,
and Jeanne Brooks-Gunn
206
Index
228
List of Illustrations
Figure 1.1Median Household Income Index and
Unemployment Rate
Figure 1.2
Local Unemployment Rates During
Interviewing Periods
Figure 2.1
Maternal Employment
Figure 2.2
Paternal Employment
Figure 2.3
Household Income ($2010)
Figure 2.4
Big Gains and Losses
Figure 2.5
Poverty Rates
Figure 2.6
Hardship (Insecurity) Rates
Figure 2.7
Household Income ($2010) by Race-Ethnicity
Figure 2.8
Employment by Education
Figure 2.9
Income by Education
Figure 2.10 Income by Race-Ethnicity and Relationship Status
Figure 2.11 Poverty Rate by Education
Figure 2.12 Hardship by Education
Figure 3.1
Public Assistance Receipt by Child’s Age-Year
Figure 3.2
Public Assistance Receipt by Education
Figure 3.3
Average Dollar Value of Public Assistance Benefits
Figure 3.4
Private Financial Transfers ($2010)
Figure 3.5
Doubling Up
Figure 3.6
Average Dollar Value of Private Assistance
Figure 3.7
Private Financial Transfers and Doubling Up
Figure 3.8
Public Transfer Receipt Rates
Figure 3.9
Effects of Transfers on Household Income
Figure 3.10 Mitigating Effects of Transfers on Poverty
Figure 4.1
Mothers’ Health Status Is Fair or Poor
Figure 4.2
Fathers’ Health Status Is Fair or Poor
Figure 4.3
Mothers’ Health Problem that Limits Work
Figure 4.4
Fathers’ Health Problem that Limits Work
Figure 4.5
Mothers’ Binge Drinking
Figure 4.6
Fathers’ Binge Drinking
Figure 4.7
Mothers’ Drug Use
Figure 4.8
Fathers’ Drug Use
Figure 4.9
Effects of a Recession on Mothers’ Health Status
Figure 4.10 Effects of a Recession on Fathers’ Health Status
Figure 4.11 Effects of a Recession on Mothers’ Health Problem
that Limits Work
Figure 4.12 Effects of a Recession on Fathers’ Health Problem
that Limits Work
Figure 4.13 Effects of a Recession on Mothers’ Binge Drinking
Figure 4.14 Effects of a Recession on Fathers’ Binge Drinking
6
9
35
36
37
38
39
39
40
42
43
43
44
44
62
63
64
65
65
66
68
69
71
71
91
92
92
93
94
94
95
95
97
97
98
98
99
100
viii
list of illustrations
Figure 4.15
Figure 4.16
Figure 5.1
Figure 5.2
Figure 5.3
Figure
Figure
Figure
Figure
Figure
Figure
Figure
Figure
5.4
5.5
5.6
5.7
5.8
5.9
5.10
5.11
Figure
Figure
Figure
Figure
5.12
5.13
5.14
5.15
Figure 5.16
Figure
Figure
Figure
Figure
Figure
Figure
Figure
Figure
Figure
Figure
6.1
6.2
6.3
6.4
6.5
6.6
6.7
6.8
7.1
7.2
Figure 7.3
Figure 7.4
Figure 7.5
Figure 7.6
Figure 7.7
Figure 7.8
Figure 7.9
Effects of a Recession on Mothers’ Drug Use
Effects of a Recession on Fathers’ Drug Use
Mothers’ Relationship Status
Marriage to Bio Fathers or New Partners
Marriage or Cohabitation to Bio Fathers
or New Partners
Mothers’ Reports of Bio Fathers’ Supportiveness
Fathers’ Reports of Bio Mothers’ Supportiveness
Mothers’ Reports of New Partners’ Supportiveness
Mothers’ Reports of Relationship with Bio Father
Fathers’ Reports of Relationship with Bio Mother
Mothers’ Marriage and Marriage or Cohabitation
Mothers’ Marriage (Bio Father or New Partner)
Mothers’ Marriage or Cohabitation (Bio Father
or New Partner)
Mothers’ Reports of Bio Fathers’ Supportiveness
Fathers’ Reports of Mothers’ Supportiveness
Mothers’ Reports of New Partners’ Supportiveness
Mothers’ Reports of Quality of Relationship
with Bio Father Fathers’ Reports of Quality of Relationship
with Bio Mother
Nonresidence Status
Father Engagement
Child Support and Visitation
Formal Child Support per Year
Informal Child Support per Year
In-Kind Child Support per Year
Visitation Days per Month
Share of Nonresident Fathers Visiting Their Children
High-Frequency Maternal Spanking by Education
High-Frequency Maternal Physical
Aggression by Education
High-Frequency Maternal Psychological
Aggression by Education
High-Frequency Maternal Warmth by Education
High-Frequency Maternal Parenting
Activities by Education
High-Frequency Paternal Spanking by Education
High-Frequency Paternal Physical
Aggression by Education
High-Frequency Paternal Psychological
Aggression by Education
High-Frequency Maternal Spanking by
Unemployment Rate
100
101
121
122
122
123
124
125
125
126
127
128
128
130
130
131
132
132
153
154
155
156
157
157
158
158
176
177
177
178
179
179
180
181
182
list of illustrationsix
Figure 7.10
Figure 7.11
Figure 7.12
Figure 7.13
Figure 7.14
Figure 7.15
Figure 7.16
Figure
Figure
Figure
Figure
Figure
Figure
Figure
Figure
8.1
8.2
8.3
8.4
8.5
8.6
8.7
8.8
Table 1.1
Table 2.A1
Table 2.A2
Table
Table
Table
Table
2.A3
2.A4
3.A1
3.A2
Table 3.A3
Table 3.A4
Table 4.A1
Table 4.A2
Table 4.A3
Table 4.A4
Table 4.A5
Table 5.A1
High-Frequency Maternal Physical Aggression
by Unemployment Rate
High-Frequency Maternal Psychological
Aggression by Unemployment Rate
High-Frequency Maternal Warmth by
Unemployment Rate
High-Frequency Maternal Parenting
Activities by Unemployment Rate
High-Frequency Paternal Spanking by
Unemployment Rate
High-Frequency Paternal Physical Aggression
by Unemployment Rate
High-Frequency Paternal Psychological
Aggression by Unemployment Rate
Child Internalizing Behavior Problems
Child Externalizing Behavior Problems
Child PPVT Scores
Child Overweight-Obese
Child Internalizing Behaviors
Child Externalizing Behaviors
Child PPVT Scores, Unemployment Rates
Child Overweight-Obese, Unemployment Rates
Fragile Families Sample Composition,
Mothers’ Education
Full Regression Results, Material Hardship
Coefficients and Standard Errors, Rate
of Change, Economic Outcomes
Sensitivity of Coefficients, Economic Outcomes
Coefficients and Standard Errors, Economic Outcomes
Full Regression Results for SNAP
Coefficients and Standard Errors, Rate of
Change for Transfers
Sensitivity of Coefficients, Transfers
Coefficients and Standard Errors, Transfers
Full Regression Results, Parents’ Physical Health
Coefficients and Standard Errors,
All Outcomes by Maternal Education
Coefficients and Standard Errors,
All Outcomes by Paternal Education
Sensitivity of Coefficients, Parents’ Health
Coefficients and Standard Errors, Model 1,
All Outcomes
Full Regression Results, Married to or Cohabiting
with Father or New Partner
182
183
183
184
185
186
186
209
209
210
211
212
212
213
213
8
48
50
53
54
75
77
82
84
106
108
110
113
114
137
x
Table 5.A2
Table 5.A3
Table 5.A4
Table 6.A1
Table 6.A2
Table 6.A3
Table 6.A4
Table 7.A1
Table 7.A2
Table 7.A3
Table 7.A4
Table 7.A5
Table 7.A6
Table 7.A7
Table 7.A8
Table 8.A1
Table 8.A2
Table 8.A3
Table 8.A4
list of illustrations
Coefficients and Standard Errors for
Unemployment Rate, Relationship Outcomes
Sensitivity of Unemployment Rate Coefficients,
Relationship Outcomes
Coefficients and Standard Errors for
Unemployment Rate, Relationship Outcomes
Full Regression Results, Child Support
and Visitation
Coefficients and Standard Errors, Rate of Change,
Father Involvement
Sensitivity of Coefficients, Child Support
and Visitation Outcomes
Coefficients and Standard Errors, Model 1,
Child Support and Visitation Outcomes
Full Regression Results, Maternal Parenting
Coefficients and Standard Errors, Rate of
Change in Unemployment for Maternal
Parenting Outcomes
Sensitivity of Coefficients, Maternal
Parenting Outcomes
Coefficients and Standard Errors, Maternal
Parenting Outcomes by Subgroup
Full Regression Results, Paternal Parenting
Coefficients and Standard Errors, Rate of
Change in Unemployment for Paternal
Parenting Outcomes
Sensitivity of Coefficients, Paternal
Parenting Outcomes
Coefficients and Standard Errors, Paternal
Parenting Outcomes by Subgroups
Full Regression Results, Child Well-Being
Coefficients and Standard Errors, Rate of Change
in Unemployment, Child Well-Being Outcomes
Sensitivity of Coefficients, Child Well-Being Outcomes
Coefficients and Standard Errors,
Child Well-Being Outcomes by Subgroup
139
143
145
164
166
168
169
191
193
196
197
198
200
202
203
219
221
223
224
Contributors
Irwin Garfinkel is Mitchell I. Ginsberg Professor of Contemporary Urban
Problems at the Columbia University School of Social Work.
Sara McLanahan is William S. Tod Professor of Sociology and Public Affairs
at Princeton University.
Christopher Wimer is co-director of the Center on Poverty and Social Policy
at Columbia University and research scientist at the Columbia Population
Research Center.
Jeanne Brooks-Gunn is Virginia and Leonard Marx Professor of Child
Development at Teachers College and the College of Physicians & Surgeons
of Columbia University.
Janet Currie is Henry Putnam Professor of Economics and Public Affairs at
Princeton University.
Elia De la Cruz Toledo is postdoctoral research scientist at the Columbia
University School of Social Work and researcher at Chaplin Hall at the University
of Chicago.
Valentina Duque is postdoctoral fellow at the University of Michigan.
Kristen Harknett is associate professor of sociology at the University of
Pennsylvania.
Ronald B. Mincy is Maurice V. Russell Professor of Social Policy and Social
Work Practice at the Columbia University School of Social Work.
Natasha Pilkauskas is assistant professor at the Ford School of Public Policy
at the University of Michigan.
Daniel Schneider is assistant professor of sociology at the University of
California, Berkeley.
William Schneider is postdoctoral fellow at Northwestern University.
Jane Waldfogel is Compton Foundation Centennial Professor at the Columbia
University School of Social Work and visiting professor at London School of
Economics.
Chapter 1
Introduction
Irwin Garfinkel, Sara McLanahan,
and Christopher Wimer
T
he first decade of the twenty-first century in the United States was
a period of enormous economic turbulence and uncertainty, beginning with a brief recession—often referred to as the dot-com recession—
followed by a dramatic housing bubble, and ending with the Great
Recession, the worst and longest economic downturn since the Great
Depression. In this book, we ask how families with young children fared
during this volatile decade. More generally, we investigate how recessions
affect family life across a wide range of domains, including economic conditions, parents’ health and health behavior, couple relationships, parenting quality, and child health and well-being. These questions are extremely
important. Opportunity and intergenerational mobility are hot topics in
both the popular press and academic research, and we have increasing
scientific evidence that childhood experiences have profound and lasting
consequences for adult lives.1 By extension, how recessions affect families
with young children is of great interest.
Equally important is the question of how recessions affect families at
different points in the income distribution. The last decade’s booms and
busts came on the heels of a sustained rise in income inequality that left the
poorest Americans increasingly struggling to get by. Thus, in addition to
asking how recessions affect the average family, we also examine social class
differences in these effects and the extent to which recessions exacerbate or
minimize pre-existing differences.
Our book is inspired by Glen Elder’s classic study, Children of the Great
Depression.2 Elder’s study followed more than 150 young people born in
the early 1920s in Oakland, California, and described their families’ struggles coping with the Great Depression and its aftermath. Elder found that
the Great Depression placed families and children under profound stress,
reducing fathers’ ability to provide for their children, disrupting parents’
relationships, altering the nature and quality of the parenting their children received, and ultimately affecting children’s long-term outcomes.
Although many families coped admirably with the economic stress, and
many families and children demonstrated fundamental resiliency in the
face of the Great Depression, Elder’s study showed that the economic
2
children of the great recession
forces families face affect the micro-level processes undergirding family
interactions and children’s development.
One of the hallmark intellectual legacies of Elder’s study is the family
stress model, which emphasized the pathways, or intermediate outcomes,
through which large drops in income or permanently low income affect
family functioning and children’s development. The model, which was also
informed by Elder and Conger’s study of the Iowa farm crisis in the early
1980s, is grounded in sociological and psychological theory, makes common sense, and produces strong statistical associations. It has also been
replicated in multiple populations, including African American families,
Finnish couples, Turkish couples, Czech couples, and Korean families.3
Although innovative, the Elder and Conger studies were limited
both by geographic particularity—Berkeley and Oakland in their Great
Depression Study and parts of rural Iowa in their Farm Crisis Study—and
by small samples—167 and 451, respectively. Moreover, the families in
these studies were primarily of white, married couples and their children.
Most important, all of the families experienced the same aggregate economic environment, and thus the effects of income declines and income
deprivation were identified by comparing families who did and did not
personally experience a large drop in income or by comparing families
who were or were not persistently poor. Unfortunately, focusing on
individual-level measures of recessions, such as job loss or income loss,
makes it difficult for researchers to truly identify causal effects of recessions. Specifically, such studies cannot rule out the possibility that the negative outcomes associated with job loss or income loss were the result of
some characteristic of the individual that caused both the job loss and the
family dysfunction. After all, people lose jobs and suffer financial shocks
for a wide variety of reasons in good times and bad. That they do makes
distinguishing between macroeconomic effects and micro-level processes quite difficult.4 Although most studies of the effects of economic
stress have this potential “omitted variable bias” problem, several recent
studies using macro-level measures—such as unemployment rates, plant
closings, and changes in income transfer policy—have found support for
elements of the Elder-Conger model.5
One of our major goals in this book is to test the various elements of
the family stress model—and to add a few new ones—using the dot-com
recession and the Great Recession as “natural experiments.” The recessions were not planned to lead to variations in local unemployment rates
across cities and over time. Thus the variation is natural as opposed to
planned. The families we study did not cause these variations but were
instead subject to their influence. Hence the term experiment. In using
local unemployment rates rather than individual-level indicators of unemployment, we eliminate the omitted variables bias and identify the shortterm causal effects of recessions on family functioning and well-being. We
introduction3
also hope to improve on earlier work by including a more recent and more
diverse sample of families. Although it stands to reason that recessions may
also have longer-term effects that take years or even decades to play out,
we seek to identify short-term, causal effects.
Many studies of economic dislocation rely on small, localized samples.
In contrast, the analyses in this volume are all based on data from the
Fragile Families and Child Wellbeing Study (FFS), which follows nearly
five thousand children born at the turn of the twenty-first century. The
FFS data are based on a probability sample of births in large U.S. cities,
are ethnically and economically diverse, and include a large number of
families formed by unmarried parents. Parents were interviewed shortly
after the birth of their child and again when the child was one, three, five,
and nine years old. By happenstance, the age nine interviews began just
before the Great Recession started and finished just after it ended. The
data’s richness and the fortuitous variation in the interviews’ timing let us
assess how families with young children were affected in the short term by
changes in economic conditions during the first decade of the twenty-first
century and how these experiences and impacts differed by social class.
The analyses in this book are not just another data point with which to
test some of the hypotheses derived from the family stress model, though
they indeed present just such a data point. Rather, the authors recognize
that much has changed in the United States since Elder and Conger’s
classic studies, requiring us to examine the interplay between families and
economic forces anew. The past forty years have seen more and more
women enter the paid labor force. According to data from the Bureau of
Labor Statistics, women’s labor force participation rates increased from
about 46 percent to nearly 60 percent between 1975 and 2000; even
greater increases occurred among women with children. Among mothers
with children under eighteen, labor force participation increased from
47 percent in 1975 to 73 percent by the end of the 1990s, where it has
remained over the last fifteen years.6 Thus, economic downturns today
can compromise both mothers’ and fathers’ economic positions and may
alter family dynamics in ways we cannot fully understand from studies
based on the male breadwinner model. The same period has also been
characterized by a steady growth in the proportion of children born to
unwed parents, from 15 percent in 1975 to about 34 percent in 2000.7
Many of these families start out with two parents living together, but do
not last, and many of the children grow up with a single mother, or, more
commonly, a mother and a series of romantic partners.8 This instability in
family structure may make families more vulnerable to economic shocks,
as they have come to stand on ever more precarious ground.
The chapters that follow take account of these demographic changes by
including a large sample of families with a child born outside marriage (all
chapters), by considering mothers’ labor force participation (chapter 2), by
4
children of the great recession
distinguishing between married and cohabiting parents (chapter 5), and
by examining the parenting behaviors of fathers who live apart from their
children (chapter 6). Last, compared with the 1930s, the United States
today has a much better-developed safety net to protect families from hard
times. The Earned Income Tax Credit provides substantial wage subsidies to parents earning low wages; the Supplemental Nutrition Assistance
Program (SNAP, formerly the Food Stamp Program) provides critical
food assistance to families who lose their jobs or who do not have enough
money to put food on the table; and Medicaid, the State Children’s Health
Insurance Program, and housing assistance help low-income families cover
their medical and housing needs. Few would argue that today’s safety
net is all-encompassing or without problems, but relative to the Great
Depression and even the 1980s, we now have a more robust set of supports
for vulnerable families.9 Despite them, it is also true that today’s safety net
does not guarantee cash assistance to the most severely disadvantaged.10
Thus, the analyses in this book also examine how these supports, both cash
and in-kind, provide a buffer to families who otherwise would have fallen
through the cracks. To address this question, we look (in chapter 3) at
the extent to which government transfers cushioned the effects of the
Great Recession.
The FFS data let us put the effects of the Great Recession in the context
of the lives of children born at the turn of the century. Repeated surveys
of family economic well-being; parents’ relationship status and quality;
parents’ health; mothers’ and fathers’ parenting; and children’s physical
health, emotional health, and cognitive development let us describe the
lives of these families and children over nine years in incredibly rich detail.
THE GREAT RECESSION
The Great Recession, which began as a financial crisis brought on by a
housing bubble and a major stock market crash, quickly metastasized into
a full-blown employment crisis. Recessions are defined by lack of growth
in the overall economy as measured by gross domestic product (GDP). A
recession begins when GDP falls for two consecutive quarters, and it ends
when GDP rises for two consecutive quarters. The Great Recession was
the worst employment crisis since the Great Depression in terms of both
its severity and duration.11 The unemployment rate, which stood at exactly
5 percent in December 2007, peaked at exactly 10 percent by late 2010.12
The unemployment rate is only a partial measure of pain in the labor
market because it excludes people who have stopped looking for work,
which people do more frequently when conditions for finding work are
bleak. Looking just at prime-age men (twenty-five to fifty-four years old),
12.5 percent of men were jobless in December 2007. Two years later,
this number had risen to 20 percent. Prime-age women fared little better:
introduction5
joblessness for this group rose from 27.4 percent at the beginning of the
recession to a high of 31.3 percent toward the end of 2011.13
The unemployment rate alone does not tell us how long a spell of
unemployment lasts for workers who find themselves out of a job. At
the height of the Great Recession, more than 40 percent of unemployed
workers had been unemployed for more than six months. Before it, only
16 percent were in this situation. Similarly, during the only other serious
recession in recent times—the so-called double-dip recession of the early
1980s—this figure reached only about 25 percent.14 In short, those who
experienced the pain of the Great Recession felt a more severe form of pain
than in the past.
The Great Recession came on the heels of a sustained rise in income
inequality. The economists Emmanuel Saez and Thomas Piketty have shown
that the share of income going to the top 10 percent of the population was
over 50 percent in 2012, the most in any year on record. According to estimates from the Congressional Budget Office, real after-tax incomes among
the bottom fifth of American households grew by only 48 percent between
1979 and 2011, versus 200 percent among the top 1 percent.15 Although
48 percent is significant progress, research on trends in poverty shows that
much of the income growth at the bottom of the income distribution came
not from increased earnings but instead from government programs and
policies such as the Earned Income Tax Credit and Food Stamps.16
Although the Great Recession is widely understood as the largest downturn since the Great Depression, it was not the only recession experienced
by families with children born at the turn of the twenty-first century. To
put the Great Recession in context and explicate the relationship between
recessions, unemployment, and household income, figure 1.1 portrays
the two recessions, the national unemployment rate and median household incomes (as a ratio of median income in the year of interest to median
income in 2000) over the period 2000 to 2015. The two recessions are
shaded in gray.
First, focus on the dot-com recession, which began in the spring of
2001 and ended in the fall of the same year. Unemployment, which was
only 4 percent at the onset, continued to creep up well after the recession
officially ended and did not peak, at a bit over 6 percent, until the summer
of 2003. Similarly, family income continued to drop long after the recession ended and unemployment rates began falling. Income reached its
nadir in the summer of 2005, two years after unemployment peaked and
four years after the recession ended.
Next, focus on the Great Recession, which began in late 2007 and ended
in July 2009. Notice that the Great Recession lasted more than twice as
long as the dot-com recession. Unemployment rates also continued to rise
after the recession ended, though they peaked at 10 percent only three
months later, whereas incomes continued to fall for the next two years.
6
children of the great recession
11
102
10
100
9
98
8
96
7
94
6
92
5
90
4
88
3
Month and Year
Recessionary periods = 104
Monthly unemployment rate
(seasonally adjusted)
Seasonally Adjusted Unemployment Rate
104
Jan 2000
Apr 2000
Jul 2000
Oct 2000
Jan 2001
Apr 2001
Jul 2001
Oct 2001
Jan 2002
Apr 2002
Jul 2002
Oct 2002
Jan 2003
Apr 2003
Jul 2003
Oct 2003
Jan 2004
Apr 2004
Jul 2004
Oct 2004
Jan 2005
Apr 2005
Jul 2005
Oct 2005
Jan 2006
Apr 2006
Jul 2006
Oct 2006
Jan 2007
Apr 2007
Jul 2007
Oct 2007
Jan 2008
Apr 2008
Jul 2008
Oct 2008
Jan 2009
Apr 2009
Jul 2009
Oct 2009
Jan 2010
Apr 2010
Jul 2010
Oct 2010
Jan 2011
Apr 2011
Jul 2011
Oct 2011
Jan 2012
Apr 2012
Jul 2012
Oct 2012
Jan 2013
Apr 2013
Jul 2013
Oct 2013
Jan 2014
Apr 2014
HII (January 2000 = 100.0)
Figure 1.1 Median Household Income Index and Unemployment Rate
Seasonally adjusted household
income index (January 2000 = 100.0)
Source: Green and Coder 2014.
In both recessions, the rapid increase in unemployment during the recession was followed by a much slower decline in unemployment during the
recovery. As a result, unemployment rates remained much higher than they
were before the recession throughout most of both recoveries. Similarly,
income was slow to recover after both recessions. Following the dot-com
recession, household incomes did not reach their earlier peak (2002) until
2007. In April 2014, following the Great Recession, income was still about
6 percent below its 2002 and 2007 peaks.
Although we know a great deal about the effects of unemployment and
the Great Recession on economic outcomes for the general population,
we know surprisingly little about the effects on families with children,
especially the noneconomic effects and particularly the effects on more
vulnerable families.
THE FRAGILE FAMILIES DATA
All of the analyses in this book are based on FFS data. The FFS is a longitudinal birth cohort study based on a stratified random sample of nearly
five thousand children born in twenty large U.S. cities between September
1998 and September 2000. The study includes a large oversample of chil-
introduction7
dren born to unmarried parents, who tend to be quite disadvantaged on
many other measures of social class. Three-quarters of the mothers in the
FFS were unmarried when their child was born.
Fifteen of the twenty cities in the study were sampled randomly from all
large U.S. cities. When weighted, the data from these cities are representative of all births in U.S. cities with populations of two hundred thousand
or more. Five additional cities were added to the fifteen-city national sample because they were of special interest to foundations. Although these
five were not chosen randomly, the births in these cities were randomly
sampled, using the same sampling design that was used in the other fifteen
cities. When weighted, data from each of the twenty cities are representative of all births in that city. The analyses in this book are based on data
from all twenty to maximize sample size.
One advantage of the FFS is that, thanks to the large oversample of births
to unmarried parents, the sample is more disadvantaged and more diverse
with respect to income, education, family structure, and race and ethnicity
than most other data sets. Black and Hispanic women and women with
low levels of education are more likely to have children outside marriage;
consequently, these groups are disproportionately represented in the FFS
data. The overrepresentation of disadvantaged families lets us formally
test for differences in the effects of recessions on better- and worse-off
families and children.
Our primary measure of disadvantage is mother’s educational attainment, the single best measure of human capital or potential earning
power. As such, it captures economists’ notion of permanent income
and sociologists’ notion of class. Table 1.1 presents the proportion of
mothers in the sample who are married, cohabiting, or single; who are
white, black, Hispanic, or other racial-ethnic group; and who are poor
as defined by mother’s completed education at their child’s birth. The
average income for each group when the child is one year old is also
displayed. (Poverty status was also measured at year one because income
was poorly measured at baseline.) Note first that more-educated mothers
are much more likely than less-educated mothers to have been married at
birth. Among college-educated mothers, fully 97 percent were married
at birth, a proportion that drops dramatically among their counterparts
with less education. Only 57 percent of mothers with some post–high
school education, 40 percent of those with a high school diploma, and
32 percent of those with less than a high school diploma were married
at the child’s birth. Eighty-six percent of mothers without a high school
diploma and 69 percent of those with only a high school diploma are
either black or Hispanic. In stark contrast, over 70 percent of the collegeeducated mothers are white. Less-educated mothers in the FFS study
have much lower household incomes than more-educated mothers and
are also much more likely to be living in poverty. One-third of mothers
8
children of the great recession
Table 1.1 Fragile Families Sample Composition, Mothers’ Education
Less than
High School
(32%)
Married
Cohabiting
Single
Non-Hispanic white
Non-Hispanic black
Hispanic
Other race-ethnicity
Poverty
Household income ($2010)
N
High
School
(26%)
Some
College
(21%)
College +
(21%)
32
32
34
13
41
45
1
40
26
34
22
46
23
9
57
23
20
32
38
24
6
97
1
2
71
7
7
15
33
$37,000
1,079
21
$47,000
791
12
$67,000
767
2
$176,000
349
Source: Authors’ calculations.
Note: Statistics are weighted using the city weights. N’s are unweighted. Sample is restricted to mothers
who are in all survey waves, N = 2,986.
with less than a high school education were poor one year after giving birth,
whereas 21 percent of mothers with a high school education, 11 percent of
those with some post–high school education, and only 2 percent of those
with a college degree or more were. In short, the overlap between education and other measures of advantage and disadvantage is huge. The
story we tell through the lens of mothers’ education could also be told via
race-ethnicity and family structure.
For our purposes, an important feature of the FFS is the remarkably
rich information on family resources, relationships, and behaviors, including data on family economic well-being; parents’ relationship status and
quality; parents’ health; mothers’ and fathers’ parenting; and children’s
physical health, emotional health, and cognitive development. No other
large data set is comparably rich in all these domains. This makes the FFS
ideal for testing many hypotheses suggested by the family stress model.
Perhaps the key advantage of the FFS data is that families in twenty cities
have been followed since the beginning of the twenty-first century, with
enormous variation in the economic contexts to which these families have
been exposed. Because data collection began at different times in different
cities and continued for up to a year in each city, and because the nine-year
interviews began in 2007 and continued into 2010, FFS data are particularly useful for studying the short-term effects of the Great Recession.
Equally important, the data are ideal for assessing the economic conditions that pertained before the Great Recession—at the children’s birth
and when they were one, three, and five years old.
The Department of Labor keeps track of unemployment rates by month
in cities across the United States. These local data were attached to the FFS
introduction9
data by the city of birth and date of the FFS interview. (For families that
moved to another city after the birth of their child, we also attached the
current city unemployment rate, though estimates reported in the text rely
on the city of birth.) Combining the local unemployment rate data with
the FFS data allows us to measure local unemployment rates at the time of
interview, average unemployment rate in the year before the interview, and
the speed at which the unemployment rate was increasing or decreasing
in the year before the interview. All of these measures are used to describe
associations between local economic conditions and various outcomes,
such as individual-level unemployment, family income, and health.
HOW WE LOOK AT THE DATA
Figure 1.2 depicts the city-level unemployment rate that prevailed in each
month that our data collection team was in the field. The figure underscores the enormous variation in the economic experiences of the families
in the FFS and illustrates how we look at the data. In some cities at some
points in time, the unemployment rate was as low as 2 to 3 percent, and
at the height of the Great Recession in one city it reached a peak of nearly
17 percent. Even within cities and interview waves, local unemployment
rates varied substantially. In 2001, when the children were approximately
one year old, unemployment rates increased substantially in most cities
and by 2 percentage points in the city with the highest unemployment
rate. In 2003, 2004, and 2005, unemployment rates were falling in all of
Unemployment Rate (%)
Figure 1.2 Local Unemployment Rates During Interviewing Periods
17
16
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
0
1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
Birth Age 1
Age 3
Source: Authors’ calculations.
Age 5
Age 9
Austin
Baltimore
Boston
Chicago
Corpus Christi
Detroit
Indianapolis
Jacksonville
Milwaukee
Nashville
New York
Newark
Norfolk
Oakland
Philadelphia
Pittsburgh
Richmond
San Antonio
San Jose
Toledo
10
children of the great recession
our cities. During the year nine interviews, unemployment rates were skyrocketing in the first set of cities, where interviewing began in 2007 and
2008, rising 4 or more percentage points in some places. In other cities,
where interviewing did not begin until 2009, unemployment rates did not
change much during the field period.
We harness all this information and variation to generate robust estimates
of the effects of recessions on families and children. In our analyses, we look
at the relationship between local unemployment rates and family outcomes
when children are approximately one, three, five and nine years old. The
longitudinal data allow us to measure the short-term effects of differences
in local economic conditions over time for the same families and children.
Each of the seven chapters that make up this book focuses on a separate domain of family well-being. Most of these—family income, parents’
mental health, parental conflict, parenting and child well-being—were
part of the original family stress model. Two chapters, reflecting social
changes described—the effects of welfare state programs on cushioning
income losses and child support payments and visitation by nonresident
fathers, the subjects of chapters 3 and 6—are new.
Each chapter follows a standard format. First, the authors review what
we know about the effects of economic downturns on a particular domain.
Next, they describe trajectories for each of their outcome measures over the
first nine years of a child’s life. Finally, they report estimates of outcomes
at two unemployment rates: 5 percent and 10 percent. The difference
between them is our best estimate of the effects of the Great Recession.
We also asked the authors to examine the disparate experiences and
impacts of recessions on more and less disadvantaged families. Although
many dimensions of disadvantage can make families more or less vulnerable to economic forces, we chose educational attainment as our measure
of family economic status. As we have said, economists view educational
attainment as the single best measure of human capital and future earnings,
and sociologists see it as a central component of social class. Education is a
strong predictor of earnings, income, health and well-being, and highly correlated with many other forms of vulnerability and disadvantage. All but one
of the chapters in this book use mother’s education to measure family education. The exception is the chapter on child support and fathers’ involvement, which uses father’s education. In addition, all of the chapters examine
differences in the effect of recessions by race-ethnicity and family structure.
These results are reported in the appendix, and the most important findings
are discussed in the individual chapters. We summarize them here.
As noted earlier, the families in our study were sampled at the time of
the focal child’s birth and again when the child was approximately one
(1999–2001, N = 4,364), three (2001–2003, N = 4,231), five (2003–2006,
N = 4,139) and nine years old (2007–2010, N = 3,515). To study the trajectories of the outcomes of interest over time, the authors use the data at each
introduction11
wave, restricting their analyses to the sample of mothers interviewed at all
survey waves (N = 2,986). To study the effects of recessions and to estimate
the effect of the Great Recession, each of the chapters uses pooled data
from the year one, three, five, and nine data (N ≈ 16,250).17 The samples
for these analyses include all mothers who were ever interviewed. Although
the sample size varies somewhat by the outcome examined, approximately
4,600 mothers contribute to the pooled estimates. As with longitudinal
data, some mothers left the study over time (or left and returned). When
we look at the mothers who left the sample and compare them with the
mothers who stayed, we see that they are slightly more likely to be Hispanic,
to be immigrants, and to have less than a high school education. Fathers
who left the sample are also more likely to be Hispanic, to be immigrants,
and to be less educated, but also to have been poor at the baseline survey,
single, and younger than fathers who remained in the sample.
The Model
Pooling data from interviews conducted when children were one, three,
five, and nine years old, the authors use local unemployment rates and individual fixed-effects models to estimate the effects of recessions on families
and children.18 As noted, variation in local unemployment rates across
cities and over time can serve as a natural experiment that lets us minimize
the problem of omitted variable bias. Individual fixed-effects models control for individual (or family) differences that may be associated with local
unemployment rates and the outcome of interest. In these models, the
association between unemployment and family outcomes is based entirely
on differences within individuals (or families) over time. Although fixedeffects estimates are generally given precedence in estimating causal effects
precisely because they are based solely on within-person comparisons, this
“purity” is purchased at a cost in sample size—for within-person comparisons, we need at least two observations for each person. Further reductions
in sample size occur when the predictor variables do not change, because
these variables must change at least once for a within-person comparison.
Dichotomous variables are less likely to change than continuous variables
are. Finally, another drawback of an individual fixed-effects model is that
all of the other control variables that are fixed—such as race, education
family structure at birth of child, and city of birth—drop out of the equation. Because these coefficients are of some interest, the first appendix
table in each chapter reports coefficients for the effect of unemployment
on one outcome of interest in each chapter, using a model that controls for
baseline education, family structure, age, race-ethnicity, nativity (foreign
or U.S. born), number of children in the household, whether mothers
lived with both parents at age fifteen, city of birth, and year of interview.
Still, because the fixed-effects model produces the best causal estimates,
12
children of the great recession
all chapters use this model to examine the magnitudes of the effects of
unemployment.
It is worth repeating that all our models have a distinct advantage over
earlier studies using individual-level unemployment to assess the family
stress model. Effects of recessions estimated via differences in local unemployment rates over time cannot be due to unmeasured characteristics
of the FFS parents, such as innate abilities and temperament. Unlike the
bulk of the family stress literature, we make no attempt to estimate the
pathways through which recessions affect parental relationships or parenting. Thus we do not distinguish between the direct and indirect effects of
unemployment. In each case, we estimate the total (combined) effects of
unemployment on each of our outcomes.
The local unemployment rate is a good indicator of the probability
that an individual is unemployed. It may not capture the stress associated
with anticipating economic adversity or with uncertainty per se, however.
Experimental research, for instance, indicates that mother monkeys parent
less well—and their offspring do less well—when foraging in poor environments versus rich environments. However, both mothers and offspring
do worse when poor and rich environments are varied randomly, suggesting that uncertainty or insecurity may be more stressful than the actual
experience of adversity.19 Among humans, anticipating significant adverse
events elicits stress or anxiety, and it impairs decision-making and increases
risk aversion and aggression.20 Hedonic adaptation theory suggests that
the emotions elicited by any particular level of unemployment depend on
the previous level.21 For example, an unemployment rate of 8 percent will
elicit hope and confidence if the previous rate was 10 percent, but fear and
anxiety if it was 6 percent. Similarly, an 8 percent rate will elicit greater fear
and anxiety if the previous rate was 4 percent rather than 7 percent because
the size (or rate) of the change is much larger for the former than the latter. Finally, research in behavioral economics demonstrates that people’s
responses to losses are greater than their responses to gains of equal size.22
This research suggests that rapidly increasing unemployment will have
more adverse effects than rapidly decreasing unemployment. Drawing on
these ideas from the behavioral sciences, we hypothesized that parenting
and other outcome measures are associated with both the direction and
the rate of change in macroeconomic conditions and that declines in economic conditions have larger effects on outcomes than improvements in
conditions. To test these hypotheses, each chapter estimates a third model
that adds two variables that measure the rate of increase or the rate of
decrease in unemployment during the previous year. All chapters report
on these results in the appendix and results that are statistically significant
are discussed in the individual chapters and summarized in this one.
All chapters also test whether the overall effect of the local unemployment rate is any greater or smaller during the Great Recession (which
introduction13
coincided with the year nine interview). If the effects of the local unemployment rate are greater during the recession period, our simulated estimates of the effects will be too low because these estimates are derived
from the full model, which relies on the dot-com as well as the recession.23
We find very little evidence of differences, except for the parenting and
child well-being chapters. Differences are discussed in the relevant chapters and summary.
CHAPTER SUMMARIES
We begin by examining trajectories in the economic circumstances of
families with young children during the first decade of the twenty-first
century and how recessions affected their economic well-being. In chapter 2, Irwin Garfinkel and Natasha Pilkauskas focus on three indicators of
family economic well-being: income, poverty, and economic insecurity
(measured as forgoing medical care, food, or housing or not paying bills
because of a lack of money—measures commonly called material hardships). Because individuals’ earnings tend to increase over time as parents
gain more working experience, we might expect to see families’ economic
outcomes improve over the nine years after a child’s birth. However, the
dot-com recession, followed by a tepid recovery and the Great Recession,
are likely to have depressed income gains and produced economic volatility
and insecurity during this decade.
The authors find that the average family income went up and down during this decade; increases were modest, and poverty rates fell somewhat for
all groups. In contrast, insecurity increased a bit. As expected, economic
well-being is strongly related to education. Most striking is the degree to
which families with a college-educated mother stand apart from the rest.
Throughout the decade, by all measures, these families did much better
than their counterparts with less-educated mothers. The family incomes
of college-educated mothers, for example, average around $180,000,
which is about 2.4, 3.6, and 4.4 times that of mothers with some college, a
high school education only, and less than a high school diploma, respectively.
(The differences in median income are nearly as striking: the ratio of highest
to lowest being 4.3 rather than 4.4.) Economic insecurity is very high for
families with the least education, at nearly 50 percent each year. What is
surprising, however, is how widespread economic insecurity is further up
the education distribution: around 40 percent for families with some college. In addition, although families with a college-educated mother did
far better than other families, 20 percent were still economically insecure.
As expected, local unemployment rates are strongly related to family
income, poverty, and economic insecurity. Thus the simulated effects of
the Great Recession on economic well-being are large, reducing family incomes and increasing poverty and economic insecurity. Again, the
14
children of the great recession
college-educated families stand apart as being the least affected. When
unemployment rates were 10 percent rather than 5 percent, their family
incomes were only 5 percent lower. The incomes of three less-educated
groups dropped three to four times that. The percentage increases in the
already high poverty rates of the three groups are dramatic—42 percent,
53 percent, and 75 percent. Insecurity also rose more in percentage terms
as education increases up to a college degree, such that distress rose up
the economic ladder. Families with some education after high school were
particularly hard hit by big recessions, especially in terms of economic
insecurity. Their rates of economic insecurity increased by nearly twothirds, becoming indistinguishable from those of families with less education. This finding may reflect that families with some education after
high school but no college degree are especially vulnerable to hard times,
whereas less-educated families experience poverty and insecurity in both
good and bad times. In short, although families with less than a college
degree fare poorly even in relatively good times, the economic impacts of
the Great Recession on these families were very large, pointing to large
potential ripple effects on other domains of family life.
In chapter 3, Pilkauskas and Garfinkel look at how the American safety
net—public programs that aim to help low-income families, plus unemployment insurance and private transfers—functioned during the first
decade of the twenty-first century. As with family income, in a healthy
economy, we would expect safety-net transfers (public and private) to
decline as children grow older and parents’ earnings rise. In difficult economic times, however, we would expect public benefit receipts to mirror
trends in family income, going up in recessions and down in recoveries.
Trends in private transfer are less certain; in bad times, the need for help
increases but the ability to help declines.
The authors find that very high proportions of families in the two groups
with the least education received benefits from Medicaid, the Earned
Income Tax Credit (EITC), and SNAP—68 percent and 77 percent,
60 percent and 62 percent, and 35 percent and 48 percent for the three programs. Recipient rates for public housing or housing voucher assistance—
26 percent and 30 percent—and Temporary Assistance for Needy Families
(TANF)—16 percent and 25 percent—were lower. Corresponding rates
for Supplemental Security Income (SSI) and Unemployment Insurance
(UI) were below 10 percent. They also find surprisingly high EITC recipient rates for the more highly educated groups—55 percent and 31 percent,
respectively, for those with some education after high school and those
with a college degree. Receipt rates from most programs were higher by
the end of the decade than they were at the beginning, reflecting the weakness of the recovery from the dot-com recession and the severity of the
Great Recession, in combination with the dynamics of individual aid programs. Recipient rates increased for entitlement programs—that is, pro-
introduction15
grams in which the federal government guarantees to pay all federal costs
and to reimburse all state expenses no matter the cost, including Medicaid,
EITC, SNAP, SSI, and UI. In stark contrast, TANF receipts declined and
housing subsidy receipts, after an initial increase when the children were
between ages one and three, were flat. These programs had federal budgets
that were declining or fixed. TANF receipt also decreased as more single
mothers went to work and as more single mothers approached the five-year
lifetime limit on TANF assistance.
Private cash transfer receipts fell between when the children were ages
one and three and then leveled off or increased somewhat, depending on
the education group. The amount transferred increased proportionally
with income, which sets the college-educated apart in terms of the average
amount transferred. Another form of private aid is to share housing, which
the authors term doubling up. The most common form of doubling up is
for the family to move in with the mother’s parent or parents. For all but
the college-educated families, doubling up decreased rapidly when the
children were between age one and three and steadily thereafter as the
child grew older.
Just as they do when it comes to economic well-being, families with
a college-educated parent stand apart with respect to public and private
transfers. They were much less likely than other families to receive incometested benefits from programs such as Medicaid, SNAP, and the EITC.
Although the chapter focuses on safety-net transfers, college-educated
families were much more likely than other families to receive public benefits through the tax system, including government-subsidized, employerprovided health insurance and deductions for home ownership. Indeed,
once employer-provided and tax benefits are counted, the total value of
cash and in-kind transfer benefits is more or less equal across all income
and education groups.24 Nonetheless, of course, less-educated families
are far more reliant on these transfers than are college-educated families
because their market incomes are so much lower.
The authors also examine how well the American safety net responded
to the economic damage wrought by the recessions. They find a strong
positive relationship between local unemployment rates and the receipt
of UI, SNAP, and Medicaid. They also find an association between local
unemployment rates and the receipt of private transfers in the form
of cash assistance from family (mostly) and friends. Not surprisingly,
the poverty-reducing effects of public transfers dwarf those of private
transfers. Indeed, the effects of private transfers are close to zero. At
the peak unemployment rate of 10 percent, poverty rates would have
been 21 to 33 percent higher for the four education groups if not for
public safety-net transfers. Interestingly, the largest mitigation effect,
33 percent, is for the group with some education after high school but
not a college degree.
16
children of the great recession
In chapter 4, Janet Currie and Valentina Duque examine mothers’ and
fathers’ physical and mental health as well as their health-related behaviors. Physical health is measured by reports of limitations in ability to
work as well as a subjective assessment of overall health. Health behaviors
include smoking, drinking, and drug use. Because adult health generally
declines with age, especially among disadvantaged populations, we would
expect to see parents’ health decline over the decade. In contrast, health
behaviors usually improve with age. Difficult economic conditions should
have exacerbated declines in health and retarded improvements in health
behaviors.
Indeed, the authors find that over the decade, both mothers and fathers
across all education groups reported an increase in health problems that
limit work and all groups except college-educated mothers self-reported
overall health declines. More generally, health disparities by education
increased over time. Binge drinking and drug use also generally increased
among both mothers and fathers, although college-educated mothers
and fathers were an exception with respect to drug use, which declined.
Smoking remained flat except among college-educated mothers, where
we see a small decline.
Local unemployment rates are strongly related to health outcomes and
behaviors, and the effects of the Great Recession were therefore quite
pronounced. For example, as a consequence of the Great Recession,
the proportion of mothers with less than a high school education and
only a high school diploma who reported their health as either poor or
fair increased, from 47 percent to 62 percent and from 37 percent to
48 percent, respectively—increases of nearly one-third in both cases.
In every group, the Great Recession also substantially increased (by 30 percent) the proportion of fathers who reported a health problem that limited their employment, the largest effect being for the group with some
education after high school. Only this group reported a decrease in overall health.
The effects on health habits are a bit more complicated. The Great
Recession increased binge drinking and drug use for all mothers, except
the college-educated, among whom drug use declined from an already
low level. Smoking also increased, but only among mothers with some
post–high school education or a college degree. Recessions did not affect
fathers’ smoking, drinking, or drug use, with one exception: fathers with
some post–high school education increased both their drug use and binge
drinking during hard economic times.
Finally, the authors also find that rapid increases in unemployment were
strongly associated with increases in health limitations, drinking, and drug
use. This may be evidence that recessions have direct effects on health via
fear or anticipation of future economic adversity, independent of their
effects on economic well-being.
introduction17
In chapter 5, Daniel Schneider, Sara McLanahan, and Kristen Harknett
examine the stability and quality of parental relationships and whether the
effects of recessions spill over into parents’ relationships. We expect relationship stability and quality to improve over time as unhappy unions dissolve and are replaced with more compatible ones. In contrast, we expect
recessions to undermine stability and increase conflict.
The authors find large disparities by education in the proportion of
mothers who were living with a partner at the time of their child’s birth.
Nearly 100 percent of college-educated mothers were married or cohabiting, compared with only 70 percent of mothers with less than a high school
education. These percentages declined slightly (by less than 10 percent)
for all mothers over the course of the decade, those with a college degree
showing a slightly steeper decline. The picture for parents’ relationship
quality is both different and similar: different in that parents’ reports of
relationship quality were quite similar across education groups, and similar
in that trends in relationship quality were quite stable over time.
Consistent with the family stress model, the authors find that high
unemployment was associated with reductions in marriages and cohabiting unions. The estimated effect of the Great Recession on residential
relationship stability was smaller, in terms of percentage change, than
the effect on either economic conditions or parents’ health. However,
the effects were far from trivial, with decreases in marriage and marriagecohabitation ranging from 7 to 17 percent. The college-educated group
again stands apart, showing no evidence of a decline in residential relationship stability. Interestingly, recessions had little to no effect on couple relationship quality as mothers reported. Fathers, on the other hand, reported
less supportiveness from mothers, an effect that was concentrated in the
lowest education group. Fathers also reported declines in overall relationship quality. The relationship effects were not statistically significant, however, except among Hispanics, where fathers reported declines
in both relationship quality and supportiveness. Again, the collegeeducated group stands apart: mother’s supportiveness, as reported by
fathers, increased as unemployment increased. The authors also found
some evidence that relationship quality was more adversely affected by
local unemployment rates during the Great Recession than in earlier years,
suggesting that their estimates of its effects may be too low. Finally, the
authors report that in previous work, they found that father’s controlling
behavior was not related to the level of unemployment but was strongly
related to the speed with which unemployment increased. This finding
provides some evidence that anticipation or fear of future economic adversity affects behaviors.
Chapter 6, by Ronald Mincy and Elia De la Cruz Toledo, examines
nonresident fathers’ monetary support and visitation and how recessions
affect these two measures of involvement in their children’s lives. Both the
18
children of the great recession
passage of time and recessions are expected to lead to declines in nonresident father involvement.
The proportion of fathers in all groups who live apart from their child
increases over time. Yet, the college-educated group, as in other chapters,
stands apart. At age nine, only 13 percent of the college-educated fathers
live apart from their child, versus between 39 and 55 percent for the lesseducated groups. Indeed, the number of college-educated fathers living
apart from their child is so small that, in analyzing the effects of unemployment on child support and visitation, the authors had to analyze them
in conjunction with fathers with some education after high school. The
authors also find, as expected, that the longer fathers have lived apart from
their child, the less likely they are to pay child support and visit.
The local unemployment rate is strongly related to court-ordered child
support payments. Thus the authors’ estimates of the effect of the Great
Recession on child support payments is substantial—a statistically significant 13 percent decrease. The effects on payments from the fathers with
a high school diploma and fathers with more than a high school education are larger—26 percent and 20 percent—and the former is statistically significant. Declines in informal support for these groups are around
16 percent (not statistically significant) and changes in in-kind support
are minimal. Declines in all kinds of child support for high school dropouts are smaller and not statistically significantly different from zero. The
authors also find that recessions have no effect on whether fathers visit
with their child in the last month for any education group.
In chapter 7, William Schneider, Jeanne Brooks-Gunn, and Jane
Waldfogel examine the quality of mothers’ and fathers’ parenting and how
it is affected by recessions. They measure parenting quality by harsh parenting (spanking, and high-frequency physical and psychological aggression), warmth, and the number of parent-child activities. Harsh parenting
is expected to increase not long after the child first becomes independent
by walking and talking—sometimes referred to colloquially as the terrible twos—but to decrease steadily sometime after age three to five as the
parents gain experience and the child matures. Warmth and the number
of parenting activities are also expected to decrease as the child ages. The
authors find evidence of all these patterns in the FFS data. The authors also
find interesting differences by mothers’ education. Warmth increases with
education at all ages. But spanking and physical aggression are unrelated
to education until children reach age nine, when both decrease steadily
with increases in education. At age nine, however, high-frequency psychological aggression is highest and activities with the child are lowest for
college-educated mothers.
Harsh parenting is expected to increase during recessions, but the
effects of recessions on warmth and parents’ activities with the child are
ambiguous. Mother’s parenting is sensitive to unemployment rates, but
introduction19
not in the expected way. The authors find no evidence that high local
unemployment rates led to worse parenting by mothers. Indeed high
unemployment was associated with less spanking and physical aggression
and was unrelated to warmth and activities with the child. Rapid increases
in unemployment rates were associated with increases in maternal warmth
and activities with their child and in some specifications with increases in
harsh parenting, whereas rapid decreases were associated with increases in
harsh parenting.
Father’s harsh parenting, like that of mothers, decreased rather than
increased when unemployment rates were high. In general, neither rapidly increasing or decreasing unemployment was associated with harsh
parenting, although among college-educated fathers, rapidly increasing
unemployment was associated with increases in spanking. (Warmth and
frequency of activities were not measured for fathers.)
In short, the dot-com recession and the Great Recession affected parenting in unexpected ways. High unemployment was associated with less, not
more, harsh parenting among both mothers and fathers. Among mothers,
rapidly increasing unemployment was associated with more warmth and
more activities with the child, and, in some specifications and in previous
research, with more harsh parenting. Rapidly decreasing unemployment
was also associated with more harsh parenting. We offer a possible explanation for the perplexing findings for parenting behavior.
Last, chapter 8, also by Schneider, Brooks-Gunn, and Waldfogel,
describes children’s developmental outcomes during the first nine years
of their life and to examine whether recessions affect these outcomes.
Behavioral problems are captured on two scales: externalizing (acting out) and internalizing (withdrawing). Cognitive development is
assessed by Peabody Picture Vocabulary Test (PPVT) scores. Finally,
the authors examine one health outcome, obesity. Externalizing behavior increases as children become more independent when they learn to
walk and talk and then decreases steadily. Internalizing behaviors also
generally decrease as children grow older. Cognitive test scores are age
normed and so not expected to trend. Obesity is expected to increase as
children age. Children’s outcomes are also expected to diverge by class
as they grow older. The authors find all of these patterns in the FFS data,
except for obesity where increase is minimal as children age, as is divergence by class.
Recessions are expected to increase behavior problems and reduce child
well-being. Yet, as with the parenting outcomes, the authors find no evidence that unemployment rates were associated with any of these outcomes. Once again, however, rapid changes in unemployment rates were
strongly associated with both improvements and reductions in child wellbeing. On the one hand, rapid increases in unemployment rates were associated with more acting out (for all education groups except those with
20
children of the great recession
college-educated mothers). On the other hand, rapid decreases in unemployment were associated with improvements in PPVT scores and internalizing behaviors (among mothers who did not complete high school).
Although the results for children are unexpected, they are consistent with
the parenting results in that both parenting and child outcomes are driven
by the rapidity of change in unemployment rates rather than the level of
the unemployment rate.
CONCLUSION
All American families with children born at the beginning of the twentyfirst century lived through turbulent economic times during their child’s
first decade. Depending on mothers’ education, however, their experiences differed dramatically. Families with a college-educated mother had
much higher incomes and much lower rates of poverty and economic
insecurity throughout the decade than families with less-educated mothers.
They also received different kinds of government-subsidized benefits,
were in much better health, and had more stable parental relationships.
Not surprisingly, the children in these families fared better than children
in less-educated families. Finally, families with a college-educated mother
stand apart because they were minimally affected by the dot-com recession
and the Great Recession.
At the other extreme, among families with a mother who did not finish
high school, poverty and economic insecurity, poor health, single parenthood, and poor child outcomes were common throughout the child’s first
decade. Families in which the mother had a high school diploma generally
fared somewhat better than those with a mother who did not, and families
in which the mother had some education after high school generally fared
better than their counterpart families. For a few outcomes, families in
which the mother had some education after high school looked more like
those with a college-educated mother than like those whose mother had
only a high school diploma; for most outcomes, however, they looked
more like the two less-educated groups.
Given the evidence presented in these chapters, we would have to
conclude that the Great Recession’s effects on two-thirds of American
families (those in which the mother did not have a college degree) were
quite large. For those with a high school diploma or less, the recession
seriously exacerbated an already bad situation. This was true not only
for families’ economic well-being but also for parents’ health. Even the
effects on family stability were notable, though smaller. The near immunity of college-educated families and the large negative consequences
for less-educated families mean that the Great Recession increased the
already large divide between families at the top and bottom of the income
distribution.
introduction21
Of particular interest are instances when the most adverse effects appear
among families with some education after high school: economic insecurity, fathers’ health (including limitations on work, binge drinking, and
drug use) and parental relationship quality. To us, these results underscore
the precariousness of this group’s position. More generally, though the
Great Recession appears to have increased economic disparities among
the most and least educated families, it also appears to have narrowed
disparities among families with less than a college degree. An important
exception is health disparities, which widened during the Great Recession
among mothers with less than a college degree.
In addition to comparing the effects of unemployment on families
with different levels of education, the chapters in this volume examine
the effects of high unemployment on families with different racial-ethnic
backgrounds (white, black, and Hispanic) and different family structures
at a child’s birth (married, cohabiting, and single). In most instances, the
estimates tell a consistent story. The negative effects of unemployment fell
disproportionately on blacks and Hispanics and on unmarried mothers.
A few exceptions prove this rule. For example, white mothers were more
likely to increase their alcohol use during periods of high unemployment,
white fathers and married fathers were more likely to see their health
decline, and high unemployment among Hispanic mothers appeared to
increase their parental warmth.
One of the most surprising findings is that high unemployment rates
were not associated with declines in either parenting quality or child wellbeing. Indeed, high unemployment rates were associated with decreases
in harsh parenting. At first glance, this would seem to contradict Elder
and Conger’s earlier findings. These findings do not mean that recessions
do not harm parenting and child well-being. Indeed, the authors found
that rapid changes in local unemployment rates increased maternal harsh
parenting and child externalizing behavior. One possible interpretation is
that in the short run, fear and uncertainty are the principle drivers of harsh
parenting and that parents reduce their harsh parenting when unemployment is stable, however high. The Great Recession is the only period in
which unemployment was high among the families in our sample. Thus, it
is likely that our unemployment rate results are driven by year nine unemployment rates. It is not too hard to imagine that as the Great Recession
set in and unemployment rates increased precipitously, the fear of another
Great Depression led to deteriorations in parenting. Once unemployment
stopped increasing and the fear of another Great Depression dissipated,
parents calmed down and their parenting improved, despite the high
unemployment rates. Finally, it bears emphasizing that the analyses of parenting and child outcomes are based on estimates of the short-run effects
of high unemployment and do not rule out that possibility that prolonged
unemployment lowers parenting quality and child well-being.
22
children of the great recession
Rapid changes in unemployment also appear to affect fathers’ controlling behavior and mothers’ health, smoking, drinking, and drug use. For
these outcomes, however, we observe a negative effect only when unemployment is increasing rather than decreasing. In these instances, then,
the anticipation of negative outcomes seems to be more important than
uncertainty per se.
Behavior stimulated by fear or uncertainty about the future is not necessarily irrational. Some or even most of those who anticipate or fear future
unemployment will actually become unemployed. The rate of change in
the local unemployment rate, after all, is a very good predictor of the
future local unemployment rate. Knowing that unemployment is coming
can be just as stressful as actually being unemployed. But why would rapidly decreasing unemployment have negative effects? Improving conditions, other things being equal, are expected to lead to positive outcomes.
But, very rapid change even in a positive direction may cause stress by
requiring rapid adaptation. For example, the prospect of going back to
work for mothers who are unemployed entails changes in child care and
other routines, and rapid changes are likely to be stressful. Because most
families with children experienced very high rates of economic insecurity,
we should not be surprised that uncertainty and anticipation or fear of
adverse future events affect their lives. Bad things happen to these families
even in good times.
Family Stress Model
Taken as a whole, what do our findings imply for the family stress model?
The model posits that large drops in income (or permanently low income)
will harm parents’ health, relationship quality, parenting quality, and child
well-being. To date, the model has been tested by relying on individuallevel differences between those who did and did not experience a big
income loss or between those who had permanently low incomes and
those who had higher incomes. The estimates produced this way may
suffer from omitted variables bias. By relying on a “natural experiment”—
variation in local unemployment rates—to measure risk of unemployment, we take a more conservative approach to estimating the effects of
the Great Recession on families and children. Our estimates suggest the
Great Recession had at least a few devastating effects: large decreases in
family income and parents’ health, large increases in poverty and economic insecurity, and modest decreases in parents’ relationship quality.
As a test of these individual components of the model, the results are an
impressive confirmation.
Less consistent and indeed puzzling from the point of view of the wellordered family stress model, we found that higher unemployment rates
were associated with less rather than more harsh parenting, and that rapid
introduction23
changes in unemployment were much stronger predictors of parenting
and child outcomes than unemployment rates per se were. The latter
suggests that some of the effects of the Great Recession, and of recessions in general, precede unemployment or income loss and likely extend
beyond the families who are affected directly. At the onset of the Great
Recession, confidence in the economy plummeted, and fear of another
Great Depression was widespread. These changes seem to have harmed
mothers’ health, relationship quality, parenting quality, and children’s
well-being independent of actual unemployment. How coping with the
stress of rapid changes fits into the family stress model is not clear, and is a
challenge for future research.
Limitations
Perhaps the greatest strength of the FFS data is the study’s longitudinal
design. Repeatedly observing the same group of families let us estimate
individual fixed-effects models, which, combined with the natural experiment design, let us derive estimates of the Great Recession’s impacts that
are not affected by omitted variables bias. The longitudinal design is also
a source of weakness, however. Longitudinal surveys are expensive and,
consequently, their sample size is typically small. Although the FFS sample
is large and diverse compared with the longitudinal samples used in the
family stress literature, it is tiny relative to the samples in studies using
pooled repeated cross-sections of the Current Population Survey (CPS)
or the American Community Survey (ACS), which are much larger. The
ACS, for example, contains millions of persons. The FFS’s relatively small
sample size makes it harder to detect statistically significant differences
among groups. In the chapters on economic outcomes, we rely on other
research based on the CPS or ACS to verify or contradict the patterns we
find in the FFS data.
The FFS is a birth cohort study of children born between 1998 and
2001. Thus it is possible that our findings can’t be generalized to families with children born either before 1998 or after 2001. Another concern is whether there were interactions between the timing of recessions
and the developmental trajectory of some of the outcomes we tracked.
For example, children’s behavior problems are expected to peak at age
three and decline markedly after age five. If these “natural” changes
coincided with dramatic increases or decreases in unemployment rates
they could have masked or exacerbated the effects of the recessions and
recoveries. That we observed two periods of rapidly increasing unemployment at the onset of the dot-com and Great Recessions and only
one of decreasing unemployment heightens this concern. That said,
that children of the same age were interviewed in twenty cities over
three years makes it less likely that our estimates are biased by a systematic
24
children of the great recession
relationship between child development and levels or rates of change in
unemployment.
The FFS data and our methods have several other limitations. As noted
in passing, the data are affected by attrition and migration. For the most
part, we do not think that attrition and migration bias our results but we
do call attention to a few specific problematic instances. Also, the timing
of the interviews may have biased our estimates of unemployment’s effects
on outcomes. Because unemployment rates were rising rapidly during the
period of the year nine interviews, families interviewed at the end of the
period were likely to have experienced higher unemployment than those
interviewed at the beginning. On the one hand, if families with the most
problems were harder to find and interview, our estimates of unemployment’s effects would be inflated. On the other hand, if the families with
the most difficulties were more likely to complete the surveys earlier in
order to receive the financial compensation, our estimates would be biased
toward zero. In several early investigations, we controlled for timing of
interviews and found that our estimated effects were unaffected.
The FFS, because it samples from births in large American cities, does
not cover rural populations and poorly represents suburban ones. We
doubt that the relationships between unemployment and family outcomes
described in this volume would be much different if these other groups
were fully represented, but that is a matter for empirical investigation.
Future Research
More research on how the rate of change in unemployment affects behaviors is clearly in order. Will the results regarding fear and anticipation or
those regarding uncertainty replicate in investigations using other data?
Future research should estimate how prolonged unemployment affects
outcomes, especially parenting quality and child well-being. This could be
done with the FFS data using individual-level unemployment, but it is not
clear to us how our preferred measure of local unemployment rates could
be used. Similarly, although our analyses focus on the short-term effects of
the Great Recession, long-term effects are also of interest. FFS data could
be used to study the long-term effects, though not with the methodology
we use in this study.
It would also be desirable to estimate the full family stress model with
pathways using the FFS data. Findings using the FFS are likely to replicate earlier findings in the literature. In this context, it would be useful
to compare the size of effects estimated using individual-level measures
of unemployment with that of those estimated using the exogenous local
unemployment rate.
Finally, estimating the costs and benefits of alternative reforms to reduce
poverty and insecurity would be a very useful contribution.
introduction25
Policy Implications
The findings reported in this book show that a large proportion of American
children born at the turn of the century are poor and economically insecure. Economic insecurity extends well beyond families formed by parents with minimal education to include those with high school diplomas
and even those with some college or other post–high school education.
These families fare much worse than those of college-educated parents in
good as well as bad times, and the Great Recession seriously exacerbated
this disparity. By themselves, our empirical findings have no strong policy
implications. However, if we value reducing poverty and economic insecurity and increasing intergenerational mobility, the chapters in this book
lead to several conclusions about how we should move forward.
That two-thirds of American families with children experience economic insecurity in good times as well as bad times suggests that existing
welfare state programs are inadequate. Programs that target low-income
families can reduce poverty and provide catastrophic insurance against a
large and prolonged economic downturn. SNAP does an excellent job in
this respect; along with UI, it played a critical role in mitigating the effects
of the Great Recession, as we see in chapter 3. But programs that target the
poor do not help the many middle-class families struggling to make ends
meet. Universal entitlement programs, such as UI, universal preschool,
paid parental leave, children’s allowances, and child support assurance,
which provide benefits to all families regardless of income, are well-suited
to this task.25
Turning now to more general policy implications, perhaps the most
obvious point is the need to increase the proportion of families with a
college-educated mother. From the mid-nineteenth century through the
1960s, America led the world in providing mass public education, first
at the elementary level and then at the secondary and college levels. 26
In 1970, one-third of Americans obtained a college degree, the highest
proportion in the world. Only Canada was close. Today the proportion
remains about the same but many other rich nations have caught up to or
surpassed it. In Canada, for example, the proportion is now 50 percent.
Increasing the proportion in the United States will require changes not
only in higher education policy and financing to make college more accessible, but also in K–12 and early childhood education, to make sure that
students are “college ready.” Analyzing various policies that might achieve
this goal is beyond the scope of this volume. However, we want to call
special attention to proposals for high-quality, universal preschool education. Universal pre-K is not just a good investment in our children’s future
productivity, it gives the current generation of young mothers a valuable
subsidy, allowing them to pursue more education as well as on-the-job
training. Because maternal education increases the quality of parenting
26
children of the great recession
and the home environment, universal preschool is a two-generation program likely to create a powerful feedback loop.27
A second implication of the book is that policymakers need to think
harder about how to discourage nonmarital childbearing and the formation
of fragile families. Such families are much more likely to be poor and economically insecure than married-parent families, even in good economic
times. They are also more prone to unemployment and declines in health
than other families during recessions. What government should do is less
obvious. Recent programs designed to increase marriage among unmarried couples had disappointing results.28 Encouraging young women to
delay their first pregnancy until they have a stable job and a stable relationship is likely to be a more successful strategy for reducing births to
unmarried women. Such programs have shown a good deal of success in
recent years.29 Increasing the human capital of girls and boys and reducing
economic insecurity are both likely to increase marriage.30
A third implication is that we need to work harder on reducing racial
and ethnic disparities in economic conditions and opportunities. Even
after accounting for differences in parents’ education and marital status,
children born to black and Hispanic parents face more economic barriers in good times and more economic disruptions during periods of high
unemployment than children born to white parents. Evidence is widespread and indisputable that minority families suffered disproportionately from the collapse in the real estate market that triggered the Great
Recession, partly because lenders pushed risky loans on them.
Assessing the benefits and costs of alternative policies to reduce poverty
and economic insecurity of American families with children is beyond the
scope of this volume. What is clear is that much can be done and much
remains to be done.
NOTES
1. Heckman 2006; Shonkoff and Phillips 2000.
2. Elder 1974.
3. Conger et al. 1992; Kinnunen and Feldt 2004; Aytaç and Rankin 2009;
Hraba, Lorenz, and Pechac̆ová 2000; Kwon et al. 2003.
4. Although the international studies that test the family stress model are more
diverse and more representative than the Elder and Conger studies, they also
suffer from the problem of omitted variable bias.
5. Wood et al. 2012; Page, Stevens, and Lindo 2009; Oreopoulos, Page, and
Stevens 2008; Milligan and Stabile 2008; Dahl and Lochner 2012.
6. BLS 2014.
7. Ventura and Bachrach 2000.
8. McLanahan and Jencks 2015.
introduction27
9. Fox et al. 2015.
10. Shaefer and Edin 2015.
11. Hout, Levanon, and Cumberworth 2011.
12. BLS 2015.
13. Hout and Cumberworth 2014.
14. Acs 2013.
15. Stone et al. 2015.
16. Wimer et al. 2013.
17. Exceptions include chapter 8 on child well-being, where the data were not
collected until year three, and chapter 6, where the sample is limited to
nonresident fathers.
18. Local unemployment rates prevailing at the time of the parent interview in the
city in which the child was born were used to measure economic conditions.
The reason for utilizing the city of birth as opposed to current residence
is that families may have moved in response to high unemployment rates,
which might lead to an underestimate of recession effects. In earlier work, the
chapters on economic and health outcomes also measured unemployment
using current city and found that the results did not change. In a few
instances, the information for a particular domain was not available in the
first follow-up interview, in which case the researchers pooled data from the
three-, five-, and nine-year interviews.
19. Rosenblum and Paully 1984; Coplan et al. 1998.
20. Loewenstein et al. 2001; Berkowitz 1990; Baumeister et al. 2007; Wilson
and Gilbert 2013.
21. Frederick and Loewenstein 1999.
22. Tversky and Kahneman 1974; Kahneman, Slovic, and Tversky 1982.
23. Similarly, as requested by a reviewer, all chapters test whether the inclusion of
individual-level measures of mothers’ and fathers’ employment can account
for the effects of aggregate unemployment rates. Inclusion of individual-level
variables of mothers’ and fathers’ employment in the week prior to the survey
had little to no effect on the local unemployment rate variable. All chapters
report on these tests in appendix table 3. Last, some additional supplemental
analyses were run by the authors of each chapter. These analyses are described
in chapter text or appendices.
24. Garfinkel, Rainwater, and Smeeding 2010; Garfinkel and Zilinawala 2015.
25. See Garfinkel, Rainwater, and Smeeding 2010; McLanahan and Garfinkel
2012; Bradbury et al. 2015.
26. Garfinkel, Rainwater, and Smeeding 2010.
27. Haskins, Garfinkel, and McClanahan 2014.
28. Haskins 2015; Wood et al. 2012.
29. Sawhill 2014.
30. Lerman and Wilcox 2014; but see Schneider 2015.
28
children of the great recession
REFERENCES
Acs, Gregory. 2013. Assessing the Factors Underlying Long-Term Unemployment
during and After the Great Recession. Washington, D.C.: The Urban Institute.
Aytaç, Isik. A., and Bruce H. Rankin. 2009. “Economic Crisis and Marital Problems
in Turkey: Testing the Family Stress Model.” Journal of Marriage and Family
71(3): 756–67.
Baumeister, Roy F., Kathleen D. Vohs, C. Nathan DeWall, and Liqing Zhang.
2007. “How Emotion Shapes Behavior: Feedback, Anticipation, and Reflection,
Rather than Direct Causation.” Personality and Social Psychology Review 11(2):
167–203.
Berkowitz, Leonard. 1990. “On the Formation and Regulation of Anger and
Aggression: A Cognitive-Neoassociationistic Analysis.” American Psychologist
45(4): 494–503.
Bradbury, Bruce, Miles Corak, Jane Waldfogel, and Elizabeth Washbrook. 2015.
Too Many Children Left Behind. New York: Russell Sage Foundation.
Bureau of Labor Statistics (BLS). 2014. Women in the Labor Force: A Databook.
BLS Report no. 1052. Washington: U.S. Department of Labor.
———. 2015. “Labor Force Statistics from the Current Population Survey.”
Accessed October 2, 2015. http://data.bls.gov/timeseries/LNS14000000.
Conger, Rand D., Katherine J. Conger, Glen H. Elder Jr., Frederick O. Lorenz,
Ronald L. Simons, and Les B. Whitbeck. 1992. “A Family Process Model
of Economic Hardship and Adjustment of Early Adolescent Boys.” Child
Development 63(3): 526–41.
Coplan, Jeremy D., Ronald C. Trost, Michael J. Owens, Thomas B. Cooper,
Jack M. Gorman, Charles B. Nemeroff, and Leonard A. Rosenblum. 1998.
“Cerebrospinal Fluid Concentrations of Somatostatin and Biogenic Amines
in Grown Primates Reared by Mothers Exposed to Manipulated Foraging
Conditions.” Archives of General Psychiatry 55(5): 473–77.
Dahl, Gordon B., and Lance Lochner. 2012. “The Impact of Family Income on
Child Achievement: Evidence from the Earned Income Tax Credit.” American
Economic Review 102(5): 1927–956.
Elder, Glen H., Jr. 1974. Children of the Great Depression: Social Change in Life
Experience. Chicago: University of Chicago Press.
Fox, Liana, Christoper Wimer, Irwin Garfinkel, Neeraj Kaushal, and Jane Waldfogel.
2015. “Waging War on Poverty: Poverty Trends Using a Historical Supplemental
Poverty Measure.” Journal of Policy Analysis and Management 43(3): 567–92.
Frederick, Shane, and George Loewenstein. 1999. “Hedonic Adaptation.” In
Well-Being: The Foundations of Hedonic Psychology, edited by Daniel Kahneman,
Ed Diener, and Norbert Schwarz. New York: Russell Sage Foundation.
Garfinkel, Irwin, Lee Rainwater, and Timothy Smeeding 2010. Wealth and Welfare
States: Is America a Laggard or Leader? New York: Oxford University Press.
Garfinkel, Irwin, and Afshin Zilanawala. 2015. “Fragile Families in the American
Welfare State.” Children and Youth Services Review 55(C): 210–21.
Green, Gordon, and John Coder. 2014. Household Income Trends: May 2014.
Annapolis, Md.: Sentier Research.
Haskins, Ron. 2015. “The Family Is Here to Stay—or Not.” Future of Children
25(2): 129–53.
introduction29
Haskins, Ron., Irwin Garfinkel, and Sara S. McLanahan. 2014. “Introduction:
Two-Generation Mechanisms of Child Development.” Future of Children
24(1): 3–12.
Heckman, James J. 2006. “Skill Formation and the Economics of Investing in
Disadvantaged Children.” Science 30(312): 1900–902.
Hout, Michael, and Erin Cumberworth. 2014. National Report Card: Labor
Markets. Stanford, Calif.: Stanford Center on Poverty and Inequality.
Hout, Michael, Asaf Levanon, and Erin Cumberworth. 2011. “Job Loss and
Unemployment.” In The Great Recession, edited by David B. Grusky and Bruce
Western. New York: Russell Sage Foundation.
Hraba, Joseph., Frederick O. Lorenz, and Zdenka Pechac̆ová. 2000. “Family
Stress During the Czech Transformation.” Journal of Marriage and the Family
62(2): 520–31.
Kahneman, Daniel, Paul Slovic, and Amos Tversky. 1982. Judgment Under
Uncertainty: Heuristics and Biases. New York: Cambridge University Press.
Kinnunen, Ulla, and Taru Feldt. 2004. “Economic Stress and Marital Adjustment
Among Couples: Analyses at the Dyadic Level.” European Journal of Social
Psychology 34(5): 519–32.
Kwon, Hee-Kyung, Martha A. Rueter, Mi-Sook Lee, Seonju Koh, and Sun Wha
Ok. 2003. “Marital Relationships Following the Korean Economic Crisis:
Applying the Family Stress Model” Journal of Marriage and Family 65(2):
316–25.
Lerman, Robert I., and W. Bradford Wilcox. 2014. For Richer, for Poorer: How
Family Structures Economic Success in America. Washington, D.C.: AEI and
Institute for Family Studies.
Loewenstein, George F., Elke U. Weber, Christopher K. Shee, and Ned Welch.
2001. “Risk as Feelings.” Psychological Bulletin 127(2): 267–86.
McLanahan, Sara S., and Irwin Garfinkel. 2012. “Fragile Families: Debates, Facts,
and Solutions.” In Marriage at the Crossroads, edited by Marsha Garrison and
Elizabeth S. Scott. New York: Cambridge University Press.
McLanahan, Sara S., and Christopher Jencks. 2015. “Was Moynihan Right?:
What Happens to Children of Unmarried Mothers.” Education Next 15(2):
17–22.
Milligan, Kevin, and Mark Stabile. 2008. “Do Child Tax Benefits Affect the
Wellbeing of Children? Evidence from Canadian Child Benefit Expansions.”
NBER working paper no. 14624. Cambridge, Mass.: National Bureau of
Economic Research.
Oreopoulos, Philip, Marianne Page, and Ann H. Stevens. 2008. “The Inter­
generational Effects of Worker Displacement.” Journal of Labor Economics
26(3): 455–83.
Page, Marianne, Ann H. Stevens, and Jason Lindo. 2009. “Parental Income
Shocks and Outcomes of Disadvantaged Youth in the United States.” In The
Problems of Disadvantaged Youth: An Economic Perspective, edited by Jonathan
Gruber. Chicago: University of Chicago Press.
Rosenblum, Leonard A., and Gayle S. Paully. 1984. “The Effects of Varying
Environmental Demands on Maternal and Infant Behavior.” Child Development
55(1): 305–14.
Sawhill, Isabel V. 2014. Generation Unbound: Drifting into Sex and Parenthood
Without Marriage. Washington, D.C.: Brookings Institution Press.
30
children of the great recession
Schneider, Daniel. 2015. “Lessons Learned from Non-Marriage Experiments.”
Future of Children 25(2): 155–78.
Shonkoff, Jack P., and Deborah A. Phillips. 2000. From Neurons to Neighborhoods:
The Science of Early Childhood Development. Washington, D.C.: National
Academies Press.
Stone, Chad, Danilo Trisi Arloc Sherman, and Brandon Debot. 2015. “A Guide
to Statistics on Historical Trends in Income Inequality.” Washington, D.C.:
Center on Budget and Policy Priorities. Accessed October 2, 2015. http://
www.cbpp.org/research/poverty-and-inequality/a-guide-to-statistics-onhistorical-trends-in-income-inequality?fa=view&id=3629.
Tversky, Amos, and Daniel Kahneman. 1974. “Judgment Under Uncertainty:
Heuristics and Biases.” Science 185(4157): 1124–131.
Ventura, Stephanie J., and Christine A. Bachrach. 2000. “Nonmarital Childbearing
in the United States, 1940–99.” National Vital Statistics Reports 48(16): 1–39.
Wilson, Timothy D., and Daniel T. Gilbert. 2013. “Comment: The Impact Bias
Is Alive and Well.” Journal of Personality and Social Psychology 105(5): 740–48.
Wimer, Christopher, Liana Fox, Irwin Garfinkel, Neeraj Kaushal, and Jane
Waldfogel. 2013. “Trends in Poverty with an Anchored Supplemental Poverty
Measure.” IRP discussion paper no. 1416-13. Madison, Wisc.: Institute for
Research on Poverty.
Wood, Joanne N., Sheyla P. Medina, Chris Feudtner, Xianqun Luan, Russell
Localio, Evan S. Fieldston, and David M. Rubin. 2012. “Local Macroeconomic
Trends and Hospital Admissions for Child Abuse, 2000–2009.” Pediatrics
130(2): e358–64.
Wood, Robert G., Quinn Moore, Andrew Clarkwest, Alexandra Killewald,
and Shannon Monahan. 2012. “The Long-Term Effects of Building Strong
Families: A Relationship Skills Education Program for Unmarried Parents.”
OPRE Report no. 2012-28A. Washington: U.S. Department of Health and
Human Services.
Chapter 2
Economic Well-Being
Irwin Garfinkel and Natasha Pilkauskas
R
ecessions are primarily an economic phenomenon. If we are to understand the effects of recessions on families and children, the first order
of business is to document how recessions affect families’ pocketbooks.
This chapter describes the economic well-being of families with children
born at the turn of the twenty-first century and how the Great Recession
affected this well-being. The economic circumstances of families are
described in terms of employment, household income, and two measures
of economic distress—poverty and economic insecurity or hardship. As
described in chapter 1, we also consider the possibility that recessions may
affect families differently depending on their initial background and level
of vulnerability. Thus we examine each of these indicators separately for
families with different social class backgrounds, measured by mother’s
education—whether less than high school, a high school diploma only,
more than high school but no college degree, or a college degree. We also
examine economic well-being separately for two other family characteristics linked with advantage and disadvantage: parents’ race-ethnicity and
whether parents were married, cohabiting, or living apart at the time of
the child’s birth.
We examine a number of different measures of economic well-being,
including numerous alternative measures of employment and earnings
and several other measures of economic well-being. All of them yield
the same overall story as the four outcomes we report. Here we describe
briefly the rationale for focusing on employment, income, poverty, and
economic insecurity and how each indicator was measured.
The most immediate effect of recessions on families is a loss of employment. Thus the first outcome we examine is the biological mother’s and
father’s employment. We study biological fathers rather than all fathers
living with the child because the biological father may contribute child
support even if he does not live in the same household, and the data on
social fathers are incomplete. We use employment in the week prior to the
survey because it is most current and therefore most likely to be the most
accurately reported labor market outcome; it is also how the Bureau of
Labor Statistics measures employment. We use employment rather than
unemployment because a change in employment picks up discouraged
32
children of the great recession
workers who have dropped out of the labor force after long-term unemployment as well as unemployed workers who looked for work in the
previous week. We also use employment because at the highest levels of
education very few mothers or fathers were unemployed and therefore
our sample is not large enough to estimate the effects of recessions for
those groups.
Our first measure of economic well-being is household income. One of
the most commonly used measures of economic well-being, it normally
includes earnings, cash government transfers, and cash transfers from family and friends, but not the Supplemental Nutrition Assistance Program
(SNAP, commonly known as Food Stamps), a near cash benefit, or the
Earned Income Tax Credit (EITC). Because both of these transfers are
widespread and substantially increase the total incomes of families who
receive them, we follow increasingly common practice among leading
researchers and include both in our household income measure. We measure total household income using the mother’s report of total income or
of the components of household income during the prior twelve months,
whichever is higher. For example, if a mother reports $40,000 in annual
household income, but summing annual earnings, SNAP, cash assistance,
and other income components yields $45,000, we treat the latter as her
true income.
Our second measure of economic well-being is poverty. Poverty is the
most common measure of whether households are in poor financial shape.
Poverty is measured using the Census Bureau’s official poverty thresholds. Unlike the official measure, our measure of household income also
includes the EITC and SNAP (as noted). The official poverty threshold
for a family of three in 2014 is a bit less than $20,000.
Our third measure of economic well-being is what is typically called
material hardship, but may be better described as economic insecurity.
The material hardship measure is newer and less commonly studied than
either household income or poverty. In the Fragile Families and Child
Wellbeing Study (FFS), families are asked whether in the past twelve
months they faced any of the following circumstances because they did
not have enough money: did not pay rent or mortgage, did not pay utilities (gas, oil, or electric), had telephone service disconnected, had gas or
electricity turned off, received free food or meals, were hungry because
they did not have enough food, moved in with other people for financial
reasons, stayed in a shelter, were evicted from their homes, or had a medical need that went unmet. A large minority of poor families respond no
to all of these questions; a large minority of nonpoor families respond yes
to at least one. Families who respond yes to any question are clearly worse
off economically than those with the same income who respond no to all
of them. These questions thus tap a dimension of economic well-being
other than poverty. Positive responses to some questions, such as hunger
economic well-being33
or homelessness are well-described as material hardships, whereas others,
such as failing to pay a bill, may or may not translate into hardship.
Families with incomes above the poverty line find themselves in such
situations sometimes because they simply do not know how to manage
money, but more often because they are near poor or experience a drop in
income at some point in the year, or because they are close to or are living
beyond their means. Indeed 20 percent of families with incomes above
three times the poverty line experience a material hardship. Clearly, these
families are worse off than families with equivalent incomes. But, material
hardship may not be the best description of what they are experiencing.
Economic insecurity appears more apt in this case. Arguably it is even a
superior description for the experience of most of the poor who report
some form of material hardship. In this context, the U.S. Department of
Agriculture (USDA) eighteen-item scale is labeled food insecurity rather
than food hardship. For these reasons, we use these material hardship
questions to measure economic insecurity. More detailed descriptions of
each of the measures studied here are available in the appendix. By looking at poverty and economic insecurity as well as family income, we can
get a more complete picture of family economic well-being during good
times and bad.
We first look at the economic well-being of families over the previous
twelve months when children were approximately one, three, five, and
nine years old. This corresponds roughly to the first decade of the twentyfirst century, ending with the Great Recession. Our purpose here is to
document families’ levels of economic well-being over the decade, as well
as how family well-being varied by social class. We then describe the effects
of the Great Recession on families’ economic well-being over the previous
twelve months, harnessing the copious data we have on families over two
recessionary periods. The core questions we seek to answer are how the
Great Recession affected the economic well-being of vulnerable families,
and how the experiences of these families compare with those of their
more-advantaged counterparts?
Our analyses indicate that employment and household incomes
increase steadily as education increases, but that the largest gap is between
those with a college degree and everyone else. We find nearly as large a difference in income when comparing white families with black and Hispanic
families and married-parent families with cohabiting and single-mother
families. The Great Recession exacerbated these differences in income by
imposing the largest percentage losses in income on the most vulnerable
groups—families formed by poorly educated, minority, and unmarried
parents. At the same time, we find that the Great Recession also narrowed
the gaps in poverty and especially insecurity rates between the more privileged and more vulnerable groups by spreading economic distress to
better-off families, in particular those with some postsecondary education.
34
children of the great recession
PRIOR RESEARCH ON RECESSIONS
AND ECONOMIC WELL-BEING
As macroeconomic conditions worsen, families’ economic circumstances
suffer. It is well established that recessions lead to more weeks of unemployment, lower average weekly earnings, lower family income, and more
poverty.1 The Great Recession is notable for the depth and severity of the
labor market crisis relative to past recessions and is the largest recession in
the United States since the Great Depression. Recent research suggests
that the relationship between unemployment and poverty during the
Great Recession was similar to that observed in earlier recessions.2 Thus,
because unemployment was higher, the effects of the Great Recession
were probably even more severe than in prior recessions.
The Great Recession had a significant impact on the economic wellbeing of American workers and their families. Mean household income
fell from 2007 to 2009 by about 2.9 percent, and median household
income fell by 3.7 percent.3 Long-term unemployment and poverty also
increased substantially, particularly among families with young children.4
Nearly 40 percent of U.S. households reported unemployment, negative
equity in housing values, or falling behind in their house payments during
the period.5
As expected, an increase in unemployment and an accompanying
decrease in income leads to increases in poverty and decreases in families’
ability to put food on the table, keep the lights on, keep current with
housing payments, and afford necessary medical care. Research on the
Great Depression shows that individual unemployment increased material
hardship; hardship likewise increased during the Great Recession.6 We
also know that food insecurity increased in response to increasing unemployment during the Great Recession, as did homelessness and household crowding and the closely related measures of consumption poverty.7
Despite a general (and expected) understanding that hardship-insecurity
increases when economic conditions decline, we know less about whether
vulnerable families, such as those who have little education and young
children, are hit harder by recessions than other groups.
However, some evidence indicates that individuals and families
with lower income and lower levels of education were hit hardest during the Great Recession—at least in terms of labor market outcomes.
Unemployment affected less-educated, low-wage workers more strongly
than other workers, especially among men.8 Estimates from the Current
Population Survey suggest that among families in the lowest 10th percentile
of the income distribution, unemployment rates were as high as 31 percent
between October and December of 2009.9 Unemployment for collegeeducated individuals rose by only 3 percentage points from 2006 to
2010, whereas for those with a high school diploma or less, it rose by
economic well-being35
nearly 7 and 9 percentage points respectively.10 This finding suggests that
increases in poverty and insecurity, and decreases in household income,
may be particularly pronounced in families whose parents have only limited
educational credentials.
We go beyond previous research by using longitudinal data that follow
the same families over the first nine years of the child’s life, which happens
to coincide with the first decade of the twenty-first century. We focus
exclusively on families with children, highlight the diversity of family
experiences, and pay special attention to families who were vulnerable
before the onset of the Great Recession.
ECONOMIC WELL-BEING OF FAMILIES
FROM BIRTH THROUGH AGE NINE
We begin by describing mother’s and father’s employment and then
describe trends over time, or trajectories, for household income, poverty,
and material hardship or economic insecurity.
Figures 2.1 and 2.2 plot mother’s and father’s employment by the
child’s age over time for the four groups of families based on mother’s educational attainment. We expect employment rates of mothers to increase
over time as their children age and need less child care. In general, we see
increasing employment for all groups of mothers as their children age. The
biggest increases are for the more poorly educated mothers. But both at
Percent of Mothers Employed
Figure 2.1 Maternal Employment
100
90
80
70
60
50
40
30
20
10
0
College +
Some college
High school
Less than
high school
1
3
5
(1999–2001) (2001–2003) (2003–2006)
9
(2007–2010)
Child’s Age-Year
Source: Authors’ calculations.
Note: Sample is restricted to mothers interviewed in all survey waves (n = 2,986). Figures
are weighted.
36
children of the great recession
Percent of Fathers Employed
Figure 2.2 Paternal Employment
100
90
80
70
60
50
40
30
20
10
0
College +
Some college
High school
Less than
high school
1
3
5
(1999–2001) (2001–2003) (2003–2006)
9
(2007–2010)
Child’s Age-Year
Source: Authors’ calculations.
Note: Sample is restricted to mothers interviewed in all survey waves (n = 2,986). Figures
are weighted.
age one and age nine, mothers with some education after high school and
with college degrees have the highest levels of employment. Employment
rates for the two most poorly educated mothers decline between age five
and nine, probably reflecting the effect of the Great Recession.
As expected, fathers’ employment rates are substantially higher than
those of mothers, ranging from 75 percent to 99 percent, versus the
41 percent to 73 percent of mothers (depending on the year and education
level). Fathers’ patterns by education are similar to those for mothers—
lower employment rates for the more poorly educated, though the differences across education groups are less pronounced. Employment rates for
fathers are flat over time, but the rate for the most poorly educated fathers
drops somewhat at age nine, perhaps reflecting a Great Recession effect.
Figure 2.3 shows the mean household income trends as the child ages.
Two patterns stand out. First, household income increases steadily with
education, the gap between families with college-educated mothers and
other families being especially large. Income for families with a mother
without a high school diploma ranges from about $36,000 to $44,000.
Income for families with a high school–educated mother, a mother with
some college, and a mother with a college degree or higher are respectively
about $47,000, $65,000, and $158,000. Second, household incomes
increase over time as the parents and their children grow older. Although
income for the college educated appears to peak at age three, the difference between three and five is not significant. The absence of an income
drop between the year five and year nine interviews may appear surprising
economic well-being37
Dollars of Household Income
Figure 2.3 Household Income ($2010)
250,000
College +
200,000
150,000
Some college
100,000
High school
50,000
Less than
high school
0
1
3
5
(1999–2001) (2001–2003) (2003–2006)
9
(2007–2010)
Child’s Age-Year
Source: Authors’ calculations.
Note: Sample is restricted to mothers interviewed in all survey waves (n = 2,986). Figures
are weighted.
at first blush. The parents in these families, though, are getting older and
therefore can be expected to earn more. Furthermore, a portion of the age
nine sample is interviewed before the Great Recession began and an even
larger portion is interviewed in the early days of the Great Recession, so
they may not have yet experienced the full effect of the recession on their
household incomes. Finally, even in the relatively good times of the first
decade of the twentieth century, big family income declines and fluctuations in household income are quite common.
That large income drops are relatively common can be seen in figure 2.4.
We examine income loss by displaying the proportion of our sample that
experienced large, moderate, and small incomes losses and gains between
interview years. Large gains or losses are those greater than 40 percent,
moderate ones are between 10 percent and 40 percent, and small ones are
less than 10 percent (labeled as no change). Twenty-seven percent of the
sample saw a 40 percent gain in income, and close to another 25 percent
saw 10 percent to 40 percent gains between years one and three. Still, more
than one in ten lost 40 percent or more of their total income. By contrast,
between years three and five, 17 percent lost 40 percent of their income
and nearly another quarter lost between 10 percent and 40 percent. Large
losses between year five and year nine interviews are actually a bit less common than losses between years three and five, though overall losses were
generally equal between three and five and five and nine. When we limit
our sample to families who completed the year nine interview later in time
(the fall of 2009 or early 2010), we see that a much larger percentage of
38
children of the great recession
Percent of Households Experiencing
Each Income Change
Figure 2.4 Big Gains and Losses
100
90
27
22
80
70
60
35
17
14
16
13
18
23
22
50
40
31
23
30
23
22
20
15
10
13
17
13
1 (1999–2001) to
3 (2001–2003)
3 (2001–2003) to
5 (2003–2006)
5 (2003–2006) to
9 (2007–2010)
0
+ 40 percent
18
20
+ 10–40 percent
No change
− 10–40 percent
− 40 percent
5 (2003–2006) to
9 (Only Fall
2009/2010)
Child’s Age-Years
Source: Authors’ calculations.
Note: Sample is restricted to mothers interviewed in all survey waves (n = 2,986). Figures
are weighted.
families, 20 percent to 13 percent, saw their incomes decrease 40 percent
or more. These findings suggest that the income drop associated with the
Great Recession will be larger than that of the 2001 recession. We observe
income drops similar in magnitude to those from 2001, but our data do
not include the postrecession years (up to two years) when we would likely
have seen the largest income drops.
In short, big drops as well as big increases in income are quite common
for urban families with young children. The Great Recession, as we have
seen, made big income losses even more common, but was not uniquely
responsible for the poor economic conditions of these families.
Figures 2.5 and 2.6 display poverty and economic insecurity trajectories by mother’s education. Both indicators are highest for the least educated and decline steadily as education increases, the largest gap, as with
income, occurring between families with a college-educated mother and
other families. More than one-third of families in which the mother has
less than a high school diploma are poor in some year. Only 1 percent to
2 percent of families in which the mother has a college degree are poor.
Economic insecurity is more common than poverty among all groups
of families. Even for families in which the mother has some college,
insecurity-hardship rates are over 40 percent. Whereas poverty rates over
time are generally steady or declining, insecurity rates increase for all groups
economic well-being39
Percent of Households in Poverty
Figure 2.5 Poverty Rates
40
35
College +
30
Some college
25
20
High school
15
10
Less than
high school
5
0
1
3
5
(1999–2001) (2001–2003) (2003–2006)
9
(2007–2010)
Child’s Age-Year
Source: Authors’ calculations.
Note: Sample is restricted to mothers interviewed in all survey waves (n = 2,986). Figures
are weighted.
Figure 2.6 Hardship (Insecurity) Rates
Percent of Households
Experiencing Insecurity
60
50
College +
40
Some college
30
High school
20
Less than
high school
10
0
1
3
5
(1999–2001)(2001–2003)(2003–2006)
9
(2007–2010)
Child’s Age-Year
Source: Authors’ calculations.
Note: Sample is restricted to mothers interviewed in all survey waves (n = 2,986). Figures
are weighted.
40
children of the great recession
Dollars of Household Income
Figure 2.7 Household Income ($2010) by Race-Ethnicity
140,000
White
120,000
Married
Hispanic
Cohabiting
Black
Single
100,000
80,000
60,000
40,000
20,000
0
1
3
5
(1999–2001) (2001–2003) (2003–2006)
9
(2007–2010)
Child’s Age-Year
Source: Authors’ calculations.
Note: Sample is restricted to mothers interviewed in all survey waves (n = 2,986). Figures
are weighted.
between years five and nine. It may appear puzzling that insecurity rates
would increase when average income is not decreasing, but if people anticipate future increases in earnings and overextend their expenses or debts,
hardship rates might increase.
Finally, to illustrate how other characteristics linked with disadvantage
are related to the economic circumstances and experiences of families,
figure 2.7 displays trajectories of household income by race-ethnicity and by
parents’ relationship status at the time the child was born. We focus on three
racial-ethnic groups (black, Hispanic, and white) and on three relationship
statuses (married, cohabiting, single). At the top of figure 2.7, with the highest incomes, are white families whose incomes range from about $110,000
to $130,000. The incomes of black and Hispanic families are relatively similar, about $48,000 and $56,000, respectively—less than 50 percent of white
household income. In terms of relationship status, families in which the parents were married at birth have the highest incomes, more than twice that
of cohabiting or single-mother families, about $63,000 to $75,000 higher
than the incomes of cohabiting-parent or single-mother families.
Although not shown, trajectories for poverty and insecurity by raceethnicity and family structure were similar to the patterns shown in figures
2.5 and 2.6. The college-educated group is the best off, followed by white
and then married-parent families. Black and Hispanic families are near the
bottom, and single-mother families always fare the worst.
In short, poorly educated, minority, and single-mother families have
the lowest incomes, the highest poverty rates, and the highest rates of
economic insecurity. How do recessions affect these families? Would they
economic well-being41
be disproportionately hard hit by a big recession, making the gaps we find
at age nine even larger than they might otherwise have been? We turn to
these questions next.
EFFECTS OF THE GREAT RECESSION
ON ECONOMIC WELL-BEING
To estimate the effects of a deep recession, such as the Great Recession, on
families’ economic conditions, we take advantage of the vast differences in
local unemployment rates among our respondents during the first decade
of the twenty-first century. As we have seen, these families lived through
the dot-com recession, a tepid recovery, and then the Great Recession.
The relatively good as well as the bad economic times are captured in these
data. As explained in chapter 1, we begin by estimating the relationship
between local unemployment rates and mothers’ and fathers’ employment
and our three indicators of economic well-being, net of a host of demographic characteristics, including mother’s age, race-ethnicity, relationship
status at birth, immigrant status, whether she grew up with both parents,
survey year, and family’s city of residence (see table 2.A1 for a detailed
example of our analyses with and without individual fixed effects). The
local unemployment rate during the month of the interview is used when
employment last week is the outcome variable. The average of the local
unemployment during the last year is used when the outcome is family
income, poverty, and economic insecurity during the past year. We then
use our estimates to predict what the economic well-being of our families
would be given an increase in the unemployment rate from 5 percent to
10 percent, which is approximately the size of the increase brought about
by the Great Recession. More detail on our methodological approach and
the regression coefficients are available in the appendix. We also examine
a number of different specifications that are reported and discussed in
the appendix (see tables 2.A2 and 2.A3). We find little evidence that the
associations between the unemployment rate and the outcomes of interest
were significantly or substantively different during the Great Recession.
Figure 2.8 displays the simulated effects of the Great Recession on mother’s and father’s level of employment. For fathers, the predicted decline in
employment is 11 percentage points, a 15 percent loss. The largest losses are
for fathers with the least education, a high school diploma or less, which is
consistent with other research. The estimated differences between groups,
however, are not statistically significant. For mothers as a whole, a big recession is predicted to decrease employment by about 9 percentage points,
or 14 percent. The smaller employment loss for mothers is also consistent
with prior research. Although differences across groups are not statistically
significant—because of small sample size, we think—two differences are
worth noting. First, consistent with prior research, college-educated mothers’ employment stands apart from the other groups in that they see no loss or
42
children of the great recession
Predicting Percent of Employment
Figure 2.8 Employment by Education
100
90
80
70
60
50
40
30
20
10
0
UR 5 percent
–11%
–9%
UR 10 percent
–15%
+3%
–11%
Some
college*
College +
All*
–14%
–10%
–12%
–9%
–5%
All*
Less than
high
school
High
school*
Mother’s Employment
Less than
high
school
High
school*
Some College +*
college*
Father’s Employment
Source: Authors’ calculations.
Note: UR = unemployment rate. Predictions based on fixed-effects regressions controlling
for time. Chow tests find no statistically significant differences between groups.
*p < .05 between UR and employment
gain in employment. Second, those with some education after high school
see the largest losses in employment. This pattern appears in other outcomes.
The effects of the Great Recession on mean family income by mother’s
education are shown in figure 2.9. The families with less-educated mothers have the highest percentage losses; families in which the mother had
less than a college degree lose 14 percent to 20 percent of their income.
Families in which the mother had a college education or more lose a much
smaller proportion of their income—5 percent. These differences are both
large and consistent with findings in other studies based on repeated cross
section data, but again are not statistically significant across groups.11
Figure 2.10 displays the effects of the Great Recession on income loss
by mothers’ race-ethnicity and relationship status at birth. Black and
Hispanic families are hit only slightly harder than their white counterparts.
Families in which the parents are cohabiting or living apart at birth have
greater losses than families with married parents. Those who are single or
cohabiting at birth have a predicted loss of about 21 percent, more than
twice that of families in which the mother is married when the child was
born. This pattern is the same as for mothers’ education: those who are
already disadvantaged see the greatest percentage losses in income.
Figure 2.11 depicts the impacts of the Great Recession on poverty. As
with income, families in which the mother has a college education or more
see little to no change in poverty. Indeed, the effect is negative, though
not statistically different from zero. Among those with less than a college
education, poverty rates increase as mother’s education increases, though
economic well-being43
Predicted Dollars of Household Income
Figure 2.9 Income by Education
160,000
–5%
140,000
120,000
100,000
80,000
60,000
UR 5 percent
–15%
–13%
UR 10 percent
–20%
–14%
40,000
20,000
0
All*
Less than
High
high school* school*
Some
college*
College +
Source: Authors’ calculations.
Note: UR = unemployment rate. Predictions based on fixed-effects regressions controlling
for time. Chow tests find no statistically significant differences between groups.
*p < .05 between UR and income
absolute poverty rates are highest among the least educated. As a consequence of the Great Recession, families in which the mother has more than
high school diploma but less than a college education see an astonishing
75 percent increase in poverty (from 8 percent to 14 percent). Thus, although
families with more-educated mothers continue to have lower poverty rates
than families with less-educated mothers, the Great Recession has the net
Predicted Dollars of
Household Income
Figure 2.10 Income by Race-Ethnicity and Relationship Status
100,000
90,000
80,000
70,000
60,000
50,000
40,000
30,000
20,000
10,000
0
–17%
–20%
Black*
–9%
–19%
Hispanic*
–20%
White
Race or Ethnicity
Married
Cohab*
–20%
UR 5 percent
UR 10 percent
Single*
Relationship Status
Source: Authors’ calculations.
Note: UR = unemployment rate. Predictions based on fixed-effects regressions controlling
for time. Chow tests show that the coefficient for unemployment for married mothers is
statistically different (p < .05) from cohabiting and single mothers.
*p < .05 between UR and income
44
children of the great recession
Predicted Percent of
Households in Poverty
Figure 2.11 Poverty Rate by Education
40
35
30
25
20
15
10
5
0
+42%
+63%
+56%
UR 5 percent
+75%
UR 10 percent
–33%
All*
Less than
High
high school* school*
Some
college*
College +
Source: Authors’ calculations.
Note: UR = unemployment rate. Predictions based on fixed-effects regressions controlling
for time. Chow tests show not statistically significant differences across groups.
*p < .05 between UR and poverty
effect of reducing educational differentials in the proportion of families
experiencing poverty—except for college-educated mothers.
Figure 2.12 depicts the effects of a big recession on economic insecurity
rates. As is true of poverty, we again see that the group hardest hit by the
Great Recession is families with mothers with some education after high
school. The increase in economic insecurity for mothers with some education after high school was 24 percentage points, or 56 percent. The effect
Predicted Percent of Households
Experiencing Insecurity
Figure 2.12 Hardship by Education
70
60
+26%
+16%
+24%
+56%
50
40
+47%
30
UR 5 percent
UR 10 percent
20
10
0
All*
Less than
High
high school school*
Some
college*
College +
Source: Authors’ calculations.
Note: UR = unemployment rate. Predictions based on fixed-effects regressions controlling
for time. Chow tests show that differences between some college and mothers with less
than a high school degree are significantly different.
*p < .05 between UR and hardship
economic well-being45
of the Great Recession was to equalize economic insecurity rates among
the three lowest education groups. College-educated mothers also see a
large increase in hardship, 47 percent, but from a much lower base. The
increase is not statistically significant, but as before, this is likely a result of
a small sample. Even if we take the increase at face value, the rate for the
college educated is less than half that for the other three groups.
We also examine the differential effects of recessions by race-ethnicity
and family relationship status at birth (regression results are available in
table 2.A4). As a consequence of the Great Recession, mothers who were
cohabiting or single at birth and blacks and Hispanics lose about 20 percent of income—a loss comparable to that of those with a high school
diploma or less. Married mothers and white mothers have losses on average similar to those of college-educated mothers, indistinguishable from
zero. Poverty rates go up for all groups, though the increases for married
mothers and Hispanic mothers are small and not statistically significant.
We suspect the Hispanic estimate is biased by attrition, because we know
that Hispanic families are more likely to attrite, and perhaps those who
lose their jobs are more likely to return to their native countries (if they
are immigrants). What is striking is how large the increase in poverty is
for single mothers, about 1.25 times larger than for the poorly educated
mother groups, suggesting a special vulnerability to increased poverty
from recessions for single-mother families.
Our results show that the most disadvantaged families see somewhat
smaller losses in income and much smaller increases in poverty and economic insecurity than the two middle groups. The smaller losses for the
most poorly educated groups are likely a result of the lower employment
rates among this group. Those who are already out of work cannot lose
earnings from a recession. Similarly, for these disadvantaged families, we
find smaller increases in poverty and economic insecurity because so many
of these families live in poverty and experience economic insecurity even
in the best of times. The poor may get poorer (lose income), but they were
poor before and during (and likely after) the recession.
In sum, families not protected by a college education (or marriage,
or white skin color) are the hardest hit in recessionary times. We find, in
particular, that mothers with more than a high school education but less
than a college degree are particularly vulnerable: the largest decrease in
employment, a poverty rate increase of 75 percent, and an insecurity rate
increase of nearly 60 percent.
Does the Great Recession exacerbate already large economic differences between vulnerable and comfortable families with children? When
comparing families with a college-educated mother with families with
less-educated mothers, the answer is yes. When comparisons are limited
to the three less-educated mother groups, the Great Recession narrowed
the gap between the vulnerable and somewhat more privileged groups by
spreading distress to the latter groups.
46
children of the great recession
APPENDIX
Measures
Employment. Mother’s employment and biological father’s employment is a measure indicating whether mothers or fathers were employed
at the time of the interview. Following the Bureau of Labor Statistics,
the parents are asked, “Last week, did you do any regular work for
pay?” If they report working or being on vacation, they are considered
employed.
Household income. Household income is a measure of mother’s household income in 2010 dollars. Mother’s total household income is calculated using the sum of the component parts of income: her earnings,
partner’s earnings, various government transfers, and child support. We
use TAXSIM to estimate the amount of the Earned Income Tax Credit
mothers would have received and add that to income (more details on
the EITC estimation are available in chapter 3).12 We also include the
near cash benefit—the Supplemental Nutrition Assistance Program in
our measure of income. Finally, we include a measure of private financial
transfers—money received from friends or family. Mothers also report
on their total household income in a single item measure. We use the
higher value of the single report or the sum of the components to create
a measure of income after transfer (analyses run on the single household
income measure and the measure without transfers were substantively
the same). We also study income before transfers, when all transfers are
subtracted from the income measure. We log the income variables in our
analyses.
Poverty. The household’s income-to-needs ratio is constructed using
the Census Bureau’s official poverty thresholds, which are adjusted by
family composition and year. Households are considered poor if they have
an income-to-needs ratio of 1 or less. We construct measures of poverty
using both pretransfer and post-transfer income.
Material hardship–economic insecurity. Material hardship measures
whether families go without basic needs in five domains: bills, utilities,
food, medical care, and housing. Specifically, families are asked whether
in the past twelve months they faced any of the following circumstances
because they did not have enough money: did not pay rent or mortgage,
did not pay utilities (gas, oil, or electric), had telephone service disconnected, had gas or electricity turned off, received free food or meals, were
hungry because they did not have enough food, moved in with other
people for financial reasons, stayed in a shelter, were evicted from their
homes, or had a medical need that went unmet. If families reported experiencing any of the ten hardship measures, they received a 1 on the hardship variable.
economic well-being47
Supplemental Analyses
A number of additional analyses tested the association between the unemployment rate and economic well-being. First, analyses including an interaction term with the unemployment rate and the year nine wave of data
collection test whether the association between the unemployment rate
and outcomes differed during the recession. In none of those analyses is
the year nine interaction term statistically significant, suggesting that the
link between unemployment and economic well-being was not distinct in
the Great Recession (see table 2.A2).
Second, to test whether the rate of change in the unemployment rate
is more closely related with economic well-being, spline models distinguished between an annual declining rate of change in the unemployment
rate and an annual increasing rate of change in the unemployment rate
and economic outcomes. Few associations are significant using the rate
of change models, and the main coefficient on unemployment does not
change from the model without rate of change indicators (see table 2.A3).
Third, instead of studying the link between the unemployment rate and
economic outcomes, we use the consumer confidence index and the foreclosure rate as indicators of the Great Recession. No associations between
the consumer sentiment index and economic outcomes are significant.
The foreclosure rate is significantly associated with economic well-being,
and the findings are very similar to those of the unemployment rate.
Fourth, additional analyses focus on years five and nine. After constructing a measure of an income drop between years five and nine, we regress
year nine economic outcomes on a measure of a 1 to 40 percent drop in
income and a 40 percent plus drop in income. These findings, as anticipated,
show that families whose incomes declined also saw a drop in economic
well-being (more hardships for example), and that the larger drop is linked
with even higher odds of hardship than the smaller drop. We also construct
income drops between waves for the other survey years and find that large
drops between survey waves are linked with higher odds of hardship.
Fifth, we consider a change in the unemployment rate between years five
and nine on year nine outcomes, distinguishing increases and decreases in
unemployment. These analyses, as expected, generally show that a decline
in unemployment is linked with better economic outcomes and an increase
is linked with poorer ones.
Sixth, we run models lagging the unemployment rate. In the first, the
average unemployment rate over the prior year is lagged two and three
years. In the second, we include the unemployment rate at the interview,
a twelve-month lag, a twenty-four-month lag, and a thirty-six-month lag.
The models show no evidence of a lag in the association between the
unemployment rate and the economic outcomes.
48
children of the great recession
Table 2.A1 Full Regression Results, Material Hardship
With Individual
Fixed Effects
Unemployment rate
Education
Less than high school
High school
Some college
Relationship status
Married
Cohabiting
Mother’s age
Race-ethnicity
Black
Hispanic
Other
Immigrant
Number of children in household
Lived with both parents at age fifteen
Interview year
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
1.17**
0.59**
0.83
0.66*
0.83
0.59**
0.73†
1.71
0.93
0.98
0.77
0.63
(5.02)
(-2.69)
(-1.28)
(-2.38)
(-1.07)
(-2.95)
(-1.90)
(1.00)
(-0.32)
(-0.14)
(-1.26)
(-1.37)
Without Individual
Fixed Effects
1.13**
(4.10)
2.85**
2.33**
2.72**
(6.94)
(5.92)
(6.97)
0.58**
1.11*
1.00
(-6.34)
(1.98)
(-0.34)
1.22*
0.96
1.20
0.77†
1.08**
0.77**
(2.32)
(-0.43)
(1.43)
(-1.87)
(3.02)
(-4.82)
0.70**
0.87**
0.72**
0.84*
0.70**
0.77**
0.89
0.92
1.02
0.82
0.77
(-4.49)
(-3.16)
(-3.87)
(-1.99)
(-3.41)
(-3.25)
(-0.35)
(-1.34)
(0.16)
(-1.26)
(-0.97)
economic well-being49
Table 2.A1 Continued
With Individual
Fixed Effects
City
Austin
Baltimore
Detroit
Newark
Philadelphia
Richmond
Corpus Christi
Indianapolis
Milwaukee
New York
San Jose
Boston
Nashville
Chicago
Jacksonville
Toledo
San Antonio
Pittsburgh
Norfolk
Constant
Observations
Number of individuals
8,392
2,280
Without Individual
Fixed Effects
1.47**
(10.58)
1.07
(0.91)
1.14*
(2.37)
0.89**
(-3.12)
1.05
(0.73)
1.27**
(2.62)
(-1.80)
0.89†
1.38**
(4.48)
1.25**
(4.26)
0.83**
(-4.88)
0.74**
(-4.70)
1.34**
(6.42)
1.06
(0.90)
0.82**
(-3.80)
0.74**
(-4.72)
1.02
(0.32)
1.33**
(4.32)
1.16*
(2.13)
1.00
(0.03)
0.23**
(-7.94)
15,860
Source: Authors’ calculations.
Note: Figures reported are odds ratios. Z-stats in parentheses. Covariates are measured at the baseline
survey (except year) and are clustered at the city and individual level. Model includes level unemployment rate. The model without individual fixed effects is clustered at city and individual level.
**p < .01; *p < .05; †p < .1
Mother’s employment odds ratios (z-stat)
Unemployment rate
0.89**
0.93
(model 1)
(-3.82)
(-1.48)
Unemployment rate
0.89**
0.94
(model 2)
(-3.73)
(-1.39)
Increasing
1.00
1.00
unemployment rate
(0.52)
(0.40)
Decreasing
1.00
1.00
unemployment rate
(-0.50)
(0.39)
Observations
8,446
3,587
Number of
2,301
991
individuals
Father’s employment odds ratios (z-stat)
Unemployment rate
0.85**
0.84**
(model 1)
(-3.77)
(-2.63)
Unemployment rate
0.84**
0.82**
(model 2)
(-3.81)
(-2.83)
Increasing
1.00
1.00
unemployment rate
(-0.52)
(-1.22)
Decreasing
1.00
0.99
unemployment rate
(-0.33)
(-0.40)
Observations
3,924
1,724
Number of
1,181
522
individuals
All
Less than
High
School
0.80**
(-3.42)
0.80**
(-3.49)
1.00
(-0.92)
1.00
(0.16)
1,856
502
0.86
(-1.56)
0.84†
(-1.77)
0.99†
(-1.67)
1.00
(-0.04)
891
264
0.88
(-1.45)
0.90
(-1.26)
1.01
(1.39)
0.99
(-0.55)
1,091
333
Some
College
0.87*
(-2.45)
0.87*
(-2.32)
1.00
(0.86)
0.98
(-1.25)
2,281
617
High
School
With Individual Fixed Effects
0.62*
(-2.21)
0.65†
(-1.90)
1.02
(1.43)
1.00
(-0.03)
218
62
1.04
(0.31)
1.04
(0.29)
1.01
(1.16)
0.98
(-0.94)
722
191
College +
0.89**
(-3.50)
0.88**
(-3.57)
1.00*
(-2.39)
1.00
(0.16)
11,588
0.93**
(-2.64)
0.93**
(-2.66)
1.00
(0.51)
1.00
(0.11)
15,851
All
Table 2.A2 Coefficients and Standard Errors, Rate of Change, Economic Outcomes
0.94
(-1.56)
0.93†
(-1.67)
1.00
(-1.48)
1.00
(-0.25)
4,236
0.96
(-1.05)
0.96
(-0.99)
1.00
(0.00)
1.01
(1.51)
6,126
Less than
High
School
0.84**
(-5.00)
0.84**
(-5.23)
1.00
(-0.25)
1.00
(0.70)
3,927
0.84**
(-3.46)
0.81**
(-3.21)
0.99**
(-3.94)
1.00
(0.20)
2,934
0.89†
(-1.83)
0.90
(-1.59)
1.00
(1.49)
1.00
(0.25)
2,864
Some
College
0.93*
(-1.98)
0.93†
(-1.92)
1.00
(0.31)
0.99
(-1.16)
4,061
High
School
Without Individual Fixed Effects
0.68**
(-3.44)
0.68**
(-3.06)
1.00
(0.31)
1.01
(0.17)
1,499
1.04
(0.43)
1.04
(0.43)
1.00
(1.10)
0.99
(-0.69)
1,733
College +
Log of household income
Unemployment rate
-0.04**
(model 1)
(0.01)
Unemployment rate
-0.03**
(model 2)
(0.01)
Increasing rate of
0.00**
unemployment
(0.00)
Decreasing rate of
0.00
unemployment
(0.00)
Observations
15,688
Number of
4,600
individuals
Poverty (odds ratios)
Unemployment rate
1.16**
(model 1)
(4.01)
Unemployment rate
1.14**
(model 2)
(3.52)
Increasing
0.99**
unemployment rate
(-3.32)
Decreasing
0.99
unemployment rate
(-0.83)
Observations
5,833
Number of
1,618
individuals
-0.05**
(0.01)
-0.05**
(0.01)
0.00†
(0.00)
-0.00
(0.00)
4,027
1,162
1.20*
(2.51)
1.20*
(2.43)
1.00
(-0.46)
1.01
(0.89)
1,571
433
-0.05**
(0.01)
-0.04**
(0.01)
0.00*
(0.00)
0.00
(0.00)
6,062
1,821
1.16**
(2.91)
1.13*
(2.39)
0.99**
(-3.30)
0.99
(-1.01)
3,277
916
1.17
(1.60)
1.16
(1.43)
0.99
(-1.05)
0.98
(-1.07)
920
251
-0.03*
(0.01)
-0.02†
(0.01)
0.00*
(0.00)
0.00
(0.00)
3,879
1,122
0.93
(-0.18)
0.92
(-0.19)
0.97
(-1.11)
0.81
(-1.15)
65
18
0.01
(0.06)
0.03
(0.07)
0.00
(0.00)
0.00
(0.00)
1,720
495
1.11**
(3.63)
1.10**
(3.44)
1.00**
(-2.68)
1.00
(-0.34)
15,656
-0.04**
(0.01)
-0.03**
(0.01)
0.00
(0.00)
-0.00
(0.00)
15,688
1.09*
(2.36)
1.08*
(2.07)
1.00**
(-3.07)
0.99
(-0.85)
6,045
-0.05**
(0.02)
-0.05**
(0.02)
0.00
(0.00)
0.00
(0.00)
6,062
1.12*
(2.20)
1.13*
(2.18)
1.00
(0.17)
1.01
(1.01)
4,004
-0.05**
(0.01)
-0.05**
(0.01)
-0.00
(0.00)
-0.00
(0.01)
4,027
0.81
(-0.67)
0.82
(-0.69)
0.99
(-0.39)
1.09
(0.93)
1,018
0.01
(0.06)
0.02
(0.07)
0.00
(0.00)
-0.00
(0.00)
1,720
(Table continues on p. 52.)
1.20**
(2.99)
1.19**
(3.03)
0.99
(-1.48)
0.99
(-0.76)
3,876
-0.03*
(0.01)
-0.03*
(0.01)
0.00*
(0.00)
0.00
(0.00)
3,879
1.17**
(5.02)
1.19**
(5.36)
1.00**
(2.75)
1.00
(0.54)
8,392
2,280
All
1.09†
(1.72)
1.10*
(2.00)
1.00†
(1.90)
1.01
(0.54)
3,523
971
Less than
High
School
1.18**
(2.68)
1.19**
(2.77)
1.00
(0.84)
1.01
(0.48)
2,341
631
High
School
1.30**
(4.13)
1.33**
(4.41)
1.01†
(1.74)
1.01
(1.11)
2,025
545
Some
College
With Individual Fixed Effects
1.16
(1.01)
1.16
(0.99)
1.01
(1.26)
0.95†
(-1.75)
503
133
College +
1.13**
(4.10)
1.13**
(4.20)
1.00
(1.23)
1.00
(0.05)
15,860
All
1.08*
(2.11)
1.09*
(2.25)
1.00
(1.08)
1.01
(1.18)
6,131
Less than
High
School
1.11**
(3.26)
1.11**
(3.17)
1.00
(0.77)
0.99
(-1.00)
4,064
High
School
1.23**
(6.09)
1.24**
(6.76)
1.00
(1.05)
1.01
(0.98)
3,925
Some
College
Without Individual Fixed Effects
1.08
(0.77)
1.07
(0.57)
1.00
(0.08)
0.96*
(-1.99)
1,740
College +
Source: Authors’ calculations.
Note: Standard errors and z-stats in parentheses. Model 1 includes the unemployment rate as a level. Model 2 includes unemployment rate as a level as well as rate of change in
unemployment rate. SEs for the OLS with fixed effects are clustered at city, for OLS and logistic models without fixed effects are clustered at city and individual.
**p < .01; *p < .05; †p < .1
Material hardship
Unemployment rate
(model 1)
Unemployment rate
(model 2)
Increasing
unemployment
Decreasing
unemployment
Observations
Number of
individuals
Table 2.A2 Continued
economic well-being53
Table 2.A3 Sensitivity of Coefficients, Economic Outcomes
With Individual
Fixed Effects
Mother’s employment odds ratios (z-stat)
Unemployment rate (model 3)
Individual unemployment
Unemployment rate (model 4)
Unemployment rate ∗ year nine
Father’s employment odds ratios (z-stat)
Unemployment rate (model 3)
Individual unemployment
Unemployment rate (model 4)
Unemployment rate ∗ year nine
Log of household income
Unemployment rate (model 3)
Individual unemployment
Unemployment rate (model 4)
Unemployment rate ∗ year nine
Poverty odds ratio (z-stat)
Unemployment rate (model 3)
Individual unemployment
Unemployment rate (model 4)
Unemployment rate ∗ year nine
Material hardship odds ratio (z-stat)
Unemployment rate (model 3)
Individual unemployment
Unemployment rate (model 4)
Unemployment rate ∗ year nine
Without Individual
Fixed Effects
—
—
0.89**
1.00
—
—
(-2.84)
(-0.00)
—
—
0.92*
1.02
—
—
(-2.19)
(0.50)
—
—
0.80**
1.09
—
—
(-3.63)
(1.27)
—
—
0.87**
1.04
—
—
(-3.69)
(1.19)
-0.03**
0.24**
-0.04**
0.01
(0.01)
(0.02)
(0.01)
(0.01)
-0.03**
0.43**
-0.05**
0.02
(0.01)
(0.02)
(0.01)
(0.01)
1.14**
0.40**
1.18**
0.97
(3.39)
(-13.16)
(3.28)
(-0.45)
1.08**
0.25**
1.12**
0.99
(3.02)
(-19.18)
(4.16)
(-0.20)
1.17**
0.92
1.15**
1.03
(4.95)
(-1.42)
(3.40)
(0.57)
1.12**
0.81**
1.09†
1.05
(4.03)
(-4.71)
(1.94)
(0.98)
Source: Authors’ calculations.
Note: Standard errors and z-stats in parentheses. Model 3 includes unemployment rate and a measure
of individual unemployment. Model 4 includes unemployment rate and an interaction between unemployment rate and year nine, when the Great Recession hit. SEs for OLS with fixed effects are clustered
at city, for OLS and logistic models without fixed effects are clustered at city and individual.
**p < .01; *p < .05; †p < .1
0.76*
(-2.39)
0.84*
(-2.15)
-0.04**
(0.01)
1.03
(0.54)
1.20**
(3.31)
0.88† (-1.93)
-0.05**
(0.01)
1.24**
(3.71)
1.20**
(3.67)
1.08
(1.01)
1.26*
(2.11)
-0.01
(0.04)
0.99
(-0.18)
White
0.91†
(-1.65)
Hispanic
0.86**
(-3.07)
Black
1.19*
(2.33)
1.09
(0.68)
0.00
(0.02)
0.73**
(-2.64)
0.96
(-0.55)
Married
1.20**
(3.66)
1.10
(1.57)
-0.05**
(0.01)
0.83**
(-2.80)
0.96
(-0.84)
Cohabiting
1.13*
(2.55)
1.25**
(4.19)
-0.05**
(0.01)
0.91
(-1.36)
0.79**
(-4.69)
Single
Source: Authors’ calculations.
Note: Standard errors and z-stats in parentheses. Model includes level unemployment rate, results include individual fixed effects and time. SEs for OLS
with fixed effects are clustered at city.
**p < .01; *p < .05; † p < .1
Material hardship odds ratio (z-stat)
Unemployment rate
Poverty odds ratio (z-stat)
Unemployment rate
Log of household income
Unemployment rate
Father’s employment odds ratio (z-stat)
Unemployment rate
Mother’s employment odds ratio (z-stat)
Unemployment rate
Table 2.A4 Coefficients and Standard Errors, Economic Outcomes
economic well-being55
NOTES
1. Blank and Blinder 1986; Blank 1989, 1993; Cutler and Katz 1991; Blank
et al. 1993; Tobin 1994; Haveman and Schwabish 2000; Freeman 2001;
Hoynes 2002; Gundersen and Ziliak 2004.
2. Bitler and Hoynes 2013.
3. Thompson and Smeeding 2013.
4. Smeeding et al. 2011.
5. Hurd and Rohwedder 2010.
6. On the Great Depression, Conger and Elder 1994; on the Great Recession,
Pilkauskas, Currie, and Garfinkel 2012
7. On food insecurity, Nord, Andrews, and Carlson 2009; on homelessness and
household crowding, Sard 2009; DeCrappeo et al. 2010; Painter 2010; Sell
et al. 2010; on consumption poverty, Meyer and Sullivan 2013.
8. Hout, Levanon, and Cumberworth 2011.
9. Sum and Khatiwada 2010.
10. Thompson and Smeeding 2013.
11. Bitler, Hoynes, and Kuka 2014.
12. TAXSIM is the National Bureau of Economic Research’s online program
for calculating liabilities under U.S. Federal and State income tax laws from
individual data, available at http://users.nber.org/~taxsim/ (accessed April
15, 2016).
REFERENCES
Bitler, Marianne, and Hilary Hoynes. 2013. “The More Things Change, the
More They Stay the Same? The Safety Net and Poverty in the Great Recession.”
NBER working paper no. w19449. Cambridge, Mass.: National Bureau of
Economic Research.
Bitler, Marianne, Hilary Hoynes, and Elira Kuka. 2014. “Child Poverty and the
Great Recession in the United States.” Innocenti occasional paper inwopa724.
Florence: UNICEF Innocenti Research Centre.
Blank, Rebecca M. 1989. “Disaggregating the Effect of the Business Cycle on the
Distribution of Income.” Economica 56(222): 141–63.
———. 1993. “Why Were Poverty Rates So High in the 1980s?” In Poverty
and Prosperity in the USA in the Late Twentieth Century, edited by Dimitri B.
Papadimitriou and Edward N. Wolff. New York: St. Martin’s Press.
Blank, Rebecca M., and Alan S. Blinder. 1986. “Macroeconomics, Income
Distribution, and Poverty.” In Fighting Poverty: What Works and What Doesn’t,
edited by Sheldon H. Danziger and Daniel H. Weinberg. Cambridge, Mass.:
Harvard University Press.
Blank, Rebecca M., David Card, Frank Levy, and James L. Medoff. 1993.
“Poverty, Income Distribution, and Growth: Are They Still Connected?”
Brookings Papers on Economic Activity 2 (1993): 285–339.
56
children of the great recession
Conger, Rand D., and Glen H. Elder Jr. 1994. Families in Troubled Times:
Adapting to Change in Rural America. Social Institutions and Social Change.
Hawthorne, N.Y.: Aldine de Gruyter.
Cutler, David M., and Lawrence F. Katz. 1991. “Macroeconomic Performance
and the Disadvantaged.” Brookings Papers on Economic Activity 2 (1991):
1–74.
DeCrappeo, Megan, Danilo Pelletiere, Sheila Crowley, and Elisabeth Teater.
2010. “Out of Reach 2010: Renters in the Great Recession, the Crisis
Continues.” Washington, D.C.: National Low Income Housing Coalition.
Freeman, Richard B. 2001. “The Rising Tide Lifts.” In Understanding Poverty,
edited by Sheldon H. Danziger and Robert H. Haveman. Cambridge, Mass.:
Harvard University Press.
Gundersen, Craig, and James P. Ziliak. 2004. “Poverty and Macroeconomic
Performance Across Space, Race, and Family Structure.” Demography 41(1):
61–86.
Haveman, Robert H., and Jonathan Schwabish. 2000. “Has Macroeconomic
Performance Regained its Antipoverty Bite?” Contemporary Economic Policy
18(4): 415–27.
Hout, Michael, Asaf Levanon, and Erin Cumberworth. 2011. “Job Loss and
Unemployment.” In The Great Recession, edited by David B. Grusky and Bruce
Western. New York: Russell Sage Foundation.
Hoynes, Hilary W. 2002. “The Employment, Earnings, and Income of Less
Skilled Workers over the Business Cycle.” In Finding Jobs: Work and Welfare
Reform, edited by David E. Card and Rebecca M. Blank. New York: Russell
Sage Foundation.
Hurd, Michael D., and Susann Rohwedder. 2010. “Effects of the Financial
Crisis and Great Recession on American Households.” NBER working paper
no. w16407. Cambridge, Mass.: National Bureau of Economic Research.
Meyer, Bruce D., and James X. Sullivan. 2013. “Consumption and Income
Inequality and the Great Recession.” American Economic Review 103(3):
178–83.
Nord, Mark, Margaret Andrews, and Steven Carlson. 2009. “Household Food
Security in the United States, 2008.” Economic Research Report no. 83.
Washington, D.C.: US Department of Agriculture.
Painter, Gary. 2010. “What Happens to Household Formation in a Recession?”
Washington, D.C.: Research Institute for Housing America.
Pilkauskas, Natasha. V., Janet Currie, and Irwin Garfinkel. 2012. “The Great
Recession, Public Transfers, and Material Hardship.” Social Service Review
86(3): 401–27.
Sard, Barbara. 2009. “Number of Homeless Families Climbing Due to Recession.”
Washington, D.C.: Center on Budget and Policy Priorities.
Sell, Katherine, Sarah Zlotnik, Kathleen Noonan, and David Rubin. 2010.
“The Effect of Recession on Child Well-Being: A Synthesis of the Evidence
by PolicyLab, the Children’s Hospital of Philadelphia.” Philadelphia, Pa.:
Children’s Hospital of Philadelphia, PolicyLab.
Smeeding, Timothy M., Jeffrey P. Thompson, Asaf Levanaon, and B. Esra Burak
Ho. 2011. “The Changing Dynamics of Work, Poverty, Income from Capital
and Income from Earnings during the Great Recession.” Working Paper presented at Stanford Poverty Conference on the Great Recession. Stanford, Calif.
(October 2012).
economic well-being57
Sum, Andrew, and Ishwar Khatiwada. 2010. “Labor Underutilization Problems
of U.S. Workers across Household Income Groups at the End of the Great
Recession: A Truly Great Depression Among the Nation’s Low Income
Workers Amidst Full Employment Among the Most Affluent.” Boston, Mass.:
Northeastern University, Center for Labor Market Studies.
Thompson, Jeffrey. P., and Timothy M. Smeeding. 2013. “Inequality and Poverty
in the United States: The Aftermath of the Great Recession.” FEDS working
paper no. 2013-51. Washington, D.C.: Board of Governors of the Federal
Reserve System Finance and Economics Discussion Series.
Tobin, James. 1994. “Poverty in Relation to Macroeconomic Trends, Cycles, and
Policies.” In Confronting Poverty: Prescriptions for Change, edited by Sheldon
H. Danziger, Gary D. Sandefur, and Daniel H. Weinberg. Cambridge, Mass.:
Harvard University Press.
Chapter 3
Public and Private Transfers
Natasha Pilkauskas and Irwin Garfinkel
T
he previous chapter documents great differences in employment,
income, poverty, and economic insecurity by education, race-ethnicity,
and marital status for families with children born at the turn of the century.
Income increases steadily—from about $41,000 to $50,000 to $76,000—
as education increases from less than a high school degree, to a high school
degree, to some education after high school, and then leaps dramatically
to $180,000 for those with a college degree. Poverty and especially economic insecurity were both common among the poorly educated, racial
minorities, and female-headed families and uncommon among collegeeducated, white, and married families. The Great Recession exacerbated
existing differences in incomes and poverty rates between those with
and without a college degree, but also narrowed the gap in economic
insecurity rates. The recession also reduced gaps in poverty and economic insecurity-hardship rates between the three least-educated groups,
increasing these rates for those with some postsecondary education but no
bachelor’s degree.
That such a large proportion of families with children experience poverty
and economic insecurity suggests that many families will also rely on both
public and private transfers to make ends meet. Recessions reduce earnings
in the labor market, but many government programs are designed to help
families cope with such shocks. At the same time, friends and family members may pitch in to assist struggling loved ones. In this chapter, we examine
assistance families receive from public transfers targeted at the poor, lowincome, and unemployed—including Medicaid, the Earned Income Tax
Credit (EITC), Supplemental Nutrition Assistance Program (SNAP, commonly known as Food Stamps), Temporary Assistance for Needy Families
(TANF), housing assistance, Supplemental Security Income (SSI), and
Unemployment Insurance (UI)—and private transfers—including both
cash and in-kind housing assistance from relatives and friends. As in the
previous chapter, we first describe transfer receipt patterns from age one
to age nine and then examine the effects of the Great Recession on public
and private transfers.
The seven public and two private transfers we examine are those most
commonly received among low-income families with young children, as
public and private transfers59
well as the largest in dollar value. Medicaid is a health insurance program
for low-income families funded by both federal and state governments.
Eligibility and generosity of benefits vary by state. In the most generous states families with incomes over 200 percent of the poverty line are
eligible for benefits, though the median is 133 percent of poverty (or
about $23,500 and $15,650 respectively, in 2014 dollars). The EITC
is a refundable federal income tax credit program for low- to moderateincome families. Some states also run their own EITC programs. We
take account of both federal and state EITC benefits. SNAP is a federally funded program that offers near-cash assistance for the purchase of
food for low-income families up to 1.3 times the poverty level. Housing
assistance is federally funded and may come in the form of public housing,
where eligible low-income families live in government-owned property,
or of Section 8 housing, where the government pays a portion of rent on
behalf of low-income families. TANF is a cash assistance program for lowincome, mostly single-mother families funded both federally and by states.
The proportion of single-mother families who benefit from the program
has plummeted dramatically since the 1996 welfare reform, which implemented strict work requirements and a lifetime limit of sixty months of
assistance. SSI, a federally funded program, provides cash assistance to
low-income people who are older than sixty-five or are blind or disabled.
Children who are disabled and live in low-income families may qualify
for SSI. Last, unemployment insurance provides individuals who work in
covered employment and become unemployed with a payment equal to
a portion of their prior salary. Unemployment insurance is not limited to
families with low incomes, but workers must lose their jobs through no
fault of their own (be laid off) and have to meet other work-level requirements (such as enough quarters employed before unemployment) to be
eligible for UI. State governments normally fund UI, but during recessions the federal government often provides extended benefits.
As noted, six of the seven public transfers we examine are targeted
at low-income families and unemployment insurance is targeted at the
unemployed. The respondents report receipt of benefits, and the annual
dollar amounts, with the exception of the EITC, which is estimated using
information on family earnings and household composition. The annual
dollar values of Medicaid are estimated using information on the value of
the program the government reports it. Housing assistance is estimated
by evaluating the difference between what mothers report paying in rent
and the fair market housing value in their city of residence. All values are
given in 2010 dollars. More details on the measurement of each public
transfer, the EITC estimation, and the dollar values of Medicaid and housing assistance are provided in the appendix.
Respondents also report about two types of private transfers—private
financial transfers and doubling up. Private financial transfers, or money
60
children of the great recession
from friends and family, are measured in annual dollars received. In-kind
housing assistance (also known as doubling up) indicates whether a family
is living with other non-nuclear family adults (such as a parent, sibling, or
unrelated person). We estimate the dollar value of doubling up by comparing rental payments among the doubled up with those who are not
doubled up. More detail on the measurement and construction of these
variables is presented in the appendix.
Not surprisingly, we find that many families rely heavily on both public
and private transfers and that a recession increases families’ likelihood of
receiving such transfers. Private transfers make up for a miniscule portion
of lost income. Public transfers increase income a great deal—nearly 30 percent among the most vulnerable groups. Public transfers also contributed
notably to preventing increases in poverty and economic insecurity.
PREVIOUS LITERATURE ON RECESSIONS AND TRANSFERS
We know that public income transfer programs respond in hard times
to help families meet their economic needs, and the Great Recession
was no exception. The stimulus bill, formally known as the American
Recovery and Reinvestment Act of 2009 (ARRA), also provided funding
for enhancements to a number of public assistance programs. In particular, Medicaid and SNAP, both entitlement programs designed to
expand in times of need, were provided with additional funding from the
federal government, above and beyond normal expansions. SNAP benefits were temporarily increased by more than 10 percent and states were
also encouraged to relax the rules to be eligible for SNAP.1 The percent
of funding coming from federal matching funds for Medicaid was also
expanded in the ARRA.2 Both SNAP and Medicaid usage expanded during the recession.3
The ARRA also provided some extra funding for the EITC (but only for
large families), state TANF block grants, SSI payments, and housing, but
these funds were relatively limited.4 Despite the limited additional funding, the EITC program grew in the Great Recession as family incomes
fell into the range where families became eligible for the benefit.5 The
average size of the benefit received also grew by about $145 between 2007
and 2010.6 TANF and SSI generally did not increase during the Great
Recession, though TANF caseloads rose in some states.7 Last, unemployment insurance was expanded dramatically, both in terms of eligibility and
length of the benefit to an unprecedented ninety-nine weeks, far beyond the
normal maximum of twenty-six weeks, and usage increased dramatically.8
Aside from the EITC, most public assistance programs are designed to
assist the poorest groups the most. Unemployment insurance, which is
not means tested, should increase among all groups, though research has
found that UI participation rates were much lower among less-educated
public and private transfers61
populations, despite their being potentially eligible for UI.9 Although
about 90 percent of jobs in the United States are covered by UI, on average just over 33 percent of unemployed workers receive benefits, as many
workers do not earn enough or have not worked for enough time to be
eligible for benefits.10
A recent study by Marianne Bitler, Hilary Hoynes, and Elira Kuka
suggests that government transfers did indeed reduce the extent of child
poverty during the recession.11 The authors found that the public safety
net significantly decreased the percent of children who would have been
poor—in large part due to the increase in the importance of Food Stamps.
A 1 percentage point increase in unemployment was linked with a 1.1 percentage point increase in pretransfer poverty but only a 0.8 percentage point
increase when taxes and transfers are taken into account, which means that
public transfers reduced the increase in poverty by about 27 percent.
Private support networks may also help families make ends meet in
times of economic crisis, but little research has estimated the quantitative
importance of such support. In previous research using Fragile Families
and Child Well-Being Study (FFS) data, we found that the odds of receiving a private financial transfer increased for most families in the Great
Recession but decreased for those with higher incomes.12 A number of
studies document that doubling up increased slightly during the Great
Recession.13 One study finds that individual unemployment was linked
with increased odds of doubling up in particular among those with less
than a high school diploma and those with some college.14 This chapter
provides evidence on both public and private transfers to see the extent to
which they helped different kinds of families make ends meet.
PUBLIC AND PRIVATE TRANSFERS
DURING THE FIRST NINE YEARS
We begin by examining the prevalence of public and private transfer
receipt in the Fragile Families data, with an eye toward disparities between
more and less vulnerable families. Figure 3.1 plots the percentage of families receiving seven forms of public benefits as their child ages from one
to nine. It is important that the ARRA described earlier, which expanded
many public programs in response to the Great Recession, was implemented in early 2009. The data in year nine were collected from 2007 to
early 2010. Thus, although we capture about one year of data post-ARRA,
much of our year nine data predates the ARRA expansions.
The proportion of families receiving benefits is high. Medicaid is the most
commonly received. About half of the families receive Medicaid. The EITC
and SNAP are the next most common—from about 30 to 50 percent. The
proportions receiving Medicaid, EITC, and SNAP increase over time,
the biggest increase being for the EITC. The increase in the EITC is
62
children of the great recession
Figure 3.1 Public Assistance Receipt by Child’s Age-Year
Percent of Households Receiving
Public Assistance
60
50
Medicaid
EITC
SNAP
TANF
UI/other
Housing
SSI
40
30
20
10
0
1
3
5
(1999–2001) (2001–2003) (2003–2006)
9
(2007–2010)
Child’s Age-Year
Source: Authors’ calculations.
Note: EITC = Earned Income Tax Credit, SNAP = Supplemental Nutrition Assistance
Program–Food Stamps, TANF = Temporary Assistance to Needy Families, UI =
Unemployment Insurance, SSI = Supplemental Security Income. Sample is restricted to
mothers who were interviewed in all survey waves (n = 2,986). Figures are weighted.
undoubtedly driven by the increase in the work and earnings of mothers
as their children grow older and enter school. Over time, smaller proportions of families receive TANF, decreasing from about 20 percent when
the child is one year old to around 10 percent by the time the child is nine.
The decline in TANF receipt is also attributable to increases in mothers’
work and earnings but is also likely partly driven by TANF lifetime time
limits and by states’ attempts to divert recipients from receiving TANF
benefits. Public housing is relatively stable at about 20 percent over
the entire period. Finally unemployment insurance and Supplementary
Security Income are quite uncommon in the early years but rise steadily to
about 10 percent total by age nine.
As we would expect, more disadvantaged families are much more likely
to receive government transfers targeted at low-income families than
other families. Figure 3.2 displays the average receipt rate over the nine
years by mothers’ education. The rates decline precipitously as education
increases for most programs. The clear exception is unemployment insurance, the only non-income-tested benefit. Especially striking is the extent
to which families with a college-educated mother differ from all other
groups. For this group receipt of any individual type of public assistance is
below 10 percent, with the exception of the EITC, which is a surprisingly
high 31 percent.
public and private transfers63
Figure 3.2 Public Assistance Receipt by Education
Percent of Households Receiving
Public Assistance
90
80
70
Less than
high school
60
High school
50
40
Some college
30
College +
20
10
0
Medicaid EITC
SNAP Housing TANF
SSI
UI/other
Source: Authors’ calculations.
Note: EITC = Earned Income Tax Credit, SNAP = Supplemental Nutrition Assistance
Program–Food Stamps, TANF = Temporary Assistance to Needy Families, UI =
Unemployment Insurance, SSI = Supplemental Security Income. Sample is restricted to
mothers who were interviewed in all survey waves (n = 2,986). Rates are averaged over
four survey years. Figures are weighted.
How much assistance do families with young children receive from
government transfers? Figure 3.3 presents the average annual value per
recipient of each of the seven public benefits. Medicaid, the most common
benefit, is also worth the most when valued at government cost (one of
many ways to assess Medicaid value)—about $9,500. Housing assistance,
which is much less common because it is not an entitlement and is underfunded, is only slightly less valuable at about $7,700. SSI, the rarest of all
the benefits but valued at $6,500, is the only cash benefit that comes close
to being worth as much as health care and housing assistance. The average
annual benefit for the EITC, SNAP, TANF, and UI is around $3,000 for
each program. Although the receipt of benefits varies tremendously by
education group, the value of these benefits varies little, conditional on
their receipt.
Figures 3.2 and 3.3 suggest that the poorly educated receive far larger
benefits from the American welfare state than the college educated. This
is because we focus in this chapter on government transfers targeted at
low-income families. But college-educated families are much more likely
than other families to receive public benefits through the tax system,
including government-subsidized, employer-provided health insurance
and deductions for home ownership. Indeed, other work shows that once
64
children of the great recession
Figure 3.3 Average Dollar Value of Public Assistance Benefits
10,000
Dollars of Public Assistance
9,000
8,000
7,000
6,000
5,000
4,000
3,000
2,000
1,000
0
Medicaid Housing
SSI
SNAP
TANF
UI
EITC
Source: Authors’ calculations.
Note: EITC = Earned Income Tax Credit, SNAP = Supplemental Nutrition Assistance
Program–Food Stamps, TANF = Temporary Assistance to Needy Families, UI =
Unemployment Insurance, SSI = Supplemental Security Income. Sample is restricted to
mothers who were interviewed in all survey waves (n = 2,986) and who received the
benefit. Values are averaged over four survey years. Figures are weighted.
employer-provided benefits and tax benefits are counted, the total value
of cash and in-kind transfer benefits is more or less equal across all income,
family structure, and education groups.15 Although transfer amounts are
roughly equal across groups, less-educated families rely far more on these
transfers than college-educated families do because their market incomes
are so much lower.
The next two figures look at what might be dubbed the private safety
net, or transfers from family and friends. Figure 3.4 plots the percent of
families receiving private cash transfers over time. In the early years, nearly
40 percent of families receive private cash transfers but this declines to
about 25 percent by age three and then increases to around 30 percent at
ages five and nine. Recall from the previous chapter that families’ incomes
were at their peak at age three, perhaps making private transfers less critical
for making ends meet at that time. Differences by education are relatively
small, and we do not see that college-educated mothers are dramatically
different from other mothers.
Figure 3.5 shows the percentage of families who double up (live with
family or friends) over time. As was true of cash transfers, doubling up is
most common in the earlier years, when nearly 30 percent of families are
public and private transfers65
Figure 3.4 Private Financial Transfers ($2010)
Percent of Households Receiving
Private Financial Transfers
45
40
College +
35
Some college
30
25
High school
20
15
Less than
high school
10
5
0
1
3
5
(1999–2001) (2001–2003) (2003–2006)
9
(2007–2010)
Child’s Age-Year
Source: Authors’ calculations.
Note: Sample is restricted to mothers interviewed in all survey waves (n = 2,986). Figures
are weighted.
Percent of Households Doubling Up
Figure 3.5 Doubling Up
45
40
College +
35
30
Some college
25
High school
20
15
Less than
high school
10
5
0
1
3
5
(1999–2001) (2001–2003) (2003–2006)
9
(2007–2010)
Child’s Age-Year
Source: Authors’ calculations.
Note: Sample is restricted to mothers interviewed in all survey waves (n = 2,986). Figures
are weighted.
66
children of the great recession
Dollars of Private Assistance
Figure 3.6 Average Dollar Value of Private Assistance
10,000
9,000
8,000
7,000
6,000
5,000
4,000
3,000
2,000
1,000
0
All
Less than
high school
High school
Some college
College +
PFTs
Doubling Up - Rental Savings
Source: Authors’ calculations.
Note: PFT = private financial transfer. Sample is restricted to mothers who were interviewed
in all survey waves (n = 2,986). Values are averaged over four survey years and are among
recipients. Figures are weighted.
doubled up. By age five, however, the rate drops below 20 percent and
by age nine to 15 percent. Unlike private financial transfers, differences by
education in doubling up are dramatic. Mothers without college degrees
are three to four times as likely to double up than college-educated mothers.
The decline in doubling up as education increases is particularly interesting when viewed in conjunction with the dollar value of private cash
transfers (among recipients), which is shown in figure 3.6. The value of
private financial transfers differs dramatically by mother’s education, ranging from just under $1,300 and $1,400 respectively for those without a
high school degree or only a high school education, to around $3,000
for those with some postsecondary education, and close to $9,000 for
those with a college degree. Nearly all private financial transfers come
from family, particularly parents providing to their children. Highly educated mothers generally come from wealthier backgrounds than poorly
educated mothers. Wealthier families, in turn, have more cash to give and
greater ability to indulge a preference for giving cash assistance rather than
sharing their homes.
To estimate the value of doubling up, we compared the rent paid by
families who were doubled up with that paid by similar families (in terms of
education, race, age, immigrant status, and city) who were not doubled up
(for more detail, see the appendix).16 The rental savings estimates reported
in figure 3.6 are the average differences in rent between mothers who are
doubled up and those who are not doubled up. Note that the value of
doubling up declines as mother’s education increases. Although college-
public and private transfers67
educated mothers are more likely to pay higher rents or mortgages, and we
might expect them to have the largest subsidy when doubled up, this might
not be the case for a couple of reasons. First, more educated mothers can
afford to pay more rent even when they are doubled up, so the subsidy they
receive may be smaller. Second, they may also be more likely to host others
when doubled up, or bring in others who are in need (such as an ailing
grandparent) who may be less able to help subsidize rent.
In short, public and private transfers to families with young children
are quite common. Private financial transfers, doubling up, and TANF are
most common in the first year the child’s life and decline steady thereafter.
Receipt of the other public transfers—particularly, Medicaid, EITC and
SNAP—increase over time.
In the last chapter, we saw that household income decreases substantially as a consequence of the Great Recession. In the next section, we
examine how public and private transfers respond to recessions. Here we
measure average responsiveness of public and private safety nets over two
recessionary periods to consider: How much do public and private transfers increase? How much worse off would these households be were it not
for public and private transfers?
EFFECTS OF THE GREAT RECESSION ON TRANSFERS
As outlined in chapter 1, we use the association between the city unemployment rate and transfers, controlling for a number of demographic characteristics of the mother to estimate the effect of the Great Recession on
public and private transfers (see table 3.A1 for an example of the model
with and without individual fixed effects). We predict the outcomes when
unemployment rates are 5 percent and 10 percent—akin to the change that
occurred during the Great Recession. We find no evidence that the effect of
unemployment during Great Recession years was different than that during
earlier periods (see appendix table 3.A3), but, as described earlier, because
the generosity of some transfer programs were increased toward the end
of our interviewing period, we underestimate the effect responsiveness of
transfers to the Great Recession.
We begin with the private safety net, depicted in figure 3.7. Private cash
transfers increase for all groups as a consequence of the Great Recession. The
increases are quite large, especially for families headed by mothers without a
high school degree and with a college degree—respectively 15 and 14 percentage points. At 10 percent unemployment, more than a third of all
education groups are estimated to be receiving at least some private financial
transfer. As noted, the value of transfers is much lower for the poorest families than for the most educated, about $1,400 versus $8,700—though in
percentage terms, private financial transfers represent a similar proportion
of income—4 percent and 5 percent of incomes, respectively. Figure 3.7
68
children of the great recession
Predicted Percent of Private
Assistance
Figure 3.7 Private Financial Transfers and Doubling Up
50
45
40
35
30
25
20
15
10
5
0
+29%
+52%
UR 5 percent
+15%
UR 10 percent
+8%
+74%
+2%
+9%
+10% +29%
–20%
All*
Less than
high
school*
High
school
PFTs
Some
college
College +*
All
Less than
high
school
High
school
Some
college
College +
Doubled Up
Source: Authors’ calculations.
Note: UR = unemployment rate, PFT = private financial transfer. Regressions include the
full set of control variables. The association between UR and PFTS is statistically significant for the full sample. Chow tests show that the high school group is statistically
different (p < 0.05) from the other groups for PFTs, no differences for doubled up are
statistically significant.
*p < .05 between UR and outcome
also shows the effect of the Great Recession on doubling up. Here we find
no significant associations. Thus, although at 10 percent unemployment
about 20 percent of families are predicted to be doubled up, this proportion
appears to be roughly the same in good and in bad economic times.
Although the private safety net is an important source of support for
many families with young children, transfers from government sources
tend to be much larger. We look at the Great Recession’s effects on these
transfers next. Figure 3.8 shows the change in the proportion of families
receiving five public transfers—Medicaid, EITC, Food Stamps or SNAP,
TANF, and UI—by education of the mother as a consequence of the Great
Recession. We do not show changes in housing transfers because such
assistance is limited by annual budgets that have never been expanded during recessions (and analyses showed no significant associations). Nor do
we show the results for SSI because too few families in our sample receive
SSI to allow us to obtain reliable estimates.
Figure 3.8 shows considerable variation across the five benefits, as
might be predicted by the laws governing receipt of each benefit. For
instance, Medicaid receipt rates increase substantially (and with statistical
significance) for those with less than a high school education and for those
with a high school diploma, but only by a small and statistically insignificant amount for those with some postsecondary education and not at all
for those with a college degree.
public and private transfers69
Predicted Percentage Point
Change in Public Assistance
Figure 3.8 Public Transfer Receipt Rates
22*
25
17*
20
16*
10*
14*
15
10 10*
9*
7*
10
5
4* 5
5
0
–1
–2
–5
–3
–5–6 –5
–10
–15
–15*
–20
Medicaid
EITC
SNAP
TANF
All
10*
4*
6
Less than
high school
High school
1
Some college
–3
College +
UI/Other
Source: Authors’ calculations.
Note: EITC = Earned Income Tax Credit; SNAP = Supplemental Nutrition Assistance
Program; TANF = Temporary Assistance for Needy Families; UI = Unemployment
Insurance. Regressions include the full set of control variables. Chow tests show that the
high school group is statistically different (p < 0.05) from the other groups for Medicaid;
for UI, the less than high school and college groups are statistically different from the
high school and some college groups.
*p < .05 between UR and outcome
As a consequence of the Great Recession, the receipt rates for the EITC
for families headed by mothers with less than a college degree, especially
those with only a high school diploma, decline and those with a college
degree increase. Recessions both increase EITC participation by reducing
the annual earnings of those not previously eligible to eligibility levels and
decrease participation by reducing annual earnings of those previously
eligible to zero, making them ineligible for the EITC.
SNAP, as expected, increases for all families except those with a collegeeducated mother. Consistent with prior research on fragile families, we
find that TANF increased in the Great Recession, but only for the mothers
with the lowest level of education.17
Last, UI receipt is strongly related to local unemployment rates for
mothers with high school or some postsecondary education. The increase
for these mothers is statistically significant and nearly twice as large as
that of mothers with a high school diploma. Eligibility for UI is based on
duration of employment and earnings prior to job loss. In addition, in
many states, part-time employees—those who work fewer than thirty-five
hours a week—are not eligible for UI. Mothers with less education are
more likely to work part-time or in low-paying jobs than mothers with
more education. Thus, that mothers with some postsecondary education
experienced the largest increase in UI might be expected because these
mothers are more likely to have worked in jobs that were eligible for UI.
70
children of the great recession
Somewhat unexpected is the negative link with UI for college-educated
mothers, though the association is not statistically different from zero.
(We also examine differences by race-ethnicity in table 3.A4.)
THE HELPING EFFECTS OF PUBLIC
AND PRIVATE TRANSFERS
How much worse off would families have been in the Great Recession
in the absence of public and private transfers? To answer this question,
we begin with our estimate of the post-transfer income at 10 percent
unemployment from chapter 2. Recall post-transfer household income
includes both cash and near-cash public benefits and private cash transfers
from family and friends. Recall also that the estimate was derived from
the relationship between household income and the local unemployment
rate when the child was ages one, three, five, and nine. To estimate how
much transfers increased household income, we introduce a new measure,
pretransfer household income, which equals post-transfer income minus
public and private transfers. Following the same procedure, we then estimate the effect of the Great Recession on pretransfer household income.
The difference between the two at 10 percent unemployment shows how
much transfers increased income in the Great Recession. Our estimates
will be too low—that is, will underestimate the increase in transfers—for
two reasons. First, as noted, SNAP benefits were temporarily increased
beginning in 2009. This increase is not reflected in data collected before
2009 and therefore is reflected in only about one year of our data because
the collection finished in the spring of 2010, and unemployment remained
very high after that. Second, our estimates include the effects on both the
dot-com and the Great Recession. Public benefits were lower in the
earlier recession: there was no expansion like the ARRA. On the other
hand, to the extent that transfers induce declines in work and earnings
among recipients, the difference between pre- and post-transfer income
overestimates the increase in income attributable to transfer programs.
Research suggests that these effects are small. We therefore ignore them.18
Figure 3.9 presents our estimates of how much transfers increase
income for our four groups of families at the peak of Great Recession. We
find that transfers are largest for the least-educated groups and decline as
education increases. At the bottom of the education distribution, household income was 18 percent higher as a result of transfers. For high school
graduates, transfers increase income by about 9 percent. By way of contrast, public and private transfers increase income much less for the two
better-educated groups—about 3 percent.
In figure 3.10, we perform a similar analysis but we compare poverty
rates (percentage of families below the poverty line) using our measures of
pre- and post-transfer income. This analysis shows the mitigating effects
public and private transfers71
Predicted Dollars of Household Income
Figure 3.9 Effects of Transfers on Household Income
160,000
+3%
140,000
120,000
Pretransfer
income
100,000
80,000
+3%
+17%
60,000
+9%
+18%
40,000
Post-transfer
income
20,000
0
All
Less than
high school
High
school
Some
college
College +
Source: Authors’ calculations.
Note: Regressions include the full set of control variables. The associations between
unemployment rate and income are significant for all groups except the college
educated. Chow tests find no significant differences across groups.
Predicted Percent of Households in Poverty
Figure 3.10 Mitigating Effects of Transfers on Poverty
60
–26%
50
40
Pretransfer
poverty
–26%
–21%
30
20
10
0
Post-transfer
poverty
–26%
–33%
All
Less than
high school
High
school
Some
college
College +
Source: Authors’ calculations.
Note: Regressions include the full set of control variables. The associations between
unemployment rate and poverty are significant for the full sample, less than high school
and high school. Chow tests find no significant differences across groups.
72
children of the great recession
of transfers on poverty. The largest absolute differences again occur at the
bottom and decline steadily as education increases—from 13 percentage
points for those without a high school diploma, to 9, 5, and 1 percentage
points respectively for the more educated. The mitigation effect measured
in percentage terms, however, is greatest for the most well-off group.
Among those with a college degree, transfers lowered the poverty rate by
33 percent. Although this decline is huge in percentage terms, the absolute
change is not: transfers moved the college-educated group from 3 percent
to 2 percent in poverty. In comparison, the mitigation effects for lesseducated groups are much smaller in percentage terms but the absolute
change in poverty—the share of families in each group saved from poverty
by transfers—is much larger. Before transfers, about 50 percent of families
with mothers lacking a high school diploma are in poverty, whereas 37 percent are poor after transfers.
Finally, we consider the mitigating effects of transfers on food insecurity. As reported in chapter 2, economic insecurity increased for all groups
in the wake of recession. We expect that economic insecurity would have
been more common and more severe were it not for government and
private transfers. How much higher would it have risen absent safety net
benefits? Using FFS data alone, it is difficult to estimate the effects of transfers on economic insecurity because we do not observe insecurity in the
absence of transfers. But because research on the effects of Food Stamps
on food hardship is quite good, we can use our data to analyze the effects
of transfers on food insecurity. By exploiting real-world variation on Food
Stamp receipt generated by state errors in under- or overpayments, Elton
Mykerezi and Bradford Mills are able to estimate that Food Stamp receipt
decreases food insecurity by 22 percent.19 Multiplying this estimate by
our estimated increase in Food Stamp usage of 10.5 percentage points as a
result of a big recession yields an increase of food insecurity of 2.3 percentage points. This calculation suggests that absent Food Stamps, recessioninduced increases in food insecurity rates would have been approximately
twice as high as they actually were.20
In contrast to the modest to substantial effects on income, poverty,
and insecurity for public transfers, we find that private financial transfers increase household income minimally. In identical analyses to those
shown, we subtracted private transfers and found that these transfers had
no substantial effects on income or poverty.
Taken together, the analyses of public transfers suggest that some parts
of the public safety net operate well—rising in hard times to help families make ends meet and providing the greatest relief to the most vulnerable. Absent public transfers, disadvantaged families would have it much
worse—less income, higher poverty rates, and increased economic insecurity. Still, poverty and insecurity rates for families with children remain
unconscionably high in both good and bad times.
public and private transfers73
APPENDIX
Measures
The values of all variables are in 2010 dollars.
Medicaid receipt indicates whether the mother was receiving Medicaid
at the time of the interview. In year nine, mother’s Medicaid was not
assessed separate from other forms of health insurance. Usage is therefore
based on whether her child received Medicaid. Tests restricting to child’s
Medicaid usage yielded similar results. The value of Medicaid is based
on state Medicaid per person expenditures for the appropriate year.21
Mothers who reported receiving Medicaid received the full value; it was
assumed that all children in the household received Medicaid.
Earned Income Tax Credit is estimated using TAXSIM version 9.2, a
program run by the National Bureau of Economic Research that uses
marital status, earnings, and dependents to estimate tax liabilities under
U.S. federal and state tax laws.
SNAP measures indicate whether mothers received the benefit the previous year. The annual dollar amount of SNAP received is constructed from
the number of months received and the amounts received each month.
Temporary Aid for Needy Families is a binary variable indicating
whether the mother received the benefit in the previous year. The amount
is constructed from the number of months received and monthly amounts
received.
Unemployment insurance or other assistance is based on whether mothers
received unemployment insurance or other assistance such as workers compensation in the previous year. We construct a binary variable for receipt
and an amount variable using data on the amount and number of months
received.
Public housing is a binary variable constructed to indicate that mother
lives in public housing if she reports either living in a housing project
or receiving federal, state, or local assistance to pay for housing—such
as Section 8. The dollar value is estimated using data from the U.S.
Department of Housing and Urban Development’s fair market housing
calculator in each metro area and year. For mothers receiving public housing or public housing assistance, the rent paid is subtracted from the fair
market value for rent to estimate the public housing subsidy.
Supplemental Security Income is a measure of whether mothers received
SSI in the previous year, the months of receipt, and monthly dollar
amounts. This information is used to construct a binary receipt variable
and an annual amount of SSI.
Private financial transfers measures the annual dollars received from
families and friends. A binary variable indicates that private transfers were
received.
74
children of the great recession
Doubling up, or moving in with others, is a measure of whether the
mother is living in a household with an adult who is not the mother, the
mother’s partner, or an adult child. To estimate its economic value, we
estimate the yearly dollar value of the rent a mother saves by doubling
up when she is living in someone else’s household. Rent is ascertained by
asking mothers how much they pay in rent, but it is not clear whether this
figure is the full household rent or what they actually pay. Because of this
ambiguity, we restrict our rental savings estimates to mothers who move
in with others because they are much more likely to report what they pay
toward rent than the full household rent. Using data on the rent paid by
mothers who are not doubled up, we generate a predicted rent variable
for the full sample of mothers for waves 3 through 5, when we have data
on whether she lives in her own or someone else’s home. Our prediction equation includes basic demographic information, such as age, race,
lagged measures of income, and city of residence. We then compare the
actual rent that doubled-up mothers pay against their predicted rent to
generate an estimate of the rental savings from doubling up.22
Supplemental Analyses
Additional analyses test the association between the unemployment rate
and public and private transfers. First, to determine whether the association between the unemployment rate and outcomes differed during the
recession, we include an interaction term with the unemployment rate and
the year nine survey wave. The only significant interaction is for doubling
up, suggesting that the odds of doubling up may have increased in the
recession years; however, the main model specification is not significant
(see table 3.A3).
Second, to determine whether the rate of change in the unemployment
rate was more closely related with transfer receipt, spline models distinguish between an annual declining rate of change in the unemployment
rate and an annual increasing rate of change in the unemployment rate and
economic outcomes. Few associations are significant and the main coefficient on unemployment is unchanged (see table 3.A2).
Third, we use the consumer confidence index and the foreclosure rate
as indicators of the Great Recession. No associations between the index
and outcomes are significant. The foreclosure rate is significantly associated with some transfers and quite similar to the unemployment rate
results, occasionally stronger and occasionally weaker.
Fourth, focusing on an income drop between years five and nine,
we regress year nine outcomes on a measure of a 1 to 40 percent drop
in income and a 40 percent plus drop in income. These findings, as
anticipated, show that a large income drop is linked with higher odds of
receiving both public and private transfers (in general, not all associa-
public and private transfers75
tions were significant), and that the larger the drop the higher the odds
of hardship.
Fifth, we consider a change in the unemployment rate between years five
and nine on year nine outcomes, distinguishing increases and decreases in
unemployment. These analyses generally show that a decline in unemployment is linked with fewer transfers and that an increase is linked with
more transfers.
Last, we lag the unemployment rate using two models. In the first, the
average unemployment rate over the prior year is lagged two and three
years. In the second, the unemployment rate at the interview is included
and lagged at twelve months, twenty-four months, and thirty-six months.
The models show no evidence of a lag in the association between the
unemployment rate and the economic outcomes.
Table 3.A1 Full Regression Results for SNAP
With Individual
Fixed Effects
Unemployment rate
Education
Less than high school
High school
Some college
Relationship status
Married
Cohabiting
Mother’s age
Race-ethnicity
Black
Hispanic
Other
Immigrant
Number of children in household
Lived with both parents at age fifteen
Interview year
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
1.18**
0.54**
0.41**
0.38**
0.43**
0.49**
0.48**
0.38
1.05
0.66*
0.50**
0.55
(4.64)
(-2.85)
(-5.83)
(-4.86)
(-4.48)
(-3.61)
(-3.96)
(-1.54)
(0.21)
(-2.18)
(-3.04)
(-1.49)
Without Individual
Fixed Effects
1.10**
(2.85)
14.08**
8.03**
4.85**
(19.93)
(14.49)
(10.16)
0.28**
0.76**
0.99†
(-13.92)
(-5.71)
(-1.66)
1.93**
1.38*
1.72*
0.50**
1.16**
0.73**
(6.32)
(2.21)
(2.21)
(-4.79)
(6.04)
(-8.13)
0.65†
0.57**
0.56*
0.60*
0.66†
0.66†
0.39*
0.93
0.81
0.71
0.88
(-1.73)
(-2.63)
(-2.40)
(-2.06)
(-1.68)
(-1.84)
(-2.34)
(-0.45)
(-0.80)
(-1.32)
(-0.42)
(Table continues on p. 76.)
76
children of the great recession
Table 3.A1 Continued
City
Austin
Baltimore
Detroit
Newark
Philadelphia
Richmond
Corpus Christi
Indianapolis
Milwaukee
New York
San Jose
Boston
Nashville
Chicago
Jacksonville
Toledo
San Antonio
Pittsburgh
Norfolk
Constant
Observations
Number of individuals
With Individual
Fixed Effects
Without Individual
Fixed Effects
6,430
1,771
1.14**
(3.52)
0.75**
(-3.44)
0.94
(-0.80)
0.75**
(-3.81)
0.94
(-0.77)
(1.88)
1.22†
1.82**
(5.98)
1.29**
(3.12)
1.88**
(9.07)
1.22*
(2.13)
0.45**
(-6.57)
1.05
(0.65)
1.35**
(3.77)
1.23*
(2.28)
(-1.82)
0.86†
1.57**
(4.72)
0.94
(-0.62)
1.19*
(2.07)
1.22*
(2.22)
0.08**
(-9.39)
15,688
4,594
Source: Authors’ calculations.
Note: Figures reported are odds ratios. Z-stats in parentheses. Covariates are measured at the baseline
survey (except year) and are clustered at the city and individual level. Model includes level unemployment rate. The model without individual fixed effects is clustered at city and individual level.
**p < .01; *p < .05; †p < .1
Medicaid odds ratios (z-stat)
Unemployment rate
1.19**
(model 1)
(5.27)
Unemployment rate
1.19**
(model 2)
(5.21)
Increasing
1.00
unemployment rate
(-0.45)
Decreasing
1.00
unemployment rate
(0.64)
Observations
7,529
Number of
2,060
individuals
EITC odds ratios (z-stat)
Unemployment rate
0.95
(model 1)
(-1.63)
Unemployment rate
0.96
(model 2)
(-1.38)
Increasing
1.00
unemployment rate
(0.52)
Decreasing
1.02*
unemployment rate
(2.37)
Observations
9,053
Number of
2,475
individuals
All
0.94
(-1.30)
0.95
(-1.07)
1.00
(1.26)
1.00
(0.42)
3,535
979
1.27**
(4.64)
1.28**
(4.77)
1.00
(1.39)
1.03*
(2.37)
3,092
855
Less than
High
School
0.84**
(-2.93)
0.85**
(-2.63)
1.00
(1.20)
1.03*
(2.10)
2,222
604
1.23**
(3.29)
1.22**
(3.19)
1.00
(-0.45)
0.99
(-0.42)
2,108
574
High
School
1.02
(0.30)
1.02
(0.39)
1.00
(0.02)
1.02
(1.25)
2,365
642
1.07
(1.08)
1.06
(0.84)
1.00
(-1.06)
0.99
(-1.01)
1,943
526
Some
College
With Individual Fixed Effects
1.14
(1.26)
1.12
(1.00)
0.99*
(-2.25)
1.04
(1.63)
931
250
0.91
(-0.63)
0.84
(-1.11)
0.98†
(-1.88)
0.95
(-1.53)
386
105
College +
All
0.97
(-0.80)
0.97
(-0.75)
1.00
(-0.54)
1.01**
(2.76)
15,884
4,605
1.11*
(2.55)
1.11*
(2.45)
1.00
(-0.38)
1.00
(0.11)
15,820
4,604
Table 3.A2 Coefficients and Standard Errors, Rate of Change for Transfers
0.98
(-0.34)
0.98
(-0.32)
1.00
(-0.02)
1.01
(1.40)
6,139
1,823
1.17*
(2.52)
1.18**
(2.72)
1.00
(1.36)
1.02
(1.42)
6,121
1,822
Less than
High
School
0.99
(-0.14)
0.99
(-0.16)
1.00
(-1.13)
1.01
(0.98)
3,931
1,124
0.89†
(-1.82)
0.90†
(-1.66)
1.00
(1.06)
1.02*
(2.26)
4,073
1,162
1.15†
(1.81)
1.12
(1.31)
0.99**
(-2.60)
1.02
(1.04)
1,741
496
0.97
(-0.30)
0.94
(-0.53)
0.99†
(-1.74)
0.97
(-1.11)
1,713
496
College +
(Table continues on p. 78.)
1.08*
(2.23)
1.06†
(1.66)
1.00†
(-1.67)
0.99
(-0.99)
3,919
1,124
Some
College
1.07
(1.17)
1.07
(1.14)
1.00
(-0.09)
0.99
(-1.03)
4,048
1,162
High
School
Without Individual Fixed Effects
SNAP odds ratios (z-stat)
Unemployment rate
(model 1)
Unemployment rate
(model 2)
Increasing
unemployment rate
Decreasing
unemployment rate
Observations
Number of
individuals
TANF odds ratios
Unemployment rate
(model 1)
Unemployment rate
(model 2)
Increasing
unemployment rate
Decreasing
unemployment rate
Observations
Number of
individuals
1.16**
(2.67)
1.17**
(2.86)
1.00
(1.27)
1.01
(1.03)
2,920
819
1.23**
(3.47)
1.25**
(3.61)
1.00
(0.51)
1.03†
(1.95)
2,775
777
1.16**
(3.34)
1.15**
(3.22)
1.00†
(-1.65)
1.02*
(2.23)
5,152
1,418
All
Less than
High
School
1.15
(1.63)
1.12
(1.35)
0.99†
(-1.83)
1.02
(0.99)
1,373
370
1.25**
(3.24)
1.23**
(3.00)
0.99†
(-1.85)
1.02
(1.53)
1,906
519
High
School
1.00
(-0.01)
0.98
(-0.17)
0.99†
(-1.91)
1.01
(0.62)
960
259
1.18*
(2.17)
1.18*
(2.18)
1.00
(-0.69)
1.02
(1.34)
1,459
394
Some
College
With Individual Fixed Effects
1.18**
(4.64)
1.19**
(4.62)
1.00
(-0.61)
1.02*
(2.06)
6,430
1,771
Table 3.A2 Continued
—
—
—
—
0.89
(-0.44)
0.79
(-0.78)
0.95*
(-2.09)
0.99
(-0.10)
145
39
College +
1.07
(1.36)
1.06
(1.30)
1.00
(-1.32)
1.00
(0.56)
15,750
4,603
1.10**
(2.85)
1.10**
(2.73)
1.00
(-0.71)
1.00
(0.52)
15,688
4,595
All
1.09
(1.52)
1.09
(1.58)
1.00
(-0.61)
1.01
(0.75)
4,026
1,159
1.00
(0.01)
1.00
(-0.04)
1.00
(-0.55)
1.01
(0.49)
4,006
1,162
1.11†
(1.86)
1.11†
(1.86)
1.00
(-0.04)
1.00
(0.01)
6,062
1,821
High
School
1.09*
(2.07)
1.09*
(2.03)
1.00
(1.02)
1.00
(-0.06)
6,022
1,817
Less than
High
School
1.07
(1.09)
1.05
(0.77)
0.99**
(-2.58)
1.01
(0.75)
3,912
1,124
1.16*
(2.20)
1.15*
(2.04)
1.00
(-1.34)
1.01
(0.39)
3,900
1,123
Some
College
Without Individual Fixed Effects
0.79
(-0.32)
0.79
(-0.38)
0.98
(-1.06)
1.20
(1.60)
740
496
1.05
(0.22)
0.99
(-0.03)
0.97*
(-2.28)
1.02
(0.44)
1,700
496
College +
Public housing or Section 8 odds ratio (z-stat)
Unemployment rate
1.00
1.01
(model 1)
(-0.04)
(0.16)
Unemployment rate
0.99
1.01
(model 2)
(-0.16)
(0.25)
Increasing
1.00
1.00
unemployment rate
(-1.09)
(0.41)
Decreasing
1.01
1.01
unemployment rate
(0.67)
(0.61)
Observations
4,427
2,296
Number of
1,218
645
individuals
Supplemental Security Income odds ratio (z-stat)
Unemployment rate
1.03
0.96
(model 1)
(0.28)
(-0.32)
Unemployment rate
1.03
0.98
(model 2)
(0.33)
(-0.17)
Increasing
1.00
1.01
unemployment rate
(0.42)
(1.15)
Decreasing
1.01
1.01
unemployment rate
(0.45)
(0.40)
Observations
1,403
701
Number of
384
193
individuals
—
—
—
—
—
—
—
—
—
—
—
1.24
(1.11)
1.24
(1.12)
1.01
(0.64)
1.03
(0.77)
389
107
—
0.98
(-0.16)
0.95
(-0.53)
0.99*
(-2.10)
0.95*
(-2.19)
751
204
0.98
(-0.28)
0.97
(-0.42)
0.99
(-1.22)
1.04†
(1.90)
1,311
349
0.97
(-0.43)
0.98
(-0.41)
1.00
(0.18)
1.00
(0.21)
6,059
1,823
0.95
(-1.04)
0.97
(-0.66)
1.01*
(2.35)
1.01
(0.90)
6,111
1,821
0.97
(-0.77)
0.96
(-0.86)
1.00
(-1.41)
1.01
(1.29)
15,739
4,604
0.94†
(-1.70)
0.95
(-1.43)
1.00†
(1.79)
1.01
(1.19)
15,828
4,602
0.90
(-0.91)
0.91
(-0.89)
1.00
(0.35)
1.01
(0.61)
4,042
1,162
0.95
(-0.76)
0.95
(-0.92)
1.00
(-0.74)
1.03**
(3.44)
4,049
1,162
0.44
(-1.26)
0.38
(-1.52)
1.04*
(2.16)
0.94
(-1.38)
1,037
495
1.32
(0.89)
1.33
(1.19)
0.97
(-0.94)
1.12
(1.60)
1,190
496
(Table continues on p. 80.)
1.07
(0.69)
1.07
(0.64)
1.00
(-0.29)
1.02
(1.44)
3,811
1,124
0.94
(-1.23)
0.91†
(-1.78)
0.99**
(-4.08)
0.98
(-1.35)
3,896
1,123
All
Less than
High
School
High
School
1.03
(0.48)
1.01
(0.23)
1.00
(-1.56)
1.36**
(2.99)
1.42**
(3.30)
1.01†
(1.90)
1.02
(0.99)
831
221
Some
College
With Individual Fixed Effects
Unemployment insurance or other odds ratio (z-stat)
Unemployment rate
1.13*
0.98
1.20
(model 1)
(2.26)
(-0.22)
(1.62)
Unemployment rate
1.15*
1.02
1.19
(model 2)
(2.45)
(0.17)
(1.52)
Increasing
1.00
1.01
0.99
unemployment rate
(1.56)
(1.62)
(-1.47)
Decreasing
1.00
0.99
0.98
unemployment rate
(-0.19)
(-0.33)
(-0.99)
Observations
2,767
973
749
Number of
743
265
199
individuals
Private financial transfers odds ratio (z-stat)
Unemployment rate
1.11**
1.18**
1.04
(model 1)
(3.37)
(3.46)
(0.67)
Unemployment rate
1.11**
1.20**
1.04
(model 2)
(3.20)
(3.68)
(0.58)
Increasing
1.00
1.00
1.00
unemployment rate
(-0.93)
(1.21)
(-0.47)
Table 3.A2 Continued
1.35*
(2.41)
1.32*
(2.15)
0.99
(-1.36)
0.75
(-1.26)
0.68
(-1.53)
0.98
(-1.18)
0.99
(-0.18)
214
58
College +
1.08**
(4.05)
1.08**
(3.75)
1.00†
(-1.87)
1.13†
(1.93)
1.14*
(2.26)
1.00*
(1.98)
1.00
(-0.69)
15,802
4,603
All
1.12*
(2.41)
1.13*
(2.38)
1.00
(0.42)
0.96
(-0.68)
1.00
(-0.01)
1.01**
(3.41)
1.01
(0.41)
6,008
1,823
Less than
High
School
1.01
(0.16)
1.00
(-0.01)
1.00
(-1.46)
1.23
(1.59)
1.21
(1.35)
0.99
(-1.50)
0.97
(-1.39)
3,990
1,161
High
School
1.07
(1.04)
1.06
(0.93)
1.00
(-1.28)
1.35**
(2.96)
1.38**
(3.29)
1.01*
(2.14)
1.00
(0.36)
3,905
1,123
Some
College
Without Individual Fixed Effects
1.20*
(2.09)
1.20*
(2.00)
1.00
(-0.07)
0.74†
(-1.96)
0.71*
(-2.12)
0.98
(-1.27)
0.97
(-0.62)
1,493
496
College +
0.99
(-1.02)
2,157
587
1.04
(0.54)
1.05
(0.70)
1.00
(1.31)
0.99
(-0.75)
1,796
484
1.02†
(1.89)
3,302
909
1.03
(0.51)
1.01
(0.23)
1.00
(-1.44)
1.00
(-0.37)
3,205
894
1.09
(1.18)
1.09
(1.16)
1.00
(0.21)
0.99
(-0.40)
1,510
409
0.99
(-0.78)
2,098
570
0.96
(-0.30)
0.95
(-0.33)
1.00
(-0.51)
1.01
(0.38)
380
103
1.00
(-0.01)
663
177
1.02
(0.64)
1.02
(0.59)
1.00
(-0.33)
1.00
(-0.78)
15,884
4,605
1.00
(0.08)
15,691
4,600
1.00
(-0.03)
0.99
(-0.20)
1.00
(-1.40)
1.00
(-0.69)
6,139
1,823
1.01†
(1.69)
6,064
1,821
1.00
(-0.06)
1.00
(-0.07)
1.00
(0.14)
0.99
(-0.92)
4,073
1,162
0.98†
(-1.90)
4,027
1,162
1.09*
(2.20)
1.10*
(2.34)
1.00†
(1.86)
1.00
(0.53)
3,931
1,124
1.00
(-0.47)
3,880
1,122
1.06
(0.60)
1.07
(0.62)
1.00
(0.35)
1.01
(0.19)
1,722
495
1.01
(0.60)
1,720
495
Source: Authors’ calculations.
Note: Standard errors and z-stats in parentheses. Model 1 includes unemployment rate as a level. Model 2 includes unemployment rate as a level as well as rate of change in
unemployment rate. SEs for OLS with fixed effects are clustered at city, for OLS and logistic models without fixed effects are clustered at city and individual.
**p < .01; *p < .05; †p < .1
Decreasing
1.00
unemployment rate
(0.19)
Observations
8,220
Number of
2,243
individuals
Doubling up odds ratio (z-stat)
Unemployment rate
1.05
(model 1)
(1.31)
Unemployment rate
1.04
(model 2)
(1.23)
Increasing
1.00
unemployment rate
(-0.24)
Decreasing
0.99
unemployment rate
(-0.67)
Observations
6,891
Number of
1,890
individuals
82
children of the great recession
Table 3.A3 Sensitivity of Coefficients, Transfers
With Individual Fixed
Effects
Medicaid odds ratios (z-stat)
Unemployment rate (model 1)
1.19**
Unemployment rate (model 3)
1.08†
Mother’s unemployment
1.85**
Bio-social fathers not employed
1.15
Unemployment rate (model 4)
1.14**
Unemployment rate * year nine
1.07
EITC odds ratios (z-stat)
Unemployment rate (model 1)
0.95
Unemployment rate (model 3)
1.04
Mother’s unemployment
0.26**
Bio-social fathers not employed
1.01
Unemployment rate (model 4)
0.97
Unemployment rate * year nine
0.96
SNAP odds ratio (z-stat)
Unemployment rate (model 1)
1.18**
Unemployment rate (model 3)
1.13**
Mother’s unemployment
2.25**
Bio-social father’s not employed
1.11
Unemployment rate (model 4)
1.17**
Unemployment rate * year nine
1.01
TANF odds ratio (z-stat)
Unemployment rate (model 1)
1.16**
Unemployment rate (model 3)
1.08
Mother’s unemployment
2.96**
Bio-social fathers not employed
1.41**
Unemployment rate (model 4)
1.11†
Unemployment rate * year nine
1.07
Public housing or Section 8 odds ratio (z-stat)
Unemployment rate (model 1)
1.00
Unemployment rate (model 3)
1.00
Mother’s unemployment
1.47**
Bio-social fathers not employed
1.22†
Unemployment rate (model 4)
0.99
Unemployment rate * year nine
1.01
Supplemental Security Income odds ratio (z-stat)
Unemployment rate (model 1)
1.03
Unemployment rate (model 3)
0.99
Mother’s unemployment
1.29
Bio-social fathers not employed
2.94**
Unemployment rate (model 4)
0.97
Unemployment rate * year nine
1.08
Without Individual
Fixed Effects
(5.27)
(1.93)
(6.77)
(1.60)
(3.08)
(1.27)
1.11*
1.05
2.57**
1.71**
1.10†
1.03
(2.55)
(1.05)
(19.74)
(11.22)
(1.68)
(0.63)
(-1.63)
(0.85)
(-14.80)
(0.15)
(-0.70)
(-0.79)
0.97
1.02
0.34**
1.05
0.97
1.00
(-0.80)
(0.47)
(-12.96)
(1.25)
(-0.63)
(-0.03)
(4.64)
(2.59)
(8.82)
(1.19)
(3.30)
(0.26)
1.10**
1.06†
3.52**
1.85**
1.09†
1.02
(2.85)
(1.75)
(20.66)
(9.20)
(1.88)
(0.46)
(3.34)
(1.25)
(11.09)
(3.40)
(1.83)
(1.00)
1.07
1.03
4.14**
1.97**
1.01
1.10
(1.36)
(0.51)
(26.86)
(8.62)
(0.20)
(1.51)
(-0.04)
(0.01)
(3.76)
(1.95)
(-0.16)
(0.20)
0.97
0.95
1.89**
1.49**
0.95
1.03
(-0.77)
(-0.81)
(11.39)
(6.70)
(-0.79)
(0.63)
(0.28)
(-0.07)
(0.90)
(3.34)
(-0.26)
(0.65)
0.94†
0.86**
2.61**
1.74**
0.93
1.02
(-1.70)
(-3.34)
(9.92)
(4.87)
(-1.04)
(0.21)
public and private transfers83
Table 3.A3 Continued
With Individual Fixed
Effects
Unemployment insurance or other odds ratio (z-stat)
Unemployment rate (model 1)
1.13*
(2.26)
Unemployment rate (model 3)
1.18*
(2.35)
Mother’s unemployment
3.49**
(9.60)
(1.66)
Bio-social fathers not employed
1.25†
(1.86)
Unemployment rate (model 4)
1.15†
Unemployment rate * year nine
0.98
(-0.28)
Private financial transfers odds ratio (z-stat)
Unemployment rate (model 1)
1.11**
(3.37)
Unemployment rate (model 3)
1.14**
(3.48)
Mother’s unemployment
1.29**
(3.14)
Bio-social fathers not employed
1.11
(1.35)
Unemployment rate (model 4)
1.07
(1.59)
Unemployment rate * year nine
1.07
(1.34)
Doubling up odds ratio (z-stat)
Unemployment rate (model 1)
1.05
(1.31)
Unemployment rate (model 3)
1.03
(0.72)
Mother’s unemployment
1.08
(0.83)
Bio-social fathers not employed
1.13
(1.34)
Unemployment rate (model 4)
0.98
(-0.35)
Unemployment rate * year nine
1.12*
(2.07)
Without Individual
Fixed Effects
1.13†
1.16**
2.68**
1.27**
1.15†
0.97
(1.93)
(2.78)
(6.58)
(3.39)
(1.75)
(-0.60)
1.08**
1.11**
1.38**
1.34**
1.05
1.06†
(4.05)
(3.90)
(7.89)
(4.60)
(1.54)
(1.66)
1.02
1.00
1.28**
1.20**
0.99
1.07
(0.64)
(0.10)
(4.42)
(3.91)
(-0.29)
(1.32)
Source: Authors’ calculations.
Note: Standard errors and z-stats in parentheses. Model 3 includes unemployment rate and a measure
of individual unemployment. Model 4 includes unemployment rate and an interaction between unemployment rate and year nine—when the Great Recession hit. SEs for the OLS with fixed effects are
clustered at city, for OLS and logistic models without fixed effects are clustered at city and individual.
**p < .01; *p < .05; †p < .1
0.91
(-0.54)
1.17
(1.44)
1.01
(0.22)
0.95
(-0.87)
1.06
(0.70)
1.19**
(3.52)
1.07
(1.19)
1.14†
(1.76)
1.31**
(3.24)
1.25**
(3.31)
0.95
(-1.03)
1.35**
(5.43)
1.12
(0.90)
0.95
(-0.90)
1.12†
(1.83)
1.22**
(3.56)
0.85**
(-3.16)
1.02
(0.38)
Hispanic
1.04
(0.52)
1.08
(1.13)
1.31*
(2.12)
1.23
(0.62)
0.85
(-1.06)
1.01
(0.07)
1.07
(0.80)
1.08
(1.22)
1.11
(1.37)
White
1.08
(0.85)
1.26**
(3.15)
1.07
(0.52)
1.49†
(1.75)
1.03
(0.18)
1.11
(0.62)
1.02
(0.15)
1.04
(0.59)
1.34**
(3.56)
Married
0.99
(-0.14)
1.04
(0.76)
1.09
(1.00)
1.23
(1.30)
1.05
(0.73)
1.24**
(3.07)
1.32**
(4.79)
0.95
(-1.03)
1.22**
(3.86)
Cohabiting
1.09
(1.64)
1.13**
(2.59)
1.16†
(1.78)
0.77†
(-1.94)
0.96
(-0.62)
1.12†
(1.80)
1.12*
(2.13)
0.92†
(-1.68)
1.13*
(2.32)
Single
Source: Authors’ calculations.
Note: Standard errors and z-stats in parentheses. Model includes level unemployment rate; results include individual fixed effects and time. SEs for OLS with
fixed effects are clustered at city.
**p < .01; *p < .05; †p < .1
Doubling up odds ratio (z-stat)
Unemployment rate
Private financial transfers odds ratio (z-stat)
Unemployment rate
Unemployment insurance or other odds ratio (z-stat)
Unemployment rate
Supplemental Security Income odds ratio (z-stat)
Unemployment rate
Public housing or Section 8 odds ratio (z-stat)
Unemployment rate
TANF odds ratio (z-stat)
Unemployment rate
SNAP odds ratio (z-stat)
Unemployment rate
EITC odds ratio (z-stat)
Unemployment rate
Medicaid odds ratio (z-stat)
Unemployment rate
Black
Table 3.A4 Coefficients and Standard Errors, Transfers
public and private transfers85
NOTES
1.Moffitt 2013.
2. Kaiser Commission on Medicaid and the Uninsured 2009.
3. Nord et al. 2010; Kaiser Commission on Medicaid and the Uninsured 2011.
4. Sell et al. 2010; Moffitt 2013.
5. Moffitt 2013.
6. Mattingly and Kneebone 2012.
7. Moffitt 2013; Pavetti and Rosenbaum 2010.
8. U.S. Senate 2009.
9. Gould-Werth and Shaefer 2012.
10. Simms and Kuehn 2008; Nichols and Simms 2012.
11. Bitler, Hoynes, and Kuka 2014.
12. Gottlieb, Pilkauskas, and Garfinkel 2014.
13. Mykyta and Macartney 2012; Taylor et al. 2011; Cherlin et al. 2013.
14. Weimers 2014.
15. Garfinkel, Rainwater, and Smeeding 2010; Garfinkel and Zilanawala 2015.
16. Pilkauskas, Garfinkel, and McLanahan 2014.
17. Pilkauskas, Currie, and Garfinkel 2012.
18. Moffitt 2013; Fox et al. 2015.
19. Mykerezi and Mills 2010.
20. For a more detailed description of the methodology and findings, see
Pilkauskas, Currie, and Gafinkel 2012.
21. Garfinkel and Zilanawala 2015.
22. For more detail on the method, see Pilkauskas, Garfinkel, and McLanahan
2014.
REFERENCES
Bitler, Marianne, Hilary Hoynes, and Elira Kuka. 2014. “Child Poverty and the
Great Recession in the United States.” Innocenti occasional paper no. 724.
Florence: UNICEF Innocenti Research Centre.
Cherlin, Andrew, Erin Cumberworth, S. Philip Morgan, and Christopher
Wimer. 2013. “The Effects of the Great Recession on Family Structure and
Fertility.” ANNALS of the American Academy of Political and Social Science
650(1): 214–31.
Fox, Liana, Christopher Wimer, Irwin Garfinkel, Neeraj Kaushal, and Jane
Waldfogel. 2015. “Waging War on Poverty: Poverty Trends Using a Historical
Supplemental Poverty Measure.” Journal of Policy Analysis and Management
34(3): 567–92.
Garfinkel, Irwin, Lee Rainwater, and Timothy Smeeding. 2010. Wealth and Welfare
States: Is America a Laggard or Leader? Oxford: Oxford University Press.
86
children of the great recession
Garfinkel, Irwin, and Afshin Zilanawala. 2015. “Fragile Families in the American
Welfare State.” Children and Youth Services Review 55(August): 210–21.
Gottlieb, Aaron, Natasha Pilkauskas, and Irwin Garfinkel. 2014. “Private Financial
Transfers, Family Income, and the Great Recession.” Journal of Marriage and
Family 76(5): 1011–24.
Gould-Werth, Alix, and H. Luke Shaefer. 2012. “Unemployment Insurance
Participation by Education and by Race and Ethnicity.” Monthly Labor Review
135 (October): 28–41.
Kaiser Commission on Medicaid and the Uninsured. 2009. American Recovery and
Reinvestment Act (ARRA): Medicaid and Health Care Provisions. Washington,
D.C.: Kaiser Family Foundation.
———. 2011. “Medicaid Enrollment: June 2010 Data Snapshot.” Publication
no. 8050–03. Washington, D.C.: Kaiser Family Foundation.
Mattingly, Marybeth, and Elizabeth Kneebone. 2012. “Share of Tax Filers
Claiming EITC Increases Across States and Place Types Between 2007 and
2010.” Issue Brief no. 57. Durham: Carsey Institute, University of New
Hampshire.
Moffitt, Robert A. 2013. “The Great Recession and the Social Safety Net.” Annals
of the American Academy of Political and Social Science 650(1): 143–66.
Mykerezi, Elton, and Bradford Mills. 2010. “The Impact of Food Stamp Program
Participation on Household Food Insecurity.” American Journal of Agricultural
Economics 92(5): 1379–91.
Mykyta, Laryssa, and Suzanne Macartney. 2012. “Sharing a Household: Household
Composition and Economic Well-Being: 2007–2010.” Current Population
Reports, series P60, no. 242. Washington: U.S. Census Bureau. Accessed March
12, 2016. https://www.census.gov/prod/2012pubs/p60-242.pdf.
Nichols, Austin, and Margaret Simms. 2012. Racial and Ethnic Differences
in Receipt of Unemployment Insurance Benefits During the Great Recession.
Washington, D.C.: The Urban Institute.
Nord, Mark, Alisha Coleman-Jensen, Margaret Andrews, and Steven Carlson. 2010.
“Household Food Security in the United States, 2009.” Economic Research Report
no. ERR-108. Washington: U.S. Dept. of Agriculture Economic Research Service.
Pavetti, Ladonna, and Dottie Rosenbaum. 2010. “Creating a Safety Net That
Works When the Economy Doesn’t: The Role of the Food Stamp and TANF
Programs.” Paper Prepared for The Georgetown University and Urban
Institute Conference on Reducing Poverty and Economic Distress after ARRA.
Washington, D.C. (January 15, 2010). Accessed March 12, 2016. http://www.
urban.org/sites/default/files/alfresco/publication-pdfs/412068-Creatinga-Safety-Net-That-Works-When-the-Economy-Doesn-t.PDF.
Pilkauskas, Natasha V., Janet Currie, and Irwin Garfinkel. 2012. “The Great
Recession, Public Transfers, and Material Hardship.” Social Service Review
86(3): 401–27.
Pilkauskas, Natasha V., Irwin Garfinkel, and Sara S. McLanahan. 2014. “The
Prevalence and Economic Value of Doubling Up.” Demography 51(5): 1667–76.
Sell, Katherine, Sarah Zlotnik, Kathleen Noonan, and David Rubin. 2010.
“The Effect of Recession on Child Well-Being: A Synthesis of the Evidence
by PolicyLab, the Children’s Hospital of Philadelphia.” Philadelphia, Pa.:
Children’s Hospital of Philadelphia, PolicyLab.
Simms, Margaret, and Daniel Kuehn. 2008. “Unemployment Insurance During
a Recession.” Recession and Recovery, no. 2. Washington, D.C.: The Urban
Institute.
public and private transfers87
Taylor, Paul, Rakesh Kochhar, D’Vera Cohn, Jeffrey S. Passel, Gabriel Velasco,
Seth Motel, and Eileen Patten. 2011. “Fighting Poverty in a Touch Economy,
Americans Move in with Their Relatives.” Washington, D.C.: Pew Research
Center.
U.S. Senate. Committee on Finance. 2009. Unemployment Insurance Benefits:
Where Do We Go from Here? 111th Cong., 1st sess., S. Hearing 111–956,
September 15 (statement of Gary Burtless, senior fellow in economics, The
Brookings Institution). Washington, D.C.: Government Printing Office. Accessed
March 12, 2016. http://permanent.access.gpo.gov/gpo8858/65459.pdf.
Wiemers, Emily E. 2014. “The Effect of Unemployment on Household
Composition and Doubling Up.” Demography 51(6): 2155–78.
Chapter 4
Mothers’ and Fathers’ Health
Janet Currie and Valentina Duque
E
conomic recessions can represent huge economic and psychological
shocks for many households, and in particular for the most vulnerable.
Because they do, these crises may have a significant impact on parents’
health. This chapter describes how the health of parents has evolved over
time, from their child’s birth up to age nine, and then investigates how
the tremendous rise in the unemployment rate at the start of the Great
Recession was associated with changes in parent’s health. A key contribution of this chapter is that it provides evidence about both mothers and
fathers. Changes in parents’ health could be an important mechanism
through which macroeconomic fluctuations could impact children in both
the short and long term.
We focus on measures of physical health and on health behaviors and
we analyze each of these outcomes for families with different levels of
maternal and paternal education at the time of the child’s birth. Physical
health is captured using self-reported indicators of physical health status
(whether parents consider their health to be good, fair, or poor rather
than excellent or very good) and by whether they report having health
problems that limit their work or study-related activities. Although these
self-reported measures have not been medically verified, they have been
widely used in previous studies of population health and have been found
to be highly correlated with medically determined health status.1
Health behaviors are measured using indicators of substance use. In
particular, we focus on binge drinking—whether a mother (father) drank
four (five) or more glasses of alcohol on one occasion in the last year—
and drug use, specifically whether they used one or more drugs (the list
includes illegal drugs, sedatives, tranquilizers, amphetamines, or other)
without a doctor’s prescription, in larger amounts than prescribed, or for a
longer period than prescribed. Further information on how we construct
these outcomes is provided in the data appendix.
We show that parents with a high school diploma or less at the time
of the child’s birth typically experience worsening physical health over
time relative to parents with more education. Interestingly, it is parents
in the middle of the education distribution who are more likely to adopt
health-compromising behaviors. We also find that the substantial rise in
mothers’ and fathers’ health89
the unemployment rate that defined the start of the Great Recession was
associated with deteriorations in physical health and increases in substance
use. Simulated predictions may suggest that the crisis accentuated gaps
in health outcomes between more and less advantaged families. Mothers
and fathers with high school or less were significantly more likely to report
worse health status and more health problems than those with at least
some college education. However, those who became more likely to binge
drink and to use drugs were not necessarily the least-educated groups.
Interestingly, our findings also suggest that mothers were more affected
than fathers in terms of physical health.
RESEARCH ON RECESSIONS AND HEALTH
Prior research on the effects of recessions on health has come to mixed
conclusions. Studies using state-level data often find positive relationships
between unemployment and aggregate measures of health. These studies
argue that one possible explanation for this result is that people adopt
better health behaviors during recessions, becoming less likely to drink or
smoke, and with more time to exercise, cook healthy meals, and sleep.2
On the other hand, a growing body of literature using individual-level
data finds that recessions are bad for people’s health, and argues that this
may be due to the stress associated with losing a job, reductions in income
and wealth, or other material hardships.3 Even among those who do not
lose their jobs or wealth, the uncertainty associated with the collapse of the
economy may have exerted some toll on people’s health. Some researchers
find that during times of high unemployment, individuals are more likely
to smoke and binge drink, especially those who are more likely to become
unemployed.4 To our knowledge, little research investigates the impacts
of economic recessions on drug use.
A number of reasons explain why the impact of recession on people’s
health might differ between groups. First, those with less education,
income, and wealth are more vulnerable to labor market fluctuations and
less capacity to buffer shocks than more advantaged groups. Second, individuals with low education and fewer economic resources are more likely
to work in low-quality jobs and precarious work environments that could
expose them to higher physical and mental health risks and less access
to health insurance.5 Third, cumulative socioeconomic disadvantage has
been shown to negatively affect people’s physical health. 6 Prolonged
periods of stress could undermine health.7 Fourth, more disadvantaged
groups such as less-educated or unmarried mothers may suffer more during recessions because their dual role of being the primary breadwinner
and the primary caretaker of their child may limit their capacity to insure
against the consequences of economic shocks.8
90
children of the great recession
As this summary shows, common threads are evident in the literature.
Most studies have focused on measuring the health impacts of recessions
on working-age males because of their traditionally strong labor force
attachment. Thus, little research has explored the effects of business cycles
on other groups such as mothers. Second, most studies have perforce
focused on the mild recessions that predated the Great Recession. These
studies may provide less conclusive evidence because the effects of milder
shocks may be harder to detect. Last, few studies have used data that allow
the researcher to track individuals over time. Studies using aggregate
state-level data or cross-sectional data cannot explore changes in individual health or health behaviors in response to changes in economic conditions, a design that provides more compelling evidence on the relationship
between recessions and health.
Janet Currie, Valentina Duque, and Irwin Garfinkel use the Fragile
Families longitudinal data to examine the impacts of the 2007 recession
on mothers’ health.9 They find that increases in state unemployment rates
decreased mothers’ self-reported health status and increased their smoking
and drug use. These declines were particularly concentrated among black
and Hispanic, less-educated, and unmarried mothers relative to mothers
with better socioeconomic prospects. Two main differences between the
Currie study and this chapter are that we provide evidence on the effects
of the Great Recession on both mothers’ and fathers’ health and health
behaviors and we focus on a much longer period (2000 to 2010), whereas
the Currie study focuses on a shorter period before the Great Recession.
The longer period provides more temporal and geographic variation in the
dynamics of the local labor market conditions.
In sum, no consensus has been reached on how changes in the economy are related to people’s health and health behaviors. Moreover, little
is known about the effects of unemployment on the health of mothers
and fathers, especially those in fragile families. This chapter helps fill
these gaps.
HEALTH OF PARENTS THROUGH AGE NINE
We begin our analysis of the effects of the Great Recession on mothers’
and fathers’ health by graphically showing trends in physical health outcomes and health behaviors across the first four waves of the survey, corresponding to child ages one, three, five, and nine. Five patterns stand
out. First, the physical health of parents has deteriorated over time, consistent with a natural aging process. Second, differences are large across
education levels. Parents with a high school diploma or less persistently
report worse physical health than more advantaged mothers and fathers.
Third, the decline in physical health has been more pronounced between
years five (2003 to 2006) and nine (2007 to 2010), the period of the
mothers’ and fathers’ health91
Great Recession, and this change has been mostly driven by less-educated
groups. Fourth, in terms of drinking and drug use, the increase over time
is persistent, and the rise in the last two waves more pronounced. Fifth,
those more likely to have health-compromising behaviors—drinking and
drug use—are not necessarily the least-educated parents.
Physical Health
Figures 4.1 and 4.2 show mean values of self-reported health status for
mothers and fathers across education groups. Overall, we find that families in which the mother has a high school diploma or less are almost
20 percentage points more likely to report worse health status over time
than those in which the mother had at least some college at the time of
the child’s birth. These differences are more striking for college-educated
mothers than for the rest (a difference of at least 25 percentage points).
We observe similar dynamics among fathers. Moreover, less-educated
women and men are more likely to experience an increase in the probability of reporting worse health between child ages five and nine (by almost
5 percentage points), a result not observed for high-skilled women and
men (who are actually less likely to report worse physical health between
years five and nine).
We find a similar pattern for other measures of physical health, with
even more divergence by parents’ educational background. Figures 4.3
and 4.4 focus on whether mothers and fathers report having a physical
Mothers’ Health Status is Fair or Poor (%)
Figure 4.1 Mothers’ Health Status Is Fair or Poor
0.6
College +
0.5
0.4
Some college
0.3
High school
0.2
Less than
high school
0.1
0
1
3
5
(1999–2001) (2001–2003) (2003–2006)
9
(2007–2010)
Child’s Age-Year
Source: Authors’ compilation based on the Fragile Families and Child Wellbeing Study.
92
children of the great recession
Figure 4.2 Fathers’ Health Status Is Fair or Poor
Fathers’ Health Status is Fair or Poor (%)
0.6
College +
0.5
0.4
Some college
0.3
High school
0.2
Less than
high school
0.1
0
1
3
5
(1999–2001) (2001–2003) (2003–2006)
9
(2007–2010)
Child’s Age-Year
Source: Authors’ compilation based on the Fragile Families and Child Wellbeing Study.
health problem that limits their ability to work or study. Mothers who had
a high school diploma or less at the child’s birth exhibit a higher incidence
of health problems than those with at least some college, but also face a
higher increase in the probability of having such a health problem in year
nine. For example, 11 percent of families in which the mother has less than
a high school diploma report a work-limiting health problem in year five, a
Figure 4.3 Mothers’ Health Problem that Limits Work
Mothers’ Health Problem that
Limits Work (%)
0.18
0.16
College +
0.14
0.12
Some college
0.10
High school
0.08
0.06
Less than
high school
0.04
0.02
0
1
3
5
(1999–2001) (2001–2003) (2003–2006)
9
(2007–2010)
Child’s Age-Year
Source: Authors’ compilation based on the Fragile Families and Child Wellbeing Study.
mothers’ and fathers’ health93
Figure 4.4 Fathers’ Health Problem that Limits Work
Fathers’ Health Problem that
Limits Work (%)
0.16
0.14
College +
0.12
Some college
0.10
0.08
High school
0.06
Less than
high school
0.04
0.02
0
1
3
5
(1999–2001) (2001–2003) (2003–2006)
9
(2007–2010)
Child’s Age-Year
Source: Authors’ compilation based on the Fragile Families and Child Wellbeing Study.
number that rises to 16 percent in year nine. The comparable percentages
for mothers with a college education are 3 percent and 4 percent. Among
fathers, the most-educated ones had very low rates of work-limiting health
conditions (3 percent). Also, a considerable increase in health problems for
those with low education coincided with the start of the Great Recession
(an increase of at least 2 percentage points).
Health Behaviors
Figures 4.5, 4.6, 4.7, and 4.8 show trajectories in important health behaviors: binge drinking and drug use. Interestingly, we find that families in
which the mother or father has low education are not necessarily those
most likely to drink or use drugs. In fact, we find that highly educated
mothers are significantly more likely to drink four or more glasses of alcohol on a given occasion (15 percent versus an average of 12 percent for the
rest of the sample), whereas mothers with high school diplomas are more
likely to use drugs on their own (16 percent versus 11 percent for the rest
of the sample). Fathers with college or more tend to binge drink more
(42 percent versus 29 percent), whereas fathers in the middle of the
education distribution are more likely to use drugs (12 percent versus
9 percent). The rise in substance use over time has also been more pronounced for families in these middle education groups. Binge drinking
and drug use have increased substantially between year five and year nine
because substance use usually declines with age.
Although it is not shown here, we also explored trajectories in smoking
and we found remarkable differences by education levels; less-educated
94
children of the great recession
Figure 4.5 Mothers’ Binge Drinking
Mothers’ Binge Drinking (%)
0.20
0.18
College +
0.16
0.14
Some college
0.12
0.10
High school
0.08
0.06
Less than
high school
0.04
0.02
0
1
3
5
(1999–2001) (2001–2003) (2003–2006)
9
(2007–2010)
Child’s Age-Year
Source: Authors’ compilation based on the Fragile Families and Child Wellbeing Study.
Figure 4.6 Fathers’ Binge Drinking
Fathers’ Binge Drinking (%)
0.50
College +
0.45
0.40
Some college
0.35
0.30
High school
0.25
0.20
Less than
high school
0.15
0.10
0.05
0
1
3
5
(1999–2001) (2001–2003) (2003–2006)
9
(2007–2010)
Child’s Age-Year
Source: Authors’ compilation based on the Fragile Families and Child Wellbeing Study.
mothers’ and fathers’ health95
Figure 4.7 Mothers’ Drug Use
0.09
Mothers’ Drug Use (%)
0.08
College +
0.07
Some college
0.06
0.05
High school
0.04
0.03
Less than
high school
0.02
0.01
0
1
3
5
(1999–2001) (2001–2003) (2003–2006)
9
(2007–2010)
Child’s Age-Year
Source: Authors’ compilation based on the Fragile Families and Child Wellbeing Study.
Figure 4.8 Fathers’ Drug Use
Fathers’ Drug Use (%)
0.25
College +
0.20
Some college
0.15
High school
0.10
Less than
high school
0.05
0
1
3
5
(1999–2001) (2001–2003) (2003–2006)
9
(2007–2010)
Child’s Age-Year
Source: Authors’ compilation based on the Fragile Families and Child Wellbeing Study.
96
children of the great recession
mothers and fathers tend to smoke more than those with some college
or more, and these differences remain relatively constant over time. On
average, 20 percent of less-educated mothers and more than 30 percent
of fathers smoke.
HEALTH OUTCOMES IN FRAGILE FAMILIES
We now turn to a more formal analysis of the relationship between economic downturns and health. As in prior chapters, we use an empirical
model that takes advantage of the longitudinal nature of the data by
accounting for an individual’s observed and unobserved characteristics as
well as other time-varying factors (for a detailed description of the empirical model, see the appendix).
Table 4.A1 focuses on the influence of the unemployment rate on
mothers’ and fathers’ self-reported health status. We find a negative
association between economic downturns and mothers’ physical health
(a positive coefficient is interpreted as a high probability of reporting fair
or poor health), which remains significant and in similar magnitude even
after accounting for a mother’s individual fixed-effect. Fathers’ physical
health, in contrast, does not seem to be affected by the fluctuations in
the unemployment rate. In tables 4.A2 and 4.A3 (model 1), we estimate
models by education groups.
For ease of interpretation, we simulate the potential effects of recessions by estimating a parent’s health outcome when the unemployment
rate is 5 and 10 percent in the last year of data, which approximates the
size of the increase in the unemployment rate during the Great Recession,
and we present these predictions in figures 4.9 through 4.15. In general,
results from the empirical analyses support our previous descriptive findings. Mothers and fathers experienced significant physical health setbacks
and were more likely to increase their use of substances; however, results
varied quite a bit by educational background. Although less-educated parents were more likely to report worsening health status and had a higher
probability of health problems when unemployment was high, groups in
the middle or high end of the education distribution were significantly
more likely to binge drink (or smoke). Moreover, we found little change
in drug use among mothers and fathers. We discuss some possible explanations for these findings later.
Physical Health
Figures 4.9, 4.10, 4.11, and 4.12 show changes in mothers’ and fathers’
physical health associated with a simulated increase of 5 percentage points
in the unemployment rate. Results for the full sample indicate that the
crisis was associated with a significant deterioration in women’s physical
health. Mothers who did not complete high school were more likely to
mothers’ and fathers’ health97
Mothers’ Health Status is Fair or
Poor (%)
Figure 4.9 Effects of a Recession on Mothers’ Health Status
0.7
+31.5%***
0.6 +20.0%***
+26.3%***
+4.0%
0.5
0.4
0.3
–10.7%
0.2
UR 5 percent
UR 10 percent
0.1
0
All
Less than
high school
High
school
Some
college
College +
Source: Authors’ compilation based on the Fragile Families and Child Wellbeing Study.
***p < .01; **p < .05; *p < .1
report good, fair, or poor health (rather than excellent or very good) than
any of their counterparts (the probability increased from 48 percent to
63 percent as unemployment went from 5 percent to 10 percent for those
with less than a high school diploma and from 46 percent to 58 percent
for those who completed high school). Less-educated mothers were also
more likely to have a health problem that limited their work (an increase
from 13 percent to 22 percent for those with a high school diploma
and from 11 percent to 15 percent for those with some postsecondary
education). College-graduate mothers actually experienced improve-
Fathers’ Health Status is Fair
or Poor (%)
Figure 4.10 Effects of a Recession on Fathers’ Health Status
0.6
0.5
0.4
–6.6%
–2.4%
+2.1%
+69.2%***
UR 5 percent
0.3
–6.2%
UR 10 percent
0.2
0.1
0
All
Less than
high school
High
school
Some
college
College +
Source: Authors’ compilation based on the Fragile Families and Child Wellbeing Study.
***p < .01; **p < .05; *p < .1
98
children of the great recession
Figure 4.11 Effects of a Recession on Mothers’ Health Problem
that Limits Work
Mothers’ Health Problem that
Limits Work (%)
0.25
+71.0%***
0.20
+45.9%***
+35.2%
0.15
UR 5 percent
+55.4%*
UR 10 percent
–85.7%***
0.10
0.05
0
All
Less than
high school
High
school
Some
college
College +
Source: Authors’ compilation based on the Fragile Families and Child Wellbeing Study.
***p < .01; **p < .05; *p < .1
ments in their physical health (that is, they were significantly less likely
to report health problems with the rise in the unemployment rate).
Results also show that changes in the unemployment rate were associated
with a decline in fathers’ physical health (fathers were 32 percent more likely
to report health problems that limited their work). In contrast to the finding that the health of highly educated women improved during the Great
Figure 4.12 Effects of a Recession on Fathers’ Health Problem
that Limits Work
Fathers’ Health Problem that
Limits Work (%)
0.25
+31.1%
0.20
0.15
+31.9%**
+43.1%
UR 5 percent
+91.2%
0.10
UR 10 percent
+24.0%
0.05
0
All
Less than
high school
High
school
Some
college
College +
Source: Authors’ compilation based on the Fragile Families and Child Wellbeing Study.
***p < .01; **p < .05; *p < .1
mothers’ and fathers’ health99
Recession, evidence suggests that fathers with some college experienced a
rise in self-reporting poor health (by almost 70 percent).
Health Behaviors
Figures 4.13, 4.14, 4.15, and 4.16 present simulations describing the
relationship between an increase in the unemployment rate and health
behaviors. We find that as the unemployment rate goes from 5 to 10 percent, mothers binge drink more (from 14 percent to 19 percent) and tend
to use more drugs (from 7 percent to 10 percent), although these associations were only marginally statistically different from zero. Moreover,
these changes do not seem to be driven by mothers in the lowest education
groups: mothers with a high school diploma only were more likely to binge
drink (an increase of more than 50 percent) whereas those with a college
degree were substantially more likely to do so (more than 120 percent) and
to increase their drug use (by 72 percent, though this effect is not statistically significant).The large rise in smoking for college-educated mothers
(result not shown) with the change in the unemployment rate (more than
120 percent) is noteworthy. Because few people begin smoking as adults,
this increase likely represents former smokers relapse in the stressful conditions of the Great Recession.
The results for fathers are somewhat weaker. Although neither binge
drinking nor smoking (result not shown) seem to respond to changes in
the unemployment rate, the increase in drug use, though insignificant, is
high (18 percent), particularly among fathers with some college education.
Figure 4.13 Effects of a Recession on Mothers’ Binge Drinking
Mothers’ Binge Drinking (%)
0.25
+122.9%**
+51.2%*
0.20
+29.2%*
+23.9%
–11.1%
0.15
UR 5 percent
UR 10 percent
0.10
0.05
0
All
Less than
high school
High
school
Some
college
College +
Source: Authors’ compilation based on the Fragile Families and Child Wellbeing Study.
***p < .01; **p < .05; *p < .1
100
children of the great recession
Fathers’ Binge Drinking (%)
Figure 4.14 Effects of a Recession on Fathers’ Binge Drinking
0.50
0.45
0.40
0.35
0.30
0.25
0.20
0.15
0.10
0.05
0
–3.8%
–11.9%
–16.1%
+8.6%
–14.5%
UR 5 percent
UR 10 percent
All
Less than
high school
High
school
Some
college
College +
Source: Authors’ compilation based on the Fragile Families and Child Wellbeing Study.
THE ROLE OF UNCERTAINTY: EFFECTS ON
PARENTS’ HEALTH
Although the local unemployment rate is a good indicator of the probability that an individual is unemployed, it may not capture the stress associated with the anticipation of economic adversity. Stress may result not
only from the actual experience of adversity but also from uncertainty and
the anticipation of adversity. We test this hypothesis by adding two terms
Figure 4.15 Effects of a Recession on Mothers’ Drug Use
Mothers’ Drug Use (%)
0.14
0.12
0.10
+62.4%
+72.5%
+46.5%*
+39.1%
0.08
–78.9%
0.06
UR 5 percent
UR 10 percent
0.04
0.02
0
All
Less than
high school
High
school
Some
college
College +
Source: Authors’ compilation based on the Fragile Families and Child Wellbeing Study.
***p < .01; **p < .05; *p < .1
mothers’ and fathers’ health101
Figure 4.16 Effects of a Recession on Fathers’ Drug Use
Fathers’ Drug Use (%)
0.25
0.20
+3.2%
+18.9%
+18.5%
+31.5%
0.15
0.10
UR 5 percent
+12.0%
UR 10 percent
0.05
0
All
Less than
high school
High
school
Some
college
College +
Source: Authors’ compilation based on the Fragile Families and Child Wellbeing Study.
to our baseline model: one that captures positive changes (increase) in
the unemployment rate (that is, things getting worse) and one that captures negative changes (decline) (things improving). Thus, if the rate and
direction of change in unemployment are significantly related to parents’
health, we could argue that uncertainty, fear, or anticipation of becoming
unemployed could be one possible mechanism through which recessions
affect people’s health.
Results are shown in tables 4.A2 and 4.A3 (model 2) and overall suggest two interesting findings. First, as the unemployment (positively)
accelerates, less-educated mothers are more likely to report worse physical
health (more health problems) than highly educated mothers. Second,
as things get worse, mothers with college degrees or more education are
significantly more likely to binge drink and use drugs (results obtained
from individual fixed-effects models), whereas high school mothers tend
to smoke more (result not shown). Interestingly, mothers with the least
education show no change in their health-related behaviors. Results for
fathers are similar but more imprecise: less-educated fathers tend to suffer
more in terms of physical health and their more advantaged counterparts
report to smoke more and use drugs as the economy deteriorates.
OTHER RESULTS
Controlling for Individual Unemployment
An additional potential pathway through which the effects of the unemployment rate could lead to changes in health is parent’s labor market status.
We explore whether this is the case by adding individual level measures of
102
children of the great recession
parents’ unemployment to our core model. Although we do not find significant results indicating a particular role of individual unemployment in
our fixed-effects models, results from our pooled regression models suggest
that mothers’ unemployment is associated with an increase in both parents’
likelihood of reporting good, fair, or poor health, having health problems
that limit work, as well as smoking and drug use. Results are shown in
table 4.A4 (model 3).
Are the Effects of the Unemployment Rate Stronger During
the Great Recession?
We now test whether the overall effect of the local unemployment rate is
stronger during the Great Recession (which coincided with the year nine
interview) than in other years. Because the Great Recession was exceptionally dramatic, this question is important. We test this by adding an interaction between year nine data and the unemployment rate to our core model.
Table 4.A4 (model 4) shows evidence that the effects of the unemployment
are not particularly different in the last wave to those from previous years
(the dot-com recession in 2001). Interestingly, however, we find that drug
use increases during the Great Recession years for both mothers and fathers.
Results by Race-Ethnicity and Relationship Status at Baseline
Last, we estimate differences by race-ethnicity and relationship status (at
baseline) in table 4.A5 using fixed-effects models. We find that mothers
who are black or Hispanic and those who were cohabitating or single at
their child’s birth were more likely to report worse physical health and to
have physical health problems associated with an increase in the unemployment rate. Black mothers were also significantly more likely to binge drink
and to use drugs than their counterparts. Those who were married did not
see significant changes in their physical health or health behaviors.
Results for fathers are somewhat different and statistically weaker.
For instance, we find that more advantaged fathers actually experienced
worse health: white fathers were more likely to report fair or poor health
and married fathers were more likely to report health problems that
limited their work. These results were only marginally statistically significant. In terms of health behaviors, the story is similar to that among
mothers: although Hispanic fathers were more likely to binge drink,
married fathers actually reported lower drinking.
DISCUSSION
This chapter highlights large differences in mothers’ and fathers’ health
by education level at the time of their child’s birth, which is mostly consistent with a large literature about the protective effects of education on
health. Using a descriptive analysis and a more formal empirical strategy that
mothers’ and fathers’ health103
accounts for an individual’s observed and unobserved characteristics as well
as time-varying characteristics, we find that mothers who had a high school
diploma or less when their child was born not only had worse initial physical health but also saw larger declines in their health as a result of economic
fluctuations. For health behaviors, the story is different. Those most likely
to smoke, binge drink, or use drugs were not necessarily the least educated.
Overall, our findings suggest stronger effects on women than on men, even
when we limit our attention to mothers for whom we also have information about the fathers. The most likely explanation for these patterns involve
changes in individual or partner unemployment, income or wealth, as well as
stress or fear. For instance, income declines, especially among the least educated, may be affecting budget constraints and therefore limiting consumption of alcohol and drugs. Moreover, increases in substance use may reflect
a different response to stress. Another possible explanation could be that
parents may lose their medical insurance during recessions, and therefore be
more vulnerable to shocks such as economic downturns; however, Currie
and her colleagues show that the Great Recession had little impact on health
insurance among mothers in fragile families.10
An intriguing result in this chapter is that high unemployment leads to
reports of better health among college-educated mothers at the same time
as they reported more binge drinking. One possible explanation for this
pattern could be related to the fact that more educated groups experienced
relatively milder changes during the crisis compared with other groups;
we discuss some reasons for why this may be the case. First, recent studies
have shown that college-educated adults were not as strongly affected by
the large increase of the unemployment rate in terms of their employment status, income, and wealth. That they experienced little change in
their physical health or that their health actually improved is therefore not
surprising.11 Second, in chapter 2 of this book, Garfinkel and Pilkauskas
find that recessions exacerbated gaps in economic well-being, reducing
family incomes and increasing poverty and economic insecurity. In particular, they found that college-educated families stood apart as being the
least affected by recessions. When unemployment rates were 10 percent
rather than 5 percent, their family incomes were only 5 percent lower.
The three less-educated groups experienced income losses three to four
times greater. Third, previous studies have documented that an increase in
binge drinking for more educated groups could be the result of their relatively better economic position (relatively higher wages) during the Great
Recession. For example, Kerwin Charles and Philip DeCicca also find that
men with high employment opportunities were more likely to drink during recessions. In other words, more binge drinking may actually lead to
exaggerations of wellness. Last, that more educated groups were actually
more sensitive to the rate of change in the unemployment rate (results
shown in tables 4.A2 and 4.A3) may provide some evidence that though
these groups had relatively few economic losses and little change in their
104
children of the great recession
physical health, they may have actually had significant stress and fear due
to vast fluctuations in the economy.
APPENDIX
Data
We use data from waves 1 through 5 of the Fragile Families and Child
Wellbeing Study. We pool the data for the analyses of the effect of the
recession (N ~15,300) and use waves 2 through 5 for measuring dependent variables and unemployment. Covariates are from the baseline—
that is, wave 1—survey.
Measures
The outcomes of interest for this study include four measures of selfreported physical health and health behaviors obtained from telephone
or in-home interviews at the moment of interview, and refer to the last
twelve months. All measures were constructed as binary indicators that
take the value of 1 when a mother or father reports a given condition and
0 otherwise. A value of 1 represents a bad health condition whereas 0 is a
good condition.
Self-rated health status: health status is good, fair, or poor versus
excellent or very good.
Health problem that limits work: has a health problem that limits
work- or study-related activities versus no problem.
Binge drinking: mother (father) drinks four (five) or more glasses of
alcohol in one occasion rather than less than that or nothing on one
occasion in the last year.
Drug use: parent uses one or more drugs (includes illegal drugs,
sedatives, tranquilizers, amphetamines, or other) whether without
a doctor’s prescription or in larger amounts than prescribed or for a
longer period rather than not at all.
The complete list of drugs includes illegal drugs (marijuana or hashish;
cocaine or crack or free base; LSD or other hallucinogens; heroin), sedatives
(including either barbiturates or sleeping pills such as Seconal, Halcion,
Methaqualone), tranquilizers or “nerve pills” (such as Librium, Valium,
Ativan, Meprobamate, Xanax), amphetamines or other stimulants (such
as methamphetamine, Preludin, Dexedrine, Ritalin, “Speed”), analgesics
or other prescription painkillers (note: this does not include normal use
mothers’ and fathers’ health105
of aspirin, Tylenol without codeine, etc., but does include use of Tylenol
with codeine and other Rx painkillers like Demerol, Darvon, Percodan,
Codeine, Morphine, and Methadone), inhalants (such as Amylnitrate,
Freon, Nitrous Oxide (“Whippets”), Gasoline, Spray paint).
Key Independent Variable
For each analysis, the unemployment rate is constructed using a measure of
the average unemployment rate in the sample city over the twelve months
before the interview. This is done to match the period of the outcome
measures.
Key Moderating Variables
We study differences in the trajectories over time, and in the effects of the
Great Recession, on health outcomes by maternal and paternal education.
Education is coded as less than a high school diploma or GED certificate,
a high school diploma, some college or an associate degree or technical
degree, and undergraduate degree or greater.
Control Variables
Our preferred models include individual fixed effects that absorb all fixed
characteristics of our subjects. In models without fixed effects, we include
a number of covariates in our models all measured at the first survey
wave (baseline). These include: mother’s age at the birth, immigrant status (foreign born), number of children in the household, a measure of
whether the mother grew up with both parents at age fifteen, as well as
city (twenty dummies for each sample city) and survey year fixed-effects
(twelve time-varying survey year dummies).
Method
The figures that plot the trajectories of each health outcome measure over
time present the mean levels of each outcome at each survey wave. All
means are weighted to be representative of births in the twenty study cities.
To study the effects of the Great Recession, we conduct linear regressions
for binary outcomes (linear probability models, or LPM) using the pooled
data (waves 2 through 5). The analyses are clustered at both the city and
individual level to account for within city and within person clustering–
nonindependence. All analyses by moderating characteristics are stratified
by mothers’ or fathers’ education (less than high school, high school, some
college, college or greater).
Unemployment rate
Relationship status
Married
Cohabiting
Mother’s age
Race-ethnicity
Black
Hispanic
Other
Immigrant
Mother’s education
High school
Some college or associates degree
College +
Number of children in household
Lived with both parents at age fifteen
Interview year
2000
2001
2002
2003
2004
2005
2006
2007
(0.005)
(0.030)
(0.021)
(0.027)
(0.026)
(0.028)
(0.026)
(0.081)
(0.033)
0.018***
-0.037
-0.036*
-0.046*
-0.075***
0.008
-0.01
-0.11
0.008
With Individual
Fixed Effects
(0.017)
(0.024)
(0.031)
(0.022)
(0.013)
(0.010)
(0.021)
(0.004)
(0.008)
0.011
0.046*
-0.005
0.025
0.053***
-0.024**
-0.176***
0.009**
-0.040***
(0.025)
(0.010)
(0.018)
(0.015)
(0.018)
(0.015)
(0.063)
(0.010)
(0.022)
(0.014)
(0.001)
-0.075***
0.008
0.007***
-0.041
-0.036***
-0.053***
-0.071***
0.00
-0.008
-0.146**
0.022**
(0.004)
Without Individual
Fixed Effects
0.016***
Mothers
Table 4.A1 Full Regression Results, Parents’ Physical Health
-0.031
-0.041*
-0.079**
-0.066**
0.006
-0.017
0.013
0.010
0.001
(0.034)
(0.025)
(0.031)
(0.030)
(0.032)
(0.029)
(0.105)
(0.039)
(0.006)
With Individual
Fixed Effects
-0.045
-0.036*
-0.076***
-0.066***
0.007
-0.016
-0.047
0.031*
(0.037)
(0.019)
(0.025)
(0.024)
(0.028)
(0.029)
(0.061)
(0.016)
(0.014)
(0.015)
(0.020)
(0.004)
(0.013)
(0.013)
(0.021)
(0.038)
(0.023)
-0.022*
0.015
0.039
0.033
0.061***
-0.053***
-0.141***
0.007*
-0.019
(0.012)
(0.015)
(0.001)
-0.026**
0.009
0.006***
(0.004)
Without Individual
Fixed Effects
0.003
Fathers
(0.027)
(0.032)
(0.055)
(0.025)
0.063**
-0.006
-0.074
-0.651***
15,362
(0.016)
(0.020)
(0.037)
(0.005)
(0.010)
(0.012)
(0.009)
(0.011)
(0.011)
(0.010)
(0.008)
(0.008)
(0.006)
(0.008)
(0.008)
(0.008)
(0.008)
(0.009)
(0.014)
(0.011)
(0.011)
(0.009)
(0.029)
0.048***
0.003
-0.077**
0.013***
-0.035***
0.008
-0.008
-0.006
0.042***
-0.051***
0.020**
0.029***
-0.029***
-0.071***
0.012
-0.01
-0.023***
-0.001
-0.047***
-0.062***
0.005
0.003
-0.802***
15,362
-0.663***
11,890
0.023
0.041
0.012
(0.029)
(0.032)
(0.035)
(0.065)
(0.035)
(0.026)
(0.051)
(0.009)
(0.017)
(0.017)
(0.014)
(0.017)
(0.017)
(0.013)
(0.012)
(0.010)
(0.010)
(0.012)
(0.013)
(0.012)
(0.012)
(0.012)
(0.015)
(0.013)
(0.016)
(0.013)
(0.041)
0.024
0.025
0.107**
-0.072***
-0.061***
-0.063***
-0.081***
-0.060***
-0.072***
-0.089***
-0.011
-0.020*
-0.093***
-0.049***
-0.089***
-0.108***
-0.069***
-0.128***
-0.020
-0.076***
-0.005
-0.055***
-0.764***
11,890
Source: Authors’ compilation based on the Fragile Families and Child Wellbeing Study.
Note: Standard errors and t-stats in parentheses. Covariates are measured at the baseline survey (except year) and are clustered at city and individual level. Model 1 includes
level unemployment rate.
***p < .01; **p < .05; *p < .1
2008
2009
2010
City
Austin
Baltimore
Detroit
Newark
Philadelphia
Richmond
Corpus Christi
Indianapolis
Milwaukee
New York
San Jose
Boston
Nashville
Chicago
Jacksonville
Toledo
San Antonio
Pittsburgh
Norfolk
Constant
N (person * wave)
Health status is fair or poor
Unemployment
0.018***
rate (model 1)
(0.005)
Unemployment
0.017***
rate (model 2)
(0.005)
Increasing
-0.000
unemployment (0.000)
rate
Decreasing
0.001
unemployment (0.001)
rate
N (individuals
8,848
* wave)
Number of
2,301
individuals
Health problem limits work
Unemployment
0.010***
rate (model 1)
(0.003)
Unemployment
-0.000
rate (model 2)
(0.001)
Increasing
0.009***
(0.003)
unemployment
rate
Decreasing
-0.000*
(0.000)
unemployment
rate
N (individuals
2,904
* wave)
Number of
789
individuals
All
0.003
(0.002)
-0.002
(0.002)
2,721
735
0.008
(0.005)
0.006
(0.005)
-0.000*
(0.000)
0.001
(0.001)
885
239
0.001
(0.002)
2,978
823
0.019***
(0.006)
0.001
(0.001)
0.015***
(0.005)
0.000
(0.000)
1,242
344
174
654
30
115
-0.003***
(0.001)
-0.002
(0.001)
4,549
15,345
-0.001
(0.001)
0.006***
(0.002)
0.005**
(0.002)
-0.000*
(0.000)
-0.014***
(0.006)
-0.018***
(0.006)
-0.001
(0.000)
0.010*
(0.005)
0.008
(0.005)
-0.000
(0.000)
15,362
0.001
(0.001)
0.016***
(0.004)
0.016***
(0.004)
0
(0.000)
All
4,549
627
-0.001
(0.003)
-0.005
(0.013)
-0.004
(0.014)
0.001
(0.001)
College +
172
567
2,108
0.004
(0.010)
0.005
(0.010)
-0.000
(0.001)
Some
College
0.024***
(0.009)
0.022**
(0.009)
-0.001
(0.001)
High
School
0.030***
(0.009)
0.030***
(0.009)
0.000
(0.000)
Less than
High
School
With Individual Fixed Effects
1,206
4,168
0.003
(0.002)
0.032***
(0.009)
0.033***
(0.009)
0.000
(0.001)
Less than
High
School
1,206
4,164
-0.000
(0.001)
1,175
4,209
-0.000
(0.001)
-0.003
(0.004)
-0.004
(0.004)
-0.000
(0.000)
1,175
4,218
-0.001
(0.002)
0.013
(0.008)
0.010
(0.007)
-0.001***
(0.000)
High
School
673
2,367
-0.002
(0.001)
0.012**
(0.005)
0.010**
(0.005)
-0.000*
(0.000)
673
2,370
0.003
(0.002)
0.007
(0.008)
0.008
(0.009)
0.000
(0.000)
Some
College
Without Individual Fixed Effects
0.014***
(0.005)
0.014***
(0.005)
-0.000
(0.000)
Table 4.A2 Coefficients and Standard Errors, All Outcomes by Maternal Education
559
1,969
-0.003**
(0.001)
-0.017**
(0.007)
-0.019**
(0.008)
-0.000
(0.000)
559
1,969
-0.001
(0.002)
0.000
(0.014)
0.002
(0.013)
0.001
(0.001)
College +
401
139
487
170
38
112
-0.001
(0.002)
-0.000
(0.001)
0.001
(0.001)
4,437
11,229
-0.001*
(0.000)
0.005
(0.004)
0.005
(0.005)
-0.000
(0.000)
0.010
(0.006)
0.010*
(0.006)
0.000
(0.000)
11,475
-0.010
(0.009)
-0.010
(0.010)
0.001**
(0.001)
208
275
216
-0.000
(0.001)
0.005
(0.007)
0.005
(0.007)
-0.000
(0.000)
595
779
0.001
(0.003)
0.008**
(0.004)
0.008**
(0.004)
-0.000
(0.000)
4,434
0.000
(0.002)
-0.000
(0.002)
0.026**
(0.013)
0.030**
(0.013)
0.002*
(0.001)
75
-0.004
(0.008)
-0.004
(0.009)
-0.001
(0.001)
0.014*
(0.008)
0.014*
(0.008)
0.000
(0.000)
1,520
3,832
-0.001
(0.001)
0.008
(0.007)
0.007
(0.007)
-0.000
(0.000)
1,518
3,821
-0.000
(0.001)
0.007
(0.009)
0.006
(0.009)
0.000
(0.000)
1,370
3,585
-0.001
(0.001)
0.006
(0.008)
0.006
(0.008)
0.000
(0.000)
1,369
3,574
-0.001
(0.001)
0.015**
(0.006)
0.014**
(0.006)
-0.000
(0.000)
1,069
2,828
-0.001
(0.001)
0.008
(0.007)
0.007
(0.007)
-0.000
(0.000)
1,069
2,826
-0.000
(0.002)
0.003
(0.006)
0.002
(0.006)
-0.001
(0.001)
473
1,243
-0.001
(0.002)
-0.015*
(0.009)
-0.016*
(0.009)
0.000
(0.001)
473
1,241
0.002
(0.003)
0.015
(0.012)
0.020
(0.013)
0.002***
(0.001)
Source: Authors’ compilation based on the Fragile Families and Child Wellbeing Study.
Note: Standard errors and t-stats in parentheses. Model 1 includes unemployment rate as a level. Model 2 includes unemployment rate as a level as well as rate of change in
unemployment rate.
***p < .01; **p < .05; *p < .1
Four or more drinks on one occasion
Unemployment
0.008*
0.007
(0.004)
(0.008)
rate (model 1)
Unemployment
0.000
0.007
(0.001)
(0.008)
rate (model 2)
0.008*
-0.000
Increasing
(0.004)
(0.000)
unemployment
rate
Decreasing
-0.000
-0.000
(0.000)
(0.002)
unemployment
rate
N (individuals
2,381
785
* wave)
Number of
844
284
individuals
Drug use
Unemployment
0.007*
0.010
(0.004)
(0.007)
rate (model 1)
Unemployment
-0.000
0.009
(0.001)
(0.007)
rate (model 2)
Increasing
0.006*
-0.000
(0.004)
(0.000)
unemployment
rate
Decreasing
0.000
-0.001
(0.000)
(0.001)
unemployment
rate
1,565
562
N (individuals
* wave)
Number of
551
203
individuals
Health status is fair or poor
Unemployment
0.001
(0.006)
rate (model 1)
Unemployment
0.002
(0.006)
rate (model 2)
Increasing
-0.000
(0.000)
unemployment
rate
Decreasing
-0.001
(0.001)
unemployment
rate
N (individuals
5,444
* wave)
Number of
1,598
individuals
Health problem limits work
Unemployment
0.007**
(0.003)
rate (model 1)
All
-0.003
(0.002)
-0.003
(0.002)
1,717
496
0.010
(0.006)
0.002
(0.002)
1,748
518
0.010
(0.007)
0.010
(0.007)
261
921
0.033***
(0.012)
0.009
(0.010)
-0.000
(0.001)
Some
College
-0.002
(0.010)
-0.000
(0.011)
0.000
(0.001)
High
School
-0.007
(0.012)
-0.010
(0.011)
0.000
(0.001)
Less than
High
School
With Individual Fixed Effects
0.002
(0.006)
173
622
-0.003
(0.002)
0.003
(0.011)
0.009
(0.010)
-0.000
(0.001)
College +
All
0.010***
(0.004)
4,026
11,890
-0.001
(0.001)
0.003
(0.004)
0.002
(0.004)
-0.000*
(0.000)
Fathers
0.016**
(0.007)
1,141
3,307
0.001
(0.002)
-0.005
(0.009)
-0.005
(0.009)
-0.000
(0.000)
Less than
High
School
0.015
(0.009)
1,123
3,450
-0.003
(0.002)
0.003
(0.006)
-0.001
(0.006)
-0.000
(0.001)
High
School
0.008
(0.007)
671
2,100
0.001
(0.002)
0.028**
(0.011)
0.031***
(0.011)
-0.000
(0.001)
Some
College
Without Individual Fixed Effects
Table 4.A3 Coefficients and Standard Errors, All Outcomes by Paternal Education
0.001
(0.006)
556
1,851
-0.002
(0.003)
-0.008
(0.008)
-0.009
(0.008)
-0.001
(0.000)
College +
Unemployment
0.008**
0.012
(0.003)
(0.007)
rate (model 2)
Increasing
0.000
0.001*
(0.000)
(0.000)
unemployment
rate
Decreasing
0.001
0.001
(0.001)
(0.002)
unemployment
rate
N (individuals
2,008
634
* wave)
Number of
595
189
individuals
Four or more drinks on one occasion
Unemployment
-0.003
-0.009
(0.007)
(0.014)
rate (model 1)
Unemployment
-0.003
-0.011
(0.007)
(0.015)
rate (model 2)
Increasing
0.000
0.000
(0.000)
(0.001)
unemployment
rate
Decreasing
-0.001
-0.002
(0.001)
(0.003)
unemployment
rate
N (individuals
2,595
777
* wave)
Number of
976
296
individuals
0.009
(0.006)
0.000
(0.000)
-0.000
(0.001)
336
96
0.007
(0.015)
-0.012
(0.012)
0.001
(0.001)
-0.002
(0.002)
443
161
0.009
(0.007)
0.000
(0.000)
-0.001
(0.001)
667
196
-0.013
(0.013)
-0.013
(0.013)
0.001
(0.001)
-0.001
(0.003)
800
300
120
334
-0.002
(0.002)
-0.013
(0.015)
-0.012
(0.012)
0.001
(0.001)
45
163
-0.000
(0.001)
0.009
(0.006)
0.000
(0.000)
3,785
8,559
-0.001
(0.002)
-0.004
(0.005)
-0.004
(0.005)
0.000
(0.000)
4,024
11,884
-0.000
(0.001)
0.009**
(0.004)
0.000
(0.000)
1,062
2,339
-0.002
(0.002)
0.003
(0.013)
0.001
(0.012)
0.001*
(0.001)
1,140
3,308
0.000
(0.002)
0.016**
(0.007)
0.000
(0.000)
1,054
2,455
-0.002
(0.004)
-0.010
(0.008)
-0.010
(0.010)
0.000
(0.001)
1,123
3,446
-0.002*
(0.001)
0.011
(0.009)
0.000
(0.000)
530
1,340
0.003
(0.003)
-0.009
(0.011)
-0.005
(0.012)
-0.000
(0.001)
556
1,851
0.001
(0.002)
0.001
(0.007)
0.000
(0.000)
(Table continues on p. 112.)
635
1,506
-0.000
(0.003)
-0.003
(0.008)
-0.003
(0.009)
0.000
(0.001)
670
2,098
0.000
(0.001)
0.009
(0.007)
0.000
(0.000)
0.001
(0.011)
0.004
(0.012)
0.001**
(0.001)
0.004
(0.002)
482
184
0.005
(0.005)
0.005
(0.005)
0.000
(0.000)
0.000
(0.001)
1,498
562
Less than
High
School
181
0.006
(0.010)
0.005
(0.010)
0
(0.001)
-0.002
(0.002)
490
High
School
0.007
(0.011)
0.013
(0.009)
0
(0.001)
0
(0.002)
—
Some
College
With Individual Fixed Effects
54
0.002
(0.010)
0.013
(0.009)
0
(0.001)
0
(0.002)
149
College +
All
3,794
0.006
(0.005)
0.007
(0.005)
0.000
(0.000)
0.001
(0.001)
8,615
Fathers
1,066
0.007
(0.010)
0.010
(0.011)
0.001
(0.001)
0.003
(0.002)
2,350
Less than
High
School
1,055
-0.001
(0.009)
-0.003
(0.008)
0.000
(0.001)
-0.000
(0.002)
2,459
High
School
636
0.015
(0.011)
0.018
(0.012)
0.001
(0.001)
0.001
(0.002)
1,512
Some
College
Without Individual Fixed Effects
530
0.007
(0.012)
0.008
(0.013)
-0.001*
(0.000)
-0.001
(0.002)
1,341
College +
Source: Authors’ compilation based on the Fragile Families and Child Wellbeing Study.
Note: Standard errors and t-stats in parentheses. Model 1 includes unemployment rate as a level. Model 2 includes unemployment rate as a level as well as rate of change in
unemployment rate.
***p < .01; **p < .05; *p < .1
Drug use
Unemployment
rate (model 1)
Unemployment
rate (model 2)
Increasing
unemployment
Decreasing
unemployment
N (individuals
* wave)
Number of
individuals
All
Table 4.A3 Continued
0.004*
0.060***
0.009***
-0.004
0.011***
0.007
0.007
0.002
0.005
0.038***
-0.000
0.007*
(0.003)
(0.007)
(0.004)
(0.004)
(0.005)
(0.012)
(0.006)
(0.006)
(0.004)
(0.009)
(0.005)
(0.005)
(0.004)
(0.005)
(0.005)
(0.004)
(0.004)
(0.008)
(0.007)
(0.006)
(0.002)
(0.007)
(0.003)
(0.004)
(0.005)
(0.009)
(0.006)
(0.007)
(0.005)
(0.013)
(0.007)
(0.005)
(0.007)
(0.018)
(0.009)
(0.007)
-0.003
-0.008
-0.006
0.004
0.005
0.025*
-0.003
0.010**
(0.003)
(0.009)
(0.004)
(0.004)
(0.006)
(0.015)
(0.007)
(0.006)
0.008**
-0.005
0.001
0.009**
0.001
0.028*
-0.002
0.005
With Individual Fixed
Effects
0.005
0.054***
-0.001
0.009**
-0.004
-0.004
-0.012**
0.011**
0.010***
0.056***
0.004
0.009*
(0.005)
(0.015)
(0.007)
(0.004)
(0.005)
(0.013)
(0.006)
(0.005)
(0.004)
(0.011)
(0.004)
(0.005)
(0.004)
(0.016)
(0.005)
(0.006)
Without Individual
Fixed Effects
0.002
0.055***
-0.001
0.007
Fathers
Source: Authors’ compilation based on the Fragile Families and Child Wellbeing Study.
Note: Standard errors in parentheses. Model 3 includes unemployment rate and a measure of individual unemployment. Model 4 includes unemployment rate and an
interaction between unemployment rate and year nine, when the Great Recession hit.
***p < .01; **p < .05; *p < .1
0.020***
0.046***
0.016**
0.001
Without Individual
Fixed Effects
(0.006)
(0.013)
(0.007)
(0.007)
With Individual
Fixed Effects
Mothers
Sensitivity of Coefficients, Parents’ Health
Health status is fair or poor
Unemployment rate (model 3)
0.022***
Individual unemployment
0.013
Unemployment rate (model 4)
0.019***
Unemployment rate * year nine
0.002
Health problem limits work
Unemployment rate (model 3)
0.005*
Individual unemployment
0.033***
Unemployment rate (model 4)
0.012***
Unemployment rate * year nine
-0.004
Four or more drinks on one occasion
Unemployment rate (model 3)
0.009
Individual unemployment
0.006
Unemployment rate (model 4)
0.010
Unemployment rate * year nine
-0.002
Drug use
Unemployment rate (model 3)
0.006
Individual unemployment
0.014
Unemployment rate (model 4)
-0.000
Unemployment rate * year nine
0.009**
Table 4.A4 Hispanic
White
-0.003
(0.010)
0.007
(0.007)
0.009
(0.006)
0.007
(0.008)
0.006
(0.006)
0.007
(0.010)
Source: Authors’ compilation based on the Fragile Families and Child Wellbeing Study.
Note: Standard errors and t-stats in parentheses. Model 1 includes level unemployment rate; results include individual fixed effects and time.
***p < .01; **p < .05; *p < .1
0.005 -0.006
(0.009) (0.009)
0.002
(0.007)
0.025* -0.022 -0.023**
(0.015) (0.014) (0.011)
0.010*
(0.005)
0.009
(0.006)
Married
0.009 -0.002
(0.006) (0.006)
White
0.014*** 0.010**
(0.005)
(0.005)
Hispanic
0.018* -0.003
(0.011) (0.009)
Black
0.006 -0.001
(0.009) (0.011)
Single
Fathers
0.030*** 0.018**
(0.008)
(0.008)
Married Cohabiting
Health status is fair or poor
Unemployment 0.023*** 0.019**
0.011 -0.003
rate
(0.008)
(0.009)
(0.010) (0.009)
Health problem limits work
Unemployment 0.013*** 0.015*** 0.004
0.003
rate
(0.005)
(0.005)
(0.006) (0.005)
Four or more drinks on one occasion
Unemployment 0.010*
-0.003
0.021* 0.009
rate
(0.006)
(0.009)
(0.011) (0.008)
Drug use
Unemployment 0.019*** 0.002
-0.008
0.001
rate
(0.006)
(0.006)
(0.008) (0.007)
Black
Mothers
Table 4.A5 Coefficients and Standard Errors, Model 1, All Outcomes
-0.003
(0.009)
0.011
(0.011)
0.007
(0.005)
-0.005
(0.009)
Cohabiting
0.016
(0.010)
0.000
(0.013)
0.002
(0.007)
0.013
(0.011)
Single
mothers’ and fathers’ health115
Our preferred specification uses a model that controls for individual
fixed effects. It exploits the longitudinal nature of the FFS to control for
observed and unobserved time-invariant characteristics of the mother and
father, which may be correlated with both their residence in a city that
has high unemployment rate and their health problems. It thus provides
more compelling evidence of the effect of unemployment on health than
the LPM. We estimate separate models by education groups and measure
education at the time of the child’s birth.
To predict the effects of the Great Recession, we estimate the predicted
probability when we set the unemployment rate to 5 percent and compare
those predictions with a 10 percent rate. We predict different probabilities
and levels for each level of parents’ education.
NOTES
1. Currie and Madrian 1999; Miilunpalo et al. 1997.
2. See, for example, Ruhm, 2000, 2003, 2005; Ruhm and Black 2002; Dehejia
and Lleras-Muney 2004.
3. Browning and Heinesen 2012; Burgard, Ailshire, and Kalousova 2013;
Charles and DeCicca 2008; Dee 2001; Eliason and Storrie 2009a, 2009b;
Ruhm 2000; Sullivan and Wachter 2009; Theodossiou 1997.
4. On binge drinking, Deb et al. 2011; Dee 2001; Dehejia and Lleras-Muney
2004; Xu and Kaestner 2010; on the unemployed, Charles and DeCicca 2008.
5. Fischer and Sousa-Poza 2009; Kim et al. 2008.
6. Geronimus 1992.
7. Acs and Nelson 2002; Ross and Van Willigen 1996.
8. Becker 1981; Lam 1988.
9. Currie, Duque, and Garfinkel 2015.
10.Ibid.
11. Grusky, Western, and Wimer 2011; Hoynes, Miller, and Schaller 2012.
REFERENCES
Acs, Gregory, and Sandi Nelson. 2002. “The Kids Are Alright? Children’s WellBeing and the Rise of Cohabitation.” Washington, D.C.: The Urban Institute.
Becker, Gary S. 1981. A Treatise on the Family. Cambridge, Mass.: Harvard
University Press.
Browning, Martin, and Eskil Heinesen. 2012. “Effect of Job Loss Due to Plant
Closure on Mortality and Hospitalization.” Journal of Health Economics 31(4):
599–616.
Burgard, Sarah A., Jennifer A. Ailshire, and Lucie Kalousova. 2013. “The Great
Recession and Health: People, Populations, and Disparities.” Annals of the
American Academy of Political and Social Science 650(1): 194–213.
116
children of the great recession
Charles, Kerwin K., and Philip DeCicca. 2008. “Local Labor Market Fluctuations
and Health: Is There a Connection and for Whom?” Journal of Health
Economics 27(6): 1532–50.
Currie, Janet, Valentina Duque, and Irwin Garfinkel. 2015. “The Great Recession
and Mother’s Health.” Economic Journal 125(588): F311–46.
Currie, Janet, and Brigitte Madrian. 1999. “Health, Health Insurance and the
Labor Market.” In The Handbook of Labor Economics, vol. 3c, edited by David
Card and Orley Ashenfelter. Amsterdam: North-Holland.
Deb, Partha, William T. Gallo, Padmaja Ayyagari, Jason M. Fletcher, and Jody
L. Sindelar. 2011. “The Effect of Job Loss on Overweight and Drinking.”
Journal of Health Economics 30(2): 317–27.
Dee, Thomas S. 2001. “Alcohol Abuse and Economic Conditions: Evidence
from Repeated Cross-Sections of Individual-Level Data.” Health Economics
10(3): 257–70.
Dehejia, Rajeev, and Adriana Lleras-Muney. 2004. “Booms, Busts, and Babies’
Health.” Quarterly Journal of Economics 119(3): 1091–30.
Eliason, Marcus, and Donald Storrie. 2009a. “Does Job Loss Shorten Life?”
Journal of Human Resources 44(2): 277–302.
———. 2009b. “Job Loss Is Bad for Your Health: Swedish Evidence on CauseSpecific Hospitalization Following Involuntary Job Loss.” Social Science and
Medicine 68(8): 1396–406.
Fischer, Justina A., and Alfonso Sousa-Poza. 2009. “Does Job Satisfaction
Improve the Health of Workers? New Evidence Using Panel Data and Objective
Measures of Health.” Health Economics 18(1): 71–89.
Geronimus, Arline T. 1992. “The Weathering Hypothesis and the Health of
African-American Women and Infants: Evidence and Speculations.” Ethnicity
& Disease 2(3): 207–21.
Grusky, David B., Bruce Western, and Christopher Wimer. 2011. The Great
Recession. New York: Russell Sage Foundation.
Hoynes, Hilary W., Douglas L. Miller, and Jessamyn Schaller. 2012. “Who Suffers
During Recessions?” 26(3): 27–48.
Kim, Myoung-Hee, Chang-yup Kim, Jin-Kyung Park, and Ichiro Kawachi.
2008. “Is Precarious Employment Damaging to Self- Rated Health? Results
of Propensity Score Matching Methods, Using Longitudinal Data in South
Korea.” Social Science & Medicine 67(12): 1982–94.
Lam, David. 1988. “Marriage Markets and Assortative Mating with Household
Public Goods: Theoretical Results and Empirical Implications.” Journal of
Human Resources 23(4): 462–87.
Miilunpalo, Seppo, Ilkka Vuori, Pekka Oja, Matti Pasanen, and Helka Urponen.
1997. “Self-Rated Health Status as a Health Measure: The Predictive Value of
Self-Reported Health Status on the Use of Physician Services and on Mortality in
the Working-age Population.” Journal of Clinical Epidemiology 50(5): 517–28.
Ross, Catherine E., and Marieke Van Willigen. 1996. “Gender, Parenthood, and
Anger.” Journal of Marriage and Family 58(3): 572–84.
Ruhm, Christopher J. 2000. “Are Recessions Good for Your Health?” Quarterly
Journal of Economics 115(2): 617–50.
———. 2003. “Good Times Make You Sick.” Journal of Health Economics 22(4):
637–58.
———. 2005. “Healthy Living in Hard Times.” Journal of Health Economics
24(2): 341–63.
mothers’ and fathers’ health117
Ruhm, Christopher J., and William E. Black. 2002. “Does Drinking Really
Decrease in Bad Times?” Journal of Health Economics 21(4): 659–78.
Sullivan, Daniel, and Till von Wachter. 2009. “Job Displacement and Mortality:
An Analysis using Administrative Data.” Quarterly Journal of Economics
124(3): 1265–306.
Theodossiou, Ioannis. 1997. “The Effects of Low-Pay and Unemployment on
Psychological Well-Being: A Logistic Regression Approach.” Journal of Health
Economics 17(1): 85–104.
Xu, Xin, and Robert Kaestner. 2010. “The Business Cycle and Health Behaviors.”
NBER working paper no. 15737. Cambridge, Mass.: National Bureau of
Economic Research.
Chapter 5
Parents’ Relationships
Daniel Schneider, Sara McLanahan,
and Kristen Harknett
I
n previous chapters, we saw that the Great Recession generated a good
deal of economic upheaval in the lives of families with young children.
Transfers from the government and family members helped stem some
suffering but did not fully make up for the recession’s economic effects.
These economic effects, in turn, are likely to have spilled over into other
areas of family life. This topic is the focus of the rest of the volume. One
of the ways economic upheaval can affect families is by generating family stress, which may in turn destabilize some relationships and lower the
quality of those that remain intact. In this chapter, we examine both outcomes. We focus on two domains: relationship status—whether the mother
is living alone or with a partner—and relationship quality—how supportive mothers and their partners are of one another as well as the overall
quality of their relationship. These domains are critical to understanding
family and child well-being, given the wealth of research documenting
the importance of stable, supportive, high-quality parent relationships on
children’s well-being and eventual life chances.
Specifically, this chapter focuses on three questions: Did high levels of
unemployment during the Great Recession reduce the likelihood that a
mother was married or living with a partner? Did high levels of unemployment during the Great Recession reduce the quality of parental relationships?
If so, did these effects differ by mothers’ education? Our goal throughout
is to understand whether the high levels of unemployment generated by
the Great Recession spilled over to affect the relationships between parents
of young children—one mechanism through which poor macroeconomic
conditions might eventually compromise children’s development.
Like the previous chapters, we begin by describing trajectories of parents’
relationship status and quality over the nine-year follow-up period. Our
goal here is primarily descriptive—to establish whether parents’ relationships change much over time and how these patterns may differ by social
class background. For relationship status, we examine variation over time
and across education groups in whether a mother is living with a partner
(child’s biological father or a new partner) or no partner. For relationship
quality, we examine the relationship between either biological mothers and
parents’ relationships119
fathers or mothers and their new partners. We then estimate the effects of
the Great Recession on relationship status and quality. We find that the
recession led to modest declines in two-parent families, and some declines
in relationship supportiveness and the overall quality of mother-father relationships. These declines are most pronounced among families in which
the mother has less than a college education.
RECESSIONS AND ROMANTIC RELATIONSHIPS
A large body of work dating back to the turn of the twentieth century shows
that more people get married when macroeconomic conditions are favorable.1 Studies spanning the 1970s, 1980s, and 1990s find that unfavorable
economic conditions lower rates of marriage, whereas favorable conditions
raise marriage rates.2 Why would this be the case? In general, people are likely
to feel more secure entering into a lasting commitment such as marriage when
they feel secure about their economic fortunes. Marriage can also be costly,
making such unions more likely when families’ budgets are not strained.
A few studies examine the effects of more recent economic downturns,
including the Great Recession, and tend to support the idea that negative
macroeconomic conditions suppress the likelihood that couples marry.3
These studies, however, tend to average results for parents and nonparents,
making them less useful for assessing effects on children. In this chapter,
we focus on parents with children, which allows us to assess how the Great
Recession affected the living arrangements and relationship contexts in
which children are raised.
Of course, poor macroeconomic conditions may cause couples to
end relationships as much as they dissuade couples from entering them.
A second set of studies thus examines the association between macro­
economic conditions and divorce. Here the evidence is more mixed, reflecting the offsetting theoretical effects recessions have.4 On the one hand,
job loss and economic hardship are expected to create financial strain and
marital conflict, which should increase the breakup of existing relationships;
on the other hand, economic hardship makes it more difficult for couples
to afford the legal fees associated with divorce and the costs of establishing
separate households, which should work in the opposite direction. In the
late 1800s and early 1900s, divorce rates were lower during recessions, suggesting that the costs of divorce outweighed the stress associated with financial hardship.5 In the post–World War II period, divorce rates have been
higher during hard times, a phenomenon attributed to the declining costs
of divorce and the increasing generosity of welfare state benefits.6 However,
the most recent studies of unemployment and divorce tend to find that
higher unemployment is associated with a decline (or at least a delay) in
divorce.7 For example, a recent study using census data finds the expected
negative effect of state-level unemployment on the divorce rate.8 Another
120
children of the great recession
study finds no association between divorce and state-level economic conditions, but does suggest a reduction in divorce during the Great Recession
relative to before the recession.9 As with studies of marriage, the studies
of the macro economy and divorce typically combine parents and non­
parents and, with one exception, average results across several decades.10
Thus, these studies do not tell us how the Great Recession affected parental
relationships and children’s family settings.
A third set of studies focuses on how economic conditions impact marital and relationship quality. Studies dating to the Great Depression show
that job loss lowers marital quality.11 The family stress model, which is
based on studies of the Great Depression and the 1980s farm crisis, argues
that economic crises lead to reductions in marital quality by increasing perceived financial strain, depression, and hostility and reducing warmth and
emotional supportiveness.12 Other studies show that economic strain is
associated with decreases in partner supportiveness and increase in intimate
partner violence.13 In addition to increasing financial strain and depression (as described in the family stress model), poor macroeconomic conditions may also reduce marital quality by undermining men’s economic role
in the family in the family. Both Shirley Hatchett and her colleagues and
Richard Patterson attribute conflict and distrust among African American
couples to black men’s attempt to seize authority to compensate for their
weak economic position in the family.14 A similar pattern was observed
among white families during the Great Depression.15
In sum, both theory and previous research give us reason to expect that
the economic shock of the Great Recession may have affected parents’ relationship status and quality. In the case of relationship status, the net effects
are ambiguous because the recession could have reduced new partnerships,
but either destabilized other couples or inhibited them from separating
because of the cost of divorce. In the case of relationship quality, we expect
to find deterioration during economic upheaval. Our analysis weighs in on
these general questions, focusing on adults with children and the household settings in which children are raised. We are also attentive to how the
effects of macroeconomic conditions vary across education groups and
thus contribute to class stratification.
TRENDS IN RELATIONSHIP STATUS AND QUALITY
We present information about the trends in relationship status and
quality in the nine-year follow-up period over which Fragile Families
parents were interviewed. The initial waves of the survey took place in
the early 2000s, and the year nine wave often coincided with the Great
Recession. The descriptive patterns we present in figures 5.1 through
5.5 are useful for providing a broad backdrop before turning to whether
the Great Recession led to changes in relationship status and quality.
parents’ relationships121
Percent of Mothers
Figure 5.1 Mothers’ Relationship Status
100
90
80
70
60
50
40
30
20
10
0
1
3
5
(1999–2001) (2001–2003) (2003–2006)
9
(2007–2010)
Child’s Age-Year
Married or cohabiting bio father
Married or cohabiting new partner
No coresidential romantic
Source: Authors’ calculations based on Fragile Families and Child Wellbeing Study data.
Figure 5.1 shows the distribution of mothers across three types of relationship status. About 77 percent were married to or cohabiting with
their child’s biological father one year after the child’s birth, declining to
56 percent by the time the child was nine years old. Two percent were
living with a new partner one year after the birth of the child, increasing to
about 16 percent by the time the child was nine years old. The proportion
who were single—that is, not in a coresidential romantic relationship—
increases from 21 percent at year one to 28 percent at year nine.
The next two figures (figures 5.2 and 5.3) show mothers’ relationship
status over time by her educational background. Figure 5.2 shows the
share of mothers married to either the child’s biological father or a new
partner at each year. The marriage gap across education groups is large:
college-educated mothers are the most likely to be married one year after
their child’s birth (about 97 percent), whereas mothers with less than a
high school diploma are the least likely to be married (about 35 percent).
These gaps persist over the next eight years. College-educated mothers
show somewhat larger declines in marriage and the least-educated mothers
show slight increases. Nevertheless, the marriage gap between collegeeducated and less-educated mothers remains substantial by the time their
children are nine years old.
Figure 5.3 shows the share of mothers married or cohabiting with either
the biological father of the focal child or a new partner. As before, education differences in the share of mothers living with a partner during the first
nine years of their children’s lives are stark and persistent. College-educated
122
children of the great recession
Figure 5.2 Marriage to Bio Fathers or New Partners
100
90
Percent of Mothers
80
70
60
50
40
30
20
10
0
1
3
5
(1999–2001) (2001–2003) (2003–2006)
9
(2007–2010)
Child’s Age-Year
College +
Some college
High school
Less than high school
Source: Authors’ calculations based on Fragile Families and Child Wellbeing Study data.
Figure 5.3 Marriage or Cohabitation to Bio Fathers or New Partners
100
90
Percent of Mothers
80
70
60
50
40
30
20
10
0
1
3
5
(1999–2001) (2001–2003) (2003–2006)
9
(2007–2010)
Child’s Age-Year
College +
Some college
High school
Less than high school
Source: Authors’ calculations based on Fragile Families and Child Wellbeing Study data.
parents’ relationships123
Figure 5.4 Mothers’ Reports of Bio Fathers’ Supportiveness
Relationship Supportiveness Score
2.0
1.8
1.6
1.4
1.2
1.0
0.8
0.6
0.4
0.2
0
1
3
5
(1999–2001) (2001–2003) (2003–2006)
9
(2007–2010)
Child’s Age-Year
College +
Some college
High school
Less than high school
Source: Authors’ calculations based on Fragile Families and Child Wellbeing Study data.
mothers are nearly all living with partners one year after their child’s birth,
though a downward trend over time is evident. Mothers with less education
are less likely to be married or cohabiting one year after the birth, and also
experience a slight downward trajectory. Significantly, although mothers
with some college education are somewhat more likely to be married or
cohabiting than their less-educated counterparts, there is very little difference in the experience of mothers without a high school diploma and those
with only a diploma.
The next set of figures examines parents’ reports of the supportiveness of their spouse or partner, as well as overall relationship quality.
Supportiveness is estimated based on each parent’s reports of a partner’s
behavior with regard to six domains: fairness and willingness to compromise; expression of affection or love; insults and criticism (reverse coded);
encouragement and helpfulness; listening when partner needs someone to
talk to; and perceptions that the other really understands one’s hurts and
joys. Quality is measured by asking parents to rate the overall quality of
their relationship, which ranges from poor to excellent. This second question is asked irrespective of whether parents live together or apart (for the
wording of the questions, see the appendix).
Figure 5.4 plots trajectories for mothers’ reports of biological fathers’
supportiveness. This figure is based on biological parents who are living
124
children of the great recession
Figure 5.5 Fathers’ Reports of Bio Mothers’ Supportiveness
Relationship Supportiveness Score
2.0
1.8
1.6
1.4
1.2
1.0
0.8
0.6
0.4
0.2
0
1
3
5
(1999–2001) (2001–2003) (2003–2006)
9
(2007–2010)
Child’s Age-Year
College +
Some college
High school
Less than high school
Source: Authors’ calculations based on Fragile Families and Child Wellbeing Study data.
together and makes it clear that mothers’ perceptions of fathers’ supportiveness are quite positive and remain steady over time. Overall, and in
comparison with the results for relationship status, differences in reports
of supportiveness across education groups are small. This is also true when
we look at fathers’ reports of mothers’ supportiveness (figure 5.5), and at
mothers’ reports of a new partner’s supportiveness (figure 5.6).
Figure 5.7, which presents mother’s assessments of relationship quality with their child’s biological father for coresident couples, shows that
more-educated mothers report higher quality relationships than lesseducated mothers. No pronounced trend is evident, but overall relationship quality declined slightly between when children were one and nine
years old for all education groups. Most of the decline for those with less
than a college education occurred between when the child was ages one
and three. Patterns were similar for biological fathers’ reports about
overall relationship quality with the mother (figure 5.8).
In short, college-educated parents are much more likely to be married and report slightly higher quality relationships with their partners
than less-educated parents. In the next section, we turn to the question
of how the Great Recession affected marriage, cohabitation, and relationship quality.
Figure 5.6 Mothers’ Reports of New Partners’ Supportiveness
Relationship Supportiveness Score
2.0
1.8
1.6
1.4
1.2
1.0
0.8
0.6
0.4
0.2
0
1
3
5
(1999–2001) (2001–2003) (2003–2006)
9
(2007–2010)
Child’s Age-Year
College +
Some college
High school
Less than high school
Source: Authors’ calculations based on Fragile Families and Child Wellbeing Study data.
Figure 5.7 Mothers’ Reports of Relationship with Bio Father
Overall Relationship Quality Score
4.0
3.5
3.0
2.5
2.0
1.5
1.0
0.5
0
1
3
5
(1999–2001) (2001–2003) (2003–2006)
9
(2007–2010)
Child’s Age-Year
College +
Some college
High school
Less than high school
Source: Authors’ calculations based on Fragile Families and Child Wellbeing Study data.
126
children of the great recession
Figure 5.8 Fathers’ Reports of Relationship with Bio Mother
Overall Relationship Quality Score
4.0
3.5
3.0
2.5
2.0
1.5
1.0
0.5
0
1
3
5
(1999–2001) (2001–2003) (2003–2006)
9
(2007–2010)
Child’s Age-Year
College +
Some college
High school
Less than high school
Source: Authors’ calculations based on Fragile Families and Child Wellbeing Study data.
EFFECTS OF THE GREAT RECESSION ON RELATIONSHIP
STATUS AND QUALITY
We follow the approach of the previous chapters and examine the relationship between area-level unemployment rates (averaged over the year prior
to interview) and parents’ relationship status and quality. As in the previous chapters, we use our model to predict relationship status and quality,
given unemployment rates of 5 percent and 10 percent. We treat the difference in predicted values as the “effect of large recessions.”
Relationship Status
We begin by looking at the effect of large recessions on mothers’ relationship status. Figure 5.9 examines changes in the probability that mothers
are married (left two columns) or married or cohabiting (right two columns) assuming unemployment rates of 5 percent and 10 percent. These
estimates are derived from regression models described in the appendix.
The full regression estimates for mothers married or cohabiting with father
or new partner are presented in appendix table 5.A1. We do not distinguish
between mothers’ relationships with biological father and new partner in
parents’ relationships127
Figure 5.9 Mothers’ Marriage and Marriage or Cohabitation
70
−7%
Percent of Mothers
60
50
40
−5%
UR 5 percent
30
UR 10 percent
20
10
0
Married
Married or Cohab†
Source: Authors’ calculations based on Fragile Families and Child Wellbeing Study data.
†
p < .1
figure 5.9, but supplementary analyses (not shown here) indicate that doing
so does not alter the results. Considering only marriage (left two bars),
39 percent of mothers are predicted to be married when unemployment is
relatively low, whereas 37 percent are predicted to be married when unemployment is twice as high, representing a 5 percent decrease in marriage.
The difference is not statistically significant. In contrast, when we include
cohabiting unions, the proportional gap is larger (61.3 percent at 5 percent
unemployment and 57.5 percent at 10 percent unemployment), and the
7 percent difference in statistically significant. These results suggest that
poor economic conditions do reduce coresidential partnerships. Whether
this difference is a result of fewer new partnerships, more breakups among
existing partnerships, or both is a question we return to later in the chapter.
Next we consider whether these effects are broadly shared across families with different class backgrounds. Figure 5.10 shows the effects of
large recessions on the share of women in a marital relationship by education. Here, we see that mothers with some postsecondary education but
no college degree are less likely to be married when unemployment rates
are high. In contrast, large recessions have no effect on marital status for
mothers with a college degree or those with a high school diploma or less.
Interestingly, although the differences between groups are not statistically
significant, the negative effects of high rates of unemployment are most
pronounced among mothers with some postsecondary education but no
college degree. This finding is similar to results reported for several of the
economic outcomes in the previous chapters, suggesting these families
may be particularly compromised by a big recession.
Figure 5.11 shows similar predictions for the share of women in a married
or cohabiting union. For all education groups combined, recessions reduce
128
children of the great recession
Figure 5.10 Mothers’ Marriage (Bio Father or New Partner)
90
+1%
Percent of Mothers
80
70
60
−17%
50
40
30
+1%
−7%
Less than
high school
High
school
UR 5 percent
UR 10 percent
20
10
0
Some
college†
College +
Source: Authors’ calculations based on Fragile Families and Child Wellbeing Study data.
Note: No significant differences in effect of unemployment between subgroups.
†
p < .1
coresidential unions, and the effect is significant. Women without a college
education are less likely to be in a coresidential union when unemployment rates are high than when they are low. Once again, college-educated
women are actually more likely to be in a union when unemployment rates
are high, but the effect is not statistically significantly different from zero.
Notice that the negative effect of high unemployment on the status of
less-educated mothers (high school diploma or less) is more pronounced
Percent of Mothers
Figure 5.11 100
90
80
70
60
50
40
30
20
10
0
Mothers’ Marriage or Cohabitation (Bio Father or New Partner)
+4%
−7%
−13%
−7%
UR 5 percent
UR 10 percent
Less than
high school
High
school
Some
college
College +
Source: Authors’ calculations based on Fragile Families and Child Wellbeing Study data.
Note: No significant differences in effect of unemployment between subgroups.
parents’ relationships129
when we consider cohabitation in addition to marriage. This difference
reflects the fact that cohabiting unions are much more common among
less-educated mothers, and these groups are apparently more sensitive
to variation in economic conditions. This finding is also true for the economic well-being outcomes reported in chapter 2.
These analyses tell us that high levels of unemployment reduce marriage
and cohabitation among women with less than a college degree. However,
they do not tell us whether the differences in relationship status are due to
increases in a mother’s chances of ending a relationship with the biological
father or decreases in her chances of entering a relationship with the father
or a new partner. To further investigate these processes, we estimate a set
of models on the effects of unemployment on the probability that a mother
would end her relationship with the biological father. We also estimate
models that look at the effects of unemployment on the probability that a
mother would enter a relationship with the child’s father or a new partner.
To examine dissolution, we focus on mothers who were living with the biological father at the time of the previous interview. To estimate entrances,
we focus on mothers not living with a father or new partner at the time of
the previous interview. The supplementary analyses suggest that the effects
of major recession on changes in relationship status are driven by a combination of small increases in dissolution and small decreases in relationship
formation during bad economic times. Although none of these estimates
are statistically distinguishable from zero, they suggest that the net results
are driven by two distinct forces.
Relationship Quality
In the next set of analyses, we look at the effects of large recessions on
relationship quality among parents. We begin by looking at biological
parents’ reports about how their coresidential partners treat them. These
analyses are restricted to biological parents who are living together, either
married or cohabiting.
According to mothers, the typical biological father is very supportive.
When we look at all mothers combined, fathers’ supportive behavior is not
particularly sensitive to increases in unemployment rates (figure 5.12). The
overall null result masks some underlying differences across education
groups. Mothers with a high school diploma report declines in fathers’
supportiveness as a result of large recessions, whereas mothers with some
postsecondary education and mothers with a college degree report slight
increases in supportiveness. Recall that the supportiveness scale ranges
from 0 (never) to 2 (often) based on a set of six supportive behaviors, and
that the average mother reports a value of around 1.6 on the scale. This set
of figures display a truncated scale to aid in visualizing differences across
groups and economic conditions.
130
children of the great recession
Relationship Supportiveness
Score
Figure 5.12 Mothers’ Reports of Bio Fathers’ Supportiveness
1.75
+1%
1.70
1.65
1.60
−5%
0%
1.55
+3%
UR 5 percent
0%
UR 10 percent
1.50
1.45
1.40
All
Less than
High
high school school*
Some
college
College +
Source: Authors’ calculations based on Fragile Families and Child Wellbeing Study data.
Note: No significant differences in effect of unemployment between subgroups.
*p < .05
As shown in figure 5.13, fathers also report high levels of support from
mothers, though here we see a more general decline in supportiveness
when unemployment rates are high (10 percent). For fathers, recessionary
conditions increase inequality in supportiveness across education groups.
Men living with mothers who have less than a college education see a drop
in supportiveness, and their counterparts see an increase.16
Figure 5.14 shows the effects of recessions on mothers’ reports of support from new partners. Overall, supportiveness from new partners is
Relationship Supportiveness
Score
Figure 5.13 Fathers’ Reports of Mothers’ Supportiveness
1.75
1.70
1.65
+5%
−2%
−8%
−4%
−3%
UR 5 percent
1.60
UR 10 percent
1.55
1.50
1.45
All*
Less than
High
high school** school
Some
college
College +†
Source: Authors’ calculations based on Fragile Families and Child Wellbeing Study data.
Note: Chow tests shows that the coefficient for unemployment for college is different
from the coefficient for unemployment for the less than high school group.
**p < .01; *p < .05; †p < .1
parents’ relationships131
Relationship Supportiveness
Score
Figure 5.14 Mothers’ Reports of New Partners’ Supportiveness
2.5
2.0
+2%
+2%
+6%
+7%
−40%
1.5
UR 5 percent
1.0
UR 10 percent
0.5
0
All
Less than
high school
High
school
Some
college
College +†
Source: Authors’ calculations based on Fragile Families and Child Wellbeing Study data.
Note: Chow tests show that the coefficient for unemployment for college is different than
the coefficient for unemployment for the less than high school group.
†
p < .1
high, even slightly higher than reports of supportiveness from fathers.
Looking at mothers as a whole, we find that high unemployment is associated with small increases in partners’ supportiveness. When we look at
the evidence by education, we see that the increase in partners’ supportiveness is concentrated among mothers with less than a college degree.
For mothers with a college degree, major increases in unemployment
are associated with large declines in new partners’ supportive behavior.
The decline in support from new partners reported by college-educated
mothers is the only case in which families with a college-educated mother
appear to be more negatively affected than other mothers by poor economic conditions. In all other analyses, these families report stability
or improvement in their relationships under recessionary conditions,
but their less-educated counterparts report modest declines. However,
this result is based on an extremely small sample size—just sixty-three
observations.
Figures 5.15 and 5.16 focus on coresident biological parents’ assessments of overall relationship quality. Looking first at mothers’ assessments
(figure 5.15), we find that large recessions lead to small increases in relationship quality among mothers in all education groups.
Looking next at fathers’ assessments of the overall quality of their
relationship with their child’s mother (figure 5.16), fathers report lower
relationship quality when unemployment rates are high, with one exception: if the mother has a college degree, fathers report higher relationship
quality. We also reestimate these relationships, broadening our focus to
include mothers’ reports of all fathers of focal children and fathers’ reports
132
children of the great recession
Overall Relationship
Quality Score
Figure 5.15 Mothers’ Reports of Quality of Relationship with
Bio Father
4.0
3.5
3.0
2.5
2.0
1.5
1.0
0.5
0
+2%
+1%
+3%
+1%
+4%
UR 5 percent
UR 10 percent
All
Less than
high school
High
school
Some
college
College +
Source: Authors’ calculations based on Fragile Families and Child Wellbeing Study data.
Note: No significant differences in the effect of unemployment between subgroups.
of all mothers of focal children, whether or not they were romantically
coresident. We find substantively similar results.
These analyses of relationship quality assume that unemployment rates
had the same type of effect on relationship quality in the early 2000s
they had during the Great Recession. In separate analyses, we relax this
assumption (see table 5.A3). We look instead at whether the effects of
unemployment were more pronounced during periods of unusually high
unemployment, such as occurred during the Great Recession. We find
that unemployment rates characteristic of the Great Recession led to
Overall Relationship Quality
Score
Figure 5.16 4.0
3.5
3.0
2.5
2.0
1.5
1.0
0.5
0
Fathers’ Reports of Quality of Relationship with Bio Mother
−2%
−1%
−4%
−6%
0%
UR 5 percent
UR 10 percent
All
Less than
high school
High
school
Some
college*
College +
Source: Authors’ calculations based on Fragile Families and Child Wellbeing Study data.
Note: Chow tests show that the coefficient for unemployment for college is different than
the coefficient for unemployment for the less than high school group.
*p < .05
parents’ relationships133
larger declines in mothers’ reports of fathers’ supportiveness and fathers’
reports of the quality of his relationship with the mother. In contrast, the
effect of unemployment on marriage and cohabitation was weaker during
the last part of the decade.
How quickly area-level unemployment rates are deteriorating (or improving) may better capture the sense of economic unease or uncertainty felt by
households than the prevailing level of unemployment. To test this idea,
spline models distinguish between percentage decline in annual unemployment rates and percentage increase in annual rates (table 5.A2). For relationship status and quality, we observe few significant effects of rapidly changing
rates. The only exception is fathers’ reports about their relationship with
their child’s mother. For this outcome, fathers in the two lowest education groups—those with less than a high school diploma and those with a
high school diploma only—report additional declines in relationship quality. Although deteriorating economic conditions did not have much effect
on relationship supportiveness and overall relationship quality, the recession
may nevertheless have increased undesirable relationship behaviors such as
being violent or controlling. Other research using the Fragile Families study
finds just that—that an increase in the unemployment rate is associated with
increases in men’s controlling behavior toward their female partners.17
SUMMARY AND CONCLUSION
A large body of research dating to the turn of the twentieth century shows
that marriage rates are positively associated with favorable macroeconomic
conditions. Further, studies of the Great Depression indicate that job loss
lowered marital quality by increasing financial strain and reducing warmth
and emotional supportiveness.18 We contribute to this body of research
by examining the effects of unemployment on the status and quality of
parental relationships during the Great Recession.
Focusing first on relationship status, we find that high rates of unemploy­
ment reduce marriage and cohabitation among mothers with less than a
college degree. In contrast, for college-educated mothers, the chances of
marriage are not affected by high unemployment. Indeed, college-educated
women are slightly more likely to be living with a partner (married or cohabiting) in difficult economic times. Our analyses of relationship status reveals
a wide marriage gap between mothers with a college education and their
counterparts with less education. Mothers with a college degree are far
more likely to be married to or living with a partner than their less-educated
counterparts in good or bad economic times. Although the relationship
changes brought about by the Great Recession were modest, the recession
widened already large marriage gaps between families with college-educated
mothers and those with less-educated mothers.
We find some evidence that unemployment rates on the order of magnitude of those during the Great Recession reduce relationship quality
134
children of the great recession
for select social class groups. Biological fathers, for example, are likely
to have less support from mothers and to see the overall quality of their
relationship with the mother of their child decline during periods of high
unemployment. If the mother has a college degree, however, she offers
more support during hard times.
A large literature in recent years argues that economic stability is a prerequisite for stable marriages and that economic distress has a destabilizing
effect. Evidence from the Great Depression suggests that the period had
major repercussions for couples and families whose incomes dropped precipitously. Given the magnitude of the shock of the Great Recession, we
might have expected to observe sizable increases in relationship instability
or relationship distress and conflict. Instead, our estimates suggest modest
negative effects on relationship status overall, and somewhat larger negative
effects on relationship status for mothers in the middle education groups.
On balance, the Great Recession tended to destabilize relationships or forestall relationship entry more so than it forced couples to stay together. It also
tended to lower the quality of relationships slightly.
APPENDIX
Measures
We examine two measures of mother’s romantic relationship status. First
we construct a measure of whether the mother is married to either the
focal child’s father or a new partner at the time of the interview. Second
we construct a measure of whether the mother is either married to or
cohabiting with the focal child’s father or a new partner at the time of
the interview.
Our relationship supportiveness measure is based on six items: partner is fair
and willing to compromise when you have a disagreement, expresses affection or love for you, insults or criticizes you or your ideas (reverse coded),
encourages or helps you to do things that are important to you, listens to you
when you need someone to talk to, and really understands your hurts and
joys. Response categories were 0 = never, 1 = sometimes, and 2 = often. We
sum these measures and divide by the number of items answered to construct
our measure, so the resulting scale ranges from 0 = never for all six items to
2 = often for all six items. We first examine reports of supportiveness from
mothers, who report on fathers with whom they are currently in romantic
coresidential relationships or on new partners with whom they are currently
in romantic coresidential relationships. We then examine reports of supportiveness from fathers, who report on mothers with whom they are currently
in romantic coresidential relationships.
We construct a measure of overall relationship quality based on mothers’
and fathers’ report of their relationship with the focal child’s other parent
parents’ relationships135
on a 5-point scale. Relationships with a score of 0 are poor and those with
a score of 4 are excellent. We limit our analysis of this measure to biological
parents who are currently in coresidential romantic relationships.
Key Independent Variable
For each analysis, the unemployment rate is a measure of the average
unemployment rate in the sample city over the twelve months before
the interview. This is calculated to match the period preceding the outcome measures.
Key Moderating Variables
We study differences in the trajectories over time, and in the effects of the
Great Recession, on relationship status and quality stratified by maternal
education at baseline. Mother’s education is coded as less than a high school
diploma or the completion of a GED, a high school diploma, some college
or an associate’s degree or technical degree, or a bachelor’s degree or greater.
Control Variables
We include a number of covariates in our models, all measured at the first
survey wave (baseline). These include mother’s age at the birth, immigrant
status (foreign born), number of children in the household, a measure of
whether the mother was living with both biological parents at age fifteen,
as well as city (twenty dummies for each sample city) and survey year fixed
effects (twelve calendar year dummies). In analyses of relationship supportiveness and overall relationship quality, we control for whether parents were
married at the time the focal child was born.
Method
The figures that plot the trajectories of each outcome measure over time
present the mean levels of each outcome at each survey wave. All means
are weighted with the wave-specific city-weights to be representative of
births in the twenty study cities; the sample is restricted to mothers who
are interviewed in all survey waves.
To study the effects of the Great Recession, we conduct logistic regressions for binary outcomes and ordinary least squares regression analyses
using the pooled data (waves 2 through 5). The standard errors are clustered at both the city and individual level to account for within city and
within person clustering–nonindependence. Analyses are conducted for
all mothers and separately for mothers with less than high school, high
school only, some college, or college degree or greater. We estimated
pooled models and also a parallel set of models with mother fixed effects.
136
children of the great recession
To predict the effects of the Great Recession, we estimate the predicted
probability (for binary outcomes) or the predicted level (for the continuous variables) when the unemployment rate is set at 5 percent, a rate
typical of the period before the recession, and compare these predictions
with when the unemployment rate is set to 10 percent, a rate typical of
the Great Recession. We predict different probabilities for each level of
mother’s education.
Supplemental Analyses
We conduct a number of additional analyses to test the association between
the unemployment rate and parents’ relationships. First, to test whether the
speed of change in the unemployment rate was related to our outcomes, we
run spline models to distinguish between the percentage decline in annual
unemployment rate and the percentage increase in annual unemployment
rates (table 5.A2). For relationship status, we observe few significant effects
of rapidly changing rates, and the negative effects of unemployment levels on
status remained largely unchanged and significant. We also find little evidence
that rapidly worsening unemployment rates affected mother’s perceptions of
supportiveness of either fathers or new partners. In contrast, rapidly worsening rates lowered fathers’ reports of the quality of his relationship with child’s
mothers but only when mothers had a high school degree or less education.
Second, we estimate a set of models that include individual-level measures of mother’s and partner’s employment status (table 5.A3). In general, we find few significant effects in the models with mother fixed effects.
The exception is that fathers report better overall relationship quality with
coresident mothers when she is unemployed, but worse quality when he is
not working.
Third, we run analyses that include an interaction term with the unemployment rate and the year nine wave of data collection to test whether the
association between unemployment and the outcomes of interest differed
during the Great Recession (table 5.A3). We find three significant interactions: one for relationship status and two for relationship quality. For relationship status, the effects of unemployment were less negative during the
Great Recession; for mother reports of father supportiveness and fathers’
assessment of overall relationship quality, however, the effects were more
negative during the Great Recession.
Finally, we estimate our preferred model stratified by race-ethnicity and
by marital status at birth rather than by education (table 5.A4). Few patterns in these results are consistent. One interesting exception is father’s
reports of mother’s supportiveness and overall quality of relationship with
mother. For those outcomes, we find significant negative subgroup effects
for men in romantic coresidential relationships with Hispanic mothers and
men who were cohabiting at the birth of the focal child.
parents’ relationships137
Table 5.A1 Full Regression Results, Married to or Cohabiting
with Father or New Partner
With Individual
Fixed Effects
Unemployment rate
Education
Less than high school
High school
Some college
Mother’s age
Race-ethnicity
Black
Hispanic
Other
Immigrant
Children in household
Lived with both parents at age fifteen
Interview year
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
-0.056† Without Individual
Fixed Effects
(0.034)
-0.043*
(0.019)
—
—
—
—
(.)
(.)
(.)
(.)
-1.188***
-1.079***
-0.998***
0.006
(0.133)
(0.108)
(0.116)
(0.004)
—
—
—
—
—
—
(.)
(.)
(.)
(.)
(.)
(.)
-1.127***
-0.502***
-0.675***
0.732***
0.016
0.241***
(0.070)
(0.084)
(0.144)
(0.147)
(0.017)
(0.059)
(0.210)
(0.156)
(0.194)
(0.188)
(0.195)
(0.186)
(0.539)
(0.245)
(0.192)
(0.225)
(0.366)
0.379*
0.020
0.101
-0.067
0.053
-0.190
-0.478
-0.447***
-0.034
-0.099
-0.583*
(0.154)
(0.143)
(0.170)
(0.130)
(0.166)
(0.139)
(0.296)
(0.103)
(0.157)
(0.157)
(0.286)
0.327
0.025
0.009
-0.095
-0.041
-0.303
-0.490
-0.802**
-0.206
-0.144
-0.304
(Table continues on p. 138.)
138
children of the great recession
Table 5.A1 Continued
With Individual
Fixed Effects
City
Austin
Baltimore
Detroit
Newark
Philadelphia
Richmond
Corpus Christi
Indianapolis
Milwaukee
New York
San Jose
Boston
Nashville
Chicago
Jacksonville
Toledo
San Antonio
Pittsburgh
Norfolk
Constant
Observations
Number of individuals
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
(.)
(.)
(.)
(.)
(.)
(.)
(.)
(.)
(.)
(.)
(.)
(.)
(.)
(.)
(.)
(.)
(.)
(.)
(.)
7,187
1,951
Without Individual
Fixed Effects
-0.143*** (0.030)
-0.008
(0.100)
-0.120
(0.094)
-0.000
(0.081)
-0.055
(0.096)
-0.263*
(0.106)
0.183*
(0.081)
-0.123
(0.082)
0.049
(0.078)
0.085
(0.062)
0.085
(0.066)
-0.356*** (0.072)
-0.079
(0.081)
0.356*** (0.079)
0.127
(0.082)
0.008
(0.088)
0.111
(0.071)
-0.373*** (0.086)
0.118
(0.085)
2.200*** (0.201)
15,855
4,603
Source: Authors’ calculations based on Fragile Families and Child Wellbeing Study data.
Note: Z-stats in parentheses. Covariates are measured at the baseline survey (except year) and are
clustered at city and individual level. Model 1 includes level unemployment rate. The model without
individual fixed effects is clustered at city and individual level.
***p < .001; **p < .01; *p < .05; †p < .1
High
School
Mother married to father or new partner
Unemployment rate
-0.017
0.053
-0.026
(model 1)
(0.044)
(0.072)
(0.085)
Unemployment rate
-0.013
0.062
-0.043
(model 2)
(0.044)
(0.072)
(0.086)
Increasing
0.001
0.009*
-0.009†
unemployment rate
(0.002)
(0.004)
(0.005)
Decreasing
0.010
-0.004
0.022
unemployment rate
(0.010)
(0.016)
(0.020)
Observations
4,120
1,543
1,076
Number of individuals
1,111
420
287
Mother married to or cohabiting with father or new partner
-0.058
-0.079
Unemployment rate
-0.056†
(model 1)
(0.034)
(0.051)
(0.065)
Unemployment rate
-0.052
-0.054
-0.076
(model 2)
(0.034)
(0.052)
(0.065)
Increasing
0.001
0.002
-0.000
unemployment rate
(0.002)
(0.003)
(0.004)
Decreasing
0.006
-0.004
0.018
unemployment rate
(0.007)
(0.011)
(0.015)
Observations
7,187
3,146
2,033
Number of individuals
1,953
863
547
All
Less than
High
School
-0.010
(0.188)
-0.021
(0.193)
-0.006
(0.007)
0.021
(0.038)
288
78
0.090
(0.172)
0.114
(0.174)
0.009
(0.011)
0.022
(0.040)
302
81
-0.061
(0.069)
-0.061
(0.070)
-0.002
(0.004)
0.005
(0.015)
1,698
460
College +
-0.148†
(0.085)
-0.140
(0.086)
0.002
(0.005)
0.014
(0.018)
1,209
325
Some
College
With Individual Fixed Effects
-0.043*
(0.019)
-0.045*
(0.018)
0.000
(0.003)
-0.001
(0.002)
15,855
4,603
-0.038*
(0.018)
0.002
(0.006)
-0.001
(0.001)
-0.040*
(0.018)
15,867
4,604
All
-0.018
(0.063)
-0.026
(0.062)
-0.005†
(0.003)
0.010
(0.007)
4,067
1,162
-0.047
(0.033)
-0.053†
(0.032)
-0.004
(0.002)
0.017†
(0.010)
4,066
1,162
-0.052†
(0.028)
-0.055†
(0.029)
0.000
(0.003)
-0.010†
(0.006)
6,128
1,823
High
School
-0.034
(0.025)
-0.034
(0.026)
0.001
(0.002)
-0.004
(0.009)
6,136
1,823
Less than
High
School
0.092
(0.117)
0.088
(0.124)
-0.002
(0.005)
0.005
(0.017)
1,735
495
0.006
(0.092)
-0.023
(0.099)
-0.010†
(0.005)
-0.001
(0.014)
1,740
496
College +
(Table continues on p. 140.)
-0.051
(0.045)
-0.053
(0.044)
-0.001
(0.002)
-0.001
(0.007)
3,922
1,122
-0.069*
(0.031)
-0.066*
(0.032)
0.001
(0.003)
0.005
(0.010)
3,924
1,123
Some
College
Without Individual Fixed Effects
Table 5.A2 Coefficients and Standard Errors for Unemployment Rate, Relationship Outcomes
All
0.008
(0.009)
0.008
(0.009)
-0.001
(0.002)
-0.000
(0.000)
1,998
771
0.026
(0.027)
0.031
(0.024)
0.008
(0.005)
-0.001†
(0.001)
408
264
0.022
(0.013)
0.021
(0.014)
0.002
(0.005)
0.002
(0.001)
541
356
Some
College
-0.017*
(0.006)
-0.014†
(0.007)
0.001
(0.003)
0.001†
(0.000)
1,791
727
High
School
With Individual Fixed Effects
Less than
High
School
Mother’s report of father’s supportiveness
Unemployment rate
-0.000
0.000
(model 1)
(0.005)
(0.007)
Unemployment rate
-0.000
-0.001
(model 2)
(0.005)
(0.007)
Increasing
0.000
-0.001
unemployment rate
(0.001)
(0.002)
Decreasing
-0.000
-0.000
unemployment rate
(0.000)
(0.000)
Observations
7,632
2,438
Number of individuals
3,024
1,080
Mother’s report of new partners’ supportiveness
Unemployment rate
0.009
0.006
(model 1)
(0.011)
(0.021)
Unemployment rate
0.008
0.003
(model 2)
(0.012)
(0.022)
Increasing
-0.000
-0.005
unemployment rate
(0.003)
(0.005)
0.001*
Decreasing
0.001†
unemployment rate
(0.000)
(0.001)
Observations
1,942
927
Number of individuals
1,259
599
Table 5.A2 Continued
0.001
(0.005)
0.001
(0.005)
0.000
(0.001)
0.000
(0.000)
7,628
3,022
-0.001
(0.008)
-0.002
(0.008)
0.000
(0.002)
-0.000
(0.000)
1,939
1,258
-0.111†
(0.063)
-0.130
(0.080)
0.014
(0.019)
-0.004†
(0.002)
63
39
All
0.005
(0.012)
0.004
(0.013)
0.002
(0.002)
-0.000
(0.001)
1,401
444
College +
-0.001
(0.013)
-0.003
(0.012)
-0.001
(0.001)
-0.002
(0.004)
927
599
-0.000
(0.008)
0.000
(0.008)
0.000
(0.000)
-0.001
(0.002)
2,438
1,080
-0.001
(0.014)
0.001
(0.014)
0.001
(0.001)
0.003
(0.005)
541
356
-0.000
(0.010)
0.001
(0.011)
0.000
(0.001)
0.004
(0.003)
1,791
727
High
School
0.011
(0.026)
0.009
(0.024)
-0.002*
(0.001)
0.001
(0.003)
408
264
0.007
(0.010)
0.007
(0.011)
0.000
(0.001)
-0.001
(0.002)
1,998
771
Some
College
Without Individual Fixed Effects
Less than
High
School
0.007
(0.048)
0.013
(0.065)
-0.001
(0.005)
0.009
(0.030)
63
39
-0.008
(0.014)
-0.012
(0.015)
-0.001†
(0.001)
-0.002
(0.002)
1,401
444
College +
Mother’s report of overall quality of relationship with bio father
Unemployment rate
0.011
0.006
0.017
(model 1)
(0.013)
(0.023)
(0.027)
Unemployment rate
0.011
-0.000
0.025
(model 2)
(0.013)
(0.025)
(0.023)
Increasing
0.000
-0.002
0.005
unemployment rate
(0.003)
(0.007)
(0.004)
Decreasing
0.000
-0.002
0.002
unemployment rate
(0.001)
(0.001)
(0.002)
Observations
7,653
2,453
1,791
Number of individuals
3,020
1,079
725
Father’s report of mother’s supportiveness
Unemployment rate
-0.008*
-0.024** -0.013
(model 1)
(0.004)
(0.008)
(0.012)
Unemployment rate
-0.009*
-0.026** -0.015
(model 2)
(0.004)
(0.007)
(0.011)
Increasing
-0.000
-0.001
-0.001
unemployment rate
(0.000)
(0.000)
(0.000)
Decreasing
-0.001
-0.004
-0.001
unemployment rate
(0.001)
(0.002)
(0.002)
Observations
6,545
2,023
1,525
Number of individuals
2,672
922
638
0.018
(0.018)
0.017
(0.017)
-0.001
(0.003)
-0.000
(0.001)
7,649
3,018
-0.010†
(0.006)
-0.010†
(0.006)
0.000
(0.000)
-0.001
(0.001)
6,542
2,670
0.028
(0.022)
0.019
(0.020)
-0.006
(0.005)
-0.002
(0.002)
1,405
444
0.017†
(0.009)
0.017†
(0.009)
-0.000
(0.001)
0.001
(0.002)
1,279
421
0.004
(0.018)
0.009
(0.017)
0.002
(0.005)
0.001*
(0.001)
2,000
770
-0.010
(0.007)
-0.009
(0.008)
0.000
(0.000)
0.000
(0.002)
1,715
689
-0.024**
(0.008)
-0.024**
(0.008)
0.000
(0.000)
-0.004
(0.003)
2,023
922
0.008
(0.017)
0.005
(0.017)
-0.001
(0.001)
-0.006
(0.004)
2,453
1,079
-0.009
(0.011)
-0.009
(0.012)
0.000
(0.000)
-0.001
(0.002)
1,525
638
0.053
(0.036)
0.055†
(0.033)
0.001
(0.002)
0.004
(0.005)
1,791
725
0.004
(0.010)
0.002
(0.011)
-0.000
(0.001)
-0.002
(0.002)
1,279
421
-0.006
(0.026)
-0.017
(0.024)
-0.003
(0.002)
-0.007†
(0.004)
1,405
444
(Table continues on p. 142.)
-0.011
(0.009)
-0.011
(0.008)
0.000
(0.000)
0.001
(0.002)
1,715
689
0.010
(0.023)
0.014
(0.023)
0.001
(0.001)
0.004
(0.005)
2,000
770
All
Less than
High
School
High
School
-0.003
(0.029)
-0.003
(0.028)
-0.001
(0.003)
0.000
(0.001)
1,298
423
College +
-0.022†
(0.012)
-0.024†
(0.013)
-0.006
(0.004)
-0.000
(0.001)
6,803
2,771
All
-0.001
(0.019)
-0.005
(0.020)
-0.001
(0.001)
-0.005
(0.020)
2,146
974
Less than
High
School
-0.025
(0.025)
-0.027
(0.025)
0.001
(0.001)
-0.027
(0.025)
1,585
666
High
School
-0.060**
(0.023)
-0.060*
(0.024)
-0.000
(0.001)
-0.060*
(0.024)
1,774
708
Some
College
Without Individual Fixed Effects
-0.014
(0.031)
-0.016
(0.032)
0.001
(0.001)
-0.016
(0.032)
1,298
423
College +
Source: Authors’ calculations based on Fragile Families and Child Wellbeing Study data.
Note: Standard errors and z-stats in parentheses. Model 1 includes unemployment rate as a level. Model 2 includes unemployment rate as a level as well as rate increasing change
and the rate of decreasing change in unemployment rate. SEs for the OLS with fixed effects are clustered at city, for OLS and logistic models without fixed effects are clustered at
city and individual.
**p < .01; *p < .05; †p < .1
-0.036*
(0.016)
-0.032†
(0.018)
0.007
(0.006)
0.001
(0.001)
1,774
708
Some
College
With Individual Fixed Effects
Father’s report of overall quality of relationship with bio mother
Unemployment rate
-0.014
-0.006
-0.026
(model 1)
(0.012)
(0.010)
(0.026)
Unemployment rate
-0.017
-0.012
-0.030
(model 2)
(0.012)
(0.010)
(0.026)
Increasing
-0.004
-0.010*
-0.010†
unemployment rate
(0.003)
(0.005)
(0.005)
Decreasing
-0.000
-0.002
0.000
unemployment rate
(0.001)
(0.001)
(0.001)
Observations
6,807
2,146
1,585
Number of individuals
2,773
974
666
Table 5.A2 Continued
Mother married to father or new partner
Unemployment rate (model 1)
Unemployment rate (model 3)
Mother’s unemployment
Bio-social fathers not employed
Unemployment rate (model 4)
Unemployment rate * year nine
Mother married to or cohabiting with father or new partner
Unemployment rate (model 1)
Unemployment rate (model 3)
Mother’s unemployment
Bio-social father’s not employed
Unemployment rate (model 4)
Unemployment rate * year nine
Mother’s report of father’s supportiveness
Unemployment rate (model 1)
Unemployment rate (model 3)
Mother’s unemployment
Bio-social father’s not employed
Unemployment rate (model 4)
Unemployment rate * year nine
Mother’s report of new partners’ supportiveness
Unemployment rate (model 1)
Unemployment rate (model 3)
Mother’s unemployment
Bio-social fathers not employed
Unemployment rate (model 4)
Unemployment rate * year nine
(0.044)
(0.053)
(0.126)
—
(0.060)
(0.069)
(0.034)
(0.079)
(0.040)
—
(0.047)
(0.051)
(0.005)
(0.004)
(0.018)
(0.018)
(0.007)
(0.007)
(0.011)
(0.014)
(0.028)
(0.029)
(0.014)
(0.014)
-0.017
0.040
-0.081
—
-0.067
0.084
-0.056†
-0.034
-0.056
—
-0.111*
0.086†
-0.000
-0.000
0.029
0.010
0.008
-0.017*
0.009
0.016
-0.029
-0.012
-0.001
0.014
With Individual Fixed
Effects
(0.008)
(0.010)
(0.015)
(0.022)
(0.013)
(0.012)
-0.001
0.007
-0.046**
-0.032
-0.010
0.011
(Table continues on p. 144.)
(0.005)
(0.007)
(0.014)
(0.017)
(0.009)
(0.008)
(0.019)
(0.024)
(0.051)
—
(0.026)
(0.024)
(0.018)
(0.025)
(0.077)
—
(0.027)
(0.029)
0.001
-0.001
-0.018
-0.048**
0.013
-0.024**
-0.043*
-0.048*
-0.230***
—
-0.063*
0.032
-0.038*
-0.022
-0.371***
—
-0.040
0.002
Without Individual
Fixed Effects
Table 5.A3 Sensitivity of Unemployment Rate Coefficients, Relationship Outcomes
(0.013)
(0.014)
(0.048)
(0.054)
(0.016)
(0.015)
(0.004)
(0.005)
(0.020)
(0.015)
(0.004)
(0.006)
(0.012)
(0.016)
(0.046)
(0.050)
(0.007)
(0.013)
0.011
0.005
0.009
-0.072
0.012
-0.002
-0.008*
-0.008
0.008
-0.006
-0.005
-0.006
-0.014
-0.038*
0.084†
-0.091†
0.010
-0.044**
With Individual Fixed
Effects
(0.018)
(0.025)
(0.031)
(0.038)
(0.022)
(0.018)
(0.006)
(0.007)
(0.014)
(0.014)
(0.006)
(0.007)
(0.012)
(0.016)
(0.039)
(0.035)
(0.017)
(0.016)
0.018
-0.002
0.011
-0.189***
0.036
-0.038*
-0.010†
-0.011
0.001
0.011
-0.003
-0.013†
-0.022†
-0.037*
-0.066†
-0.033
0.025
-0.085***
Without Individual
Fixed Effects
Source: Authors’ calculations based on Fragile Families and Child Wellbeing Study data.
Note: Standard errors and z-stats in parentheses. Model 3 includes unemployment rate and a measure of individual unemployment. Model 4
includes unemployment rate and an interaction between unemployment rate and year nine, when the Great Recession hit. SEs for the OLS
with fixed effects are clustered at city, for OLS and logistic models without fixed effects are clustered at city and individual.
***p < .001; **p < .01; *p < .05; †p < .1
Mother’s report of overall quality of relationship with bio father
Unemployment rate (model 1)
Unemployment rate (model 3)
Mother’s unemployment
Bio-social father’s not employed
Unemployment rate (model 4)
Unemployment rate * year nine
Father’s report of mother’s supportiveness
Unemployment rate (model 1)
Unemployment rate (model 3)
Mother’s unemployment
Bio-social fathers not employed
Unemployment rate (model 4)
Unemployment rate * year nine
Father’s report of overall quality of relationship with bio mother
Unemployment rate (model 1)
Unemployment rate (model 3)
Mother’s unemployment
Bio-social fathers not employed
Unemployment rate (model 4)
Unemployment rate * year nine
Table 5.A3 Continued
-0.018***
(0.003)
-0.031*
(0.013)
0.013
(0.022)
0.005
(0.009)
-0.025
(0.035)
0.015
(0.021)
-0.006
(0.005)
0.000
(0.023)
0.012
(0.020)
-0.011
(0.022)
-0.006
(0.015)
0.000
(0.005)
0.014
(0.063)
0.032
(0.075)
Hispanic
0.048†
(0.025)
-0.002
(0.011)
-0.023
(0.051)
-0.132
(0.082)
0.003
(0.007)
-0.011
(0.081)
-0.074
(0.087)
Black
-0.015
(0.025)
-0.035**
(0.010)
-0.026***
(0.006)
-0.007
(0.017)
0.021†
(0.012)
0.008
(0.005)
0.009
(0.026)
-0.006
(0.007)
-0.081
(0.056)
-0.040
(0.066)
Cohabiting
at Baseline
-0.060
(0.037)
0.006
(0.008)
-0.041
(0.119)
-0.051
(0.120)
Married at
Baseline
0.041
(0.031)
-0.026
(0.021)
0.021
(0.037)
0.016
(0.011)
-0.002
(0.010)
-0.087†
(0.048)
-0.040
(0.088)
Single at
Baseline
Source: Authors’ calculations based on Fragile Families and Child Wellbeing Study data.
Note: Standard errors and z-stats in parentheses. Model 1 includes level unemployment rate; results include individual fixed effects and time. SEs for the OLS with
fixed effects are clustered at city.
***p < .001; **p < .01; *p < .05; †p < .1
Father’s report of overall quality of relationship with bio mother
Unemployment rate
Father’s report of mother’s supportiveness
Unemployment rate
Mother’s report of overall quality of relationship with bio father
Unemployment rate
Mother’s report of new partners’ supportiveness
Unemployment rate
Mother’s report of father’s supportiveness
Unemployment rate
Mother married to or cohabiting with father or new partner
Unemployment rate
Mother married to father or new partner
Unemployment rate
White
Table 5.A4 Coefficients and Standard Errors for Unemployment Rate, Relationship Outcomes
146
children of the great recession
NOTES
1. Ogburn and Nimkoff 1955; Cherlin 1992.
2. Lichter, McLaughlin, and Ribar 2002; Blau, Kahn, and Waldfogel 2000;
Moffitt 2000.
3. Schaller 2012; Schneider and Hastings 2015.
4. Ogburn and Nimkoff 1955.
5. Willcox 1893; Ogburn and Thomas 1922; Gulden 1939.
6. Conger and Elder 1994; South 1985; Fischer and Liefbroer 2006.
7. Amato and Beattie 2011; Hellerstein and Morrill 2011; Schaller 2012.
8. Cherlin et al. 2013.
9. Cohen 2014.
10. Cherlin et al. 2013.
11. Komarovsky 1940.
12. Conger et al. 1999; Conger and Elder 1994.
13. Fox et al. 2002; Benson et al. 2003; Vinokur et al. 1996.
14. Patterson 1998; Hatchett et al. 1995.
15. Bakke 1940; Komarovsky 1940.
16. The father sample is positively selected on seriousness of relationship with
mother. Fathers who were interviewed tended to have closer relationships with
mothers (for example, to be married or cohabiting) than those who were not.
The positive selection explains why fathers tend to report higher mean levels
of relationship quality, but why this selectivity would bias the comparison of
father reports of supportiveness under strong and weak economic conditions
is unclear (see figure 5.13). Therefore, the decline in fathers’ reports of
mothers’ supportiveness but not mothers’ reports of fathers’ supportiveness
likely reflects a differential response by gender to the recession conditions.
17. Schneider, McLanahan, and Harknett 2016.
18. Conger and Elder 1994.
REFERENCES
Amato, Paul R., and Brett Beattie. 2011. “Does the Unemployment Rate Affect
the Divorce Rate? An Analysis of State Data 1960–2005.” Social Science
Research 40(3): 705–15.
Bakke, E. Wight. 1940. Citizens Without Work: A Study of the Effects of
Unemployment Upon the Workers’ Social Relations and Practices. New Haven,
Conn.: Yale University Press.
Benson, Michael L., Greer L. Fox, Alfred DeMaris, and Judy Van Wyk. 2003.
“Neighborhood Disadvantage, Individual Economic Distress and Violence
Against Women in Intimate Relationships.” Journal of Quantitative Crim­
inology 19(3): 207–35.
parents’ relationships147
Blau, Francine D., Laurence M. Kahn, and Jane Waldfogel. 2000. “Understanding
Young Women’s Marriage Decisions: The Role of Labor and Marriage Market
Conditions.” Industrial and Labor Relations Review 58(4): 624–47.
Cherlin, Andrew J. 1992. Marriage, Divorce, Remarriage. Cambridge, Mass.:
Harvard University Press.
Cherlin, Andrew, Erin Cumberworth, S. Philip Morgan, and Christopher
Wimer. 2013. “The Effects of the Great Recession on Family Structure and
Fertility.” Annals of the American Academy of Political and Social Science
650(1): 214–31.
Cohen, Phillip N. 2014. “Recession and Divorce in the United States, 2008–2011.”
Population Research and Policy Review 33(5): 615–28.
Conger, Rand D., and Glenn H. Elder Jr. 1994. Families in Troubled Times:
Adapting to Change in Rural America. Social Institutions and Social Change.
New York: Aldine de Gruyter.
Conger, Rand D., Martha A. Rueter, and Glenn H. Elder Jr. 1999. “Couple
Resilience to Economic Pressure.” Journal of Personality and Social Psychology
76(1): 54–71.
Fischer, Tamar, and Aart C. Liefbroer. 2006. “For Richer, for Poorer: The Impact
of Macroeconomic Conditions on Union Dissolution Rates in the Netherlands
1972–1996.” European Sociological Review 22(5): 519–32.
Fox, Greer L., Michael L. Benson, Alfred A. DeMaris, and Judy Van Wyk.
2002. “Economic Distress and Intimate Violence: Testing Family Stress and
Resources Theories.” Journal of Marriage and the Family 64(3): 793–807.
Gulden, Tees. 1939. “Divorce and Business Cycles.” American Sociological
Review 4(2): 217–23.
Hatchett, Shirley J., Joseph Veroff, and Elizabeth Douvan. 1995. “Marital
Instability among Black and White Couples in Early Marriage.” In The Decline
in Marriage Among African Americans, edited by M. Belinda Tucker and
Claudia Mitchell-Kernan. New York: Russell Sage Foundation.
Hellerstein, Judith K., and Melinda S. Morrill. 2011. “Booms, Busts, and
Divorce.” B.E. Journal of Economic Analysis & Policy 11(1): ISSN (Online)
1935–1682.2914.
Komarovsky, Mirra. 1940. The Unemployed Man and His Family. New York:
Dryden Press.
Lichter, Daniel T., Diane K. McLaughlin, and David C. Ribar. 2002. “Economic
Restructuring and the Retreat from Marriage.” Social Science Research 31(2):
230–56.
Moffitt, Robert A. 2000. “Welfare Benefits and Female Headship in US Time
Series.” American Economic Review 90(2): 373–77.
Ogburn, William F., and Meyer F. Nimkoff. 1955. Technology and the Changing
Family. Westport, Conn.: Greenwood Press.
Ogburn, William F., and Dorothy S. Thomas. 1922. “The Influence of the Business
Cycle on Certain Social Conditions.” Journal of the American Statistical
Association 18(139): 324–40.
Patterson, Orlando. 1998. Rituals of Blood: Consequences of Slavery in Two
American Centuries. Washington, D.C.: Civitas/CounterPoint.
Schaller, Jessamyn. 2012. “For Richer, if Not for Poorer? Marriage and Divorce
over the Business Cycle.” Journal of Population Economics 26(3): 1007–33.
Schneider, Daniel, Sara S. McLanahan, and Kristen Harknett. 2016. “Intimate
Partner Violence in the Great Recession.” Demography 53(2): 471–505.
148
children of the great recession
Schneider, Daniel, and Orestes P. Hastings. 2015. “Socio-Economic Variation
in the Effect of Economic Conditions on Marriage and Non-marital Fertility:
Evidence from the Great Recession.” Demography 52(6): 1983–15.
South, Scott J. 1985. “Economic Conditions and the Divorce Rate: A Time Series
Analysis of the Postwar United States.” Journal of Marriage and Family 47(1):
31–41.
Vinokur, Amiram D., Richard H. Price and Robert D. Caplan. 1996. “Hard
Times and Hurtful Partners: How Financial Strain Affects Depression and
Relationship Satisfaction of Unemployed Persons and their Spouses.” Journal
of Personality and Social Psychology 71(1): 166–79
Willcox, Walter F. 1893. “A Study in Vital Statistics.” Political Science Quarterly
8(1): 69–96.
Chapter 6
Nonresident Father Involvement
Ronald B. Mincy and Elia De la Cruz Toledo
P
revious chapters in this volume show that big increases in unemployment detrimentally affect the economic well-being of families with
young children, and also lead to a decrease in the probability that mothers
are partnered with either children’s biological fathers or a new partner.
That a father is not physically present in the home of a child, however, does
not mean that he is not involved in his life. This chapter thus examines the
impact unemployment had on the involvement of nonresident fathers in
the lives of their children.
Children born in the late 1990s were born to fathers who reached fertile age following a fifteen-year decline in the average earnings of men
lacking a college education. Nonmarital births became normative among
such men and their partners and the proportion of children with nonresident fathers reached an all-time high. Over the same period, however, the
social expectation that nonresident fathers should be financially responsible for children also became widespread throughout the United States,
along with the legal and administrative apparatus to secure those expectations. Consistent with this expectation, contact between nonresident
fathers and their children has been growing over time, calling into question the idea that declining nonresident father-child contact as children
age is the typical pattern.1 But this was before the Great Recession, the
worst economic downturn in the postwar period. Did this bring about a
collision between the commitment that nonresident fathers pay child support and their ability to do so? Did it reduce father-child contact, at least
while unemployment remained high?
This chapter examines changes in formal, informal, and in-kind child
support and visitation among nonresident fathers during the Great
Recession. We first review the theoretical and empirical literature about
the effect of unemployment on formal and informal child support and
visitation among nonresident fathers, including the few studies that focus
on the Great Recession.
150
children of the great recession
HOW DOES UNEMPLOYMENT AFFECT FINANCIAL
SUPPORT AND VISITATION?
Economic factors are one of the primary determinants of child support.2
Economic downturns result in declines in earnings, sometimes to zero,
at least temporarily. This can lead to new child support orders by parents
who divorce or by never-married parents who initiate formal child support
orders because fathers reduce their informal agreement compliance during recessions.
Economic downturns can also lead to reductions in formal child support payments among noncustodial parents (NCP) with existing orders.
Because it is costly and time-consuming and the outcome is uncertain,
many NCPs do not attempt to reduce their child support orders when
their earnings decline.3 Those who do may get a downward modification
only after a long delay. Immediately after a reduction in earnings, many
NCPs therefore pay less than the full amount of child support due.
Economic downturns may also affect fathers’ nonfinancial involvement with their children. Studies consistently show that people typically
experience stress as a result of unemployment, with men and blue-collar
workers more likely to do so than women and white-collar workers.4
Because unemployment makes it difficult for nonresident fathers to provide for their children, unemployment leads to stress among nonresident
fathers. Researchers who focus on chronically unemployed, especially
black, nonresident fathers label this effect provider role strain.5 Much
qualitative literature supports these ideas about provider role strain. For
example, Elijah Anderson finds that the chronic unemployment among
nonresident black fathers reduces their contact with children.6 Many
qualitative studies find that shame, associated with lack of resources, discourages these fathers from seeking visitation.7 Other qualitative studies
find that unemployed, nonresident fathers seek visitation, but that, lacking the resources to provide adequate support for their children, they are
unable to pass critical gatekeepers, such as mothers and their relatives, to
see their children.8 Despite generally positive feelings for their children
early on, recurring spells of unemployment may result in less and less
visitation over time.9
Unemployment might also affect visitation indirectly, through the effect
of unemployment on child support compliance or payments. Fathers who
pay child support may want to monitor how custodial mothers spend their
child support payments. A reduction in compliance following an increase
in unemployment would reduce fathers’ incentive to monitor child support payments.10 Second, fathers who pay child support meet the expectations of mothers, children’s relatives, and other community stakeholders,
and thereby garner permission to see their children.11
nonresident father involvement151
EMPIRICAL EVIDENCE
A large literature focuses on the determinants of fathers’ financial and
nonfinancial involvement in the lives of their children. Unfortunately,
changes in unemployment rates have not played a prominent role in this
literature. Economic downturns affect financial contributions from fathers
through their effects on earnings, which most studies control or proxy by
demographic and human capital characteristics (age, race-ethnicity, and
education). Studies of visitation by nonresident fathers prominently feature these same variables because these characteristics are good proxies for
fathers’ provider role strain, readiness to play responsible parenting roles,
and cultural factors that should affect visitation. When unemployment
rates do appear in these studies, they often represent contextual factors,
which authors sometimes use to identify the child support equation in a
system in which both it and visitation are simultaneously determined.12
Child Support Compliance
Evidence that economic downturns spur new child support orders among
divorced or never-married parents is scant.13 Many studies, however, show
that child support payments and compliance are positively associated with
earnings capacity or income, as predicted.14 Chi-Fang Wu, for instance,
shows that NCPs who experienced larger reductions in earnings between
2006 and 2009 were also less likely to pay child support.15
Not much direct empirical evidence supports the association between
economic downturns and child support compliance. Some studies have
examined the association between the unemployment rate and child support compliance (or payments) after controlling for other variables.16
Among these, only one finds a statistically significant association between
unemployment and child support compliance; in this case, the association
has an unexpected (positive) sign.17
One reason for these results is that most studies using nationally representative data that include an unemployment rate typically include controls for a variety of demographic characteristics (such as age, race, and
education), which are also good proxies for earnings. After including such
controls, unemployment may account for little of the remaining variation
in child support compliance. What is more, these studies were estimated
over sample periods of economic growth or mild recessions, during which
the employment rates are growing or falling modestly.18 During more
severe recessions, many more workers become separated from their jobs
and from immediate wage withholding, which causes a drop in compliance. Therefore, after immediate wage withholding took effect in 1988, we
should expect compliance to be more responsive to economic downturns.
152
children of the great recession
In general, the literature on economic downturns and child support
tends to focus on nonrepresentative samples or methods not explicitly
designed to capture the effect of economic conditions. One recent study,
however, uses data from the Current Population Survey-Child Support
Supplement (CPS-CSS) to examine the association between unemployment and the probability that mothers received any child support payments, all payments, and all payments in full.19 On average, an increase in
the unemployment rate was associated with a decrease in the probability that
a mother received all payments due to her that were not passed through the
welfare system and a decrease in the probability that all non-pass-through
payments were for the full amount. These associations appeared to have
been driven by less-advantaged mothers, who were probably owed child
support by less-advantaged fathers. No association was evident between
unemployment and any payments, or between unemployment and either
measure of full compliance for more-advantaged mothers.
Informal Support
Although empirical studies of formal child support compliance rarely
focus on the role of unemployment rates, one empirical study of informal
child support does. Lenna Nepomnyaschy and Irwin Garfinkel include
a control for the unemployment rate in their models designed to examine the effect of enforcement on informal child support.20 They find that
mothers who lived in cities with higher unemployment rates tended to
receive more informal cash child support. Unemployment rates were not
significantly associated with in-kind support, however.
Visitation
Studies based on individual data show that employed nonresident fathers
are more likely to visit or engage with their children than those who are
not employed.21 However, few studies incorporate a measure of aggregate unemployment among the control variables in studies of nonresident father visitation. Two important exceptions are Jonathan Veum and
Judith Seltzer and her colleagues, who estimate the causal relationship
between child support and visitation using longitudinal data.22 Although
they reach different conclusions about the causal relationship between
child support and visitation, neither finds a statistically significant association between the unemployment rate and visitation. To our knowledge,
no study has estimated the association between unemployment and visitation during the Great Recession.
Child-Support Orders, Payments, and Visitation over Time
We begin by documenting trends in child support over children’s first
nine years. We measure changes in formal child support over the previous
nonresident father involvement153
Figure 6.1 Nonresidence Status
70
Percentage of Fathers
60
College +
50
Some college
40
30
High school
20
Less than
high school
10
0
1
3
5
9
Child’s Age-Year
Source: Authors’ calculations using Fragile Families and Child Wellbeing Study data circa
1998 to 2010.
year among nonresident fathers who have a court order to provide child
support. We also measure informal (not court ordered) cash child support and in-kind child support. Both formal and informal cash support
are measured in dollars; in-kind support is measured as a binary variable
that takes the value of one if any in-kind support such as toys, medicines,
clothes, food, and school-related items is provided. All monetary units are
in constant dollars from 1999. Nonresident fathers sometimes provide
more than one type of support. It is common to observe combinations of
informal and in-kind child support or formal and in-kind child support.
Figure 6.1 shows changes in the proportion of nonresident fathers by the
age of the child. From the time focal children in the Fragile Families and
Child Wellbeing Survey were one year old until they were nine, the proportion of nonresident fathers at different levels of education (high school
dropouts, high school graduates, some college attendees, and college
graduates) increases. Less-educated fathers were consistently the most
likely to be nonresident; more-educated fathers were less likely at every
age to be nonresident. Fathers with a high school diploma showed the
highest rates of nonresidency, at 36 percent when the child was one year
old, which rose to 55 percent when the child was nine years old. Fathers
with a college education or more had the lowest rates of nonresidency,
between 6 percent at age one and 13 percent at age nine.
As the proportion of nonresident fathers grows, so does the importance
of understanding monetary support and patterns of fathers’ visitation. We
show changes over time in several indicators of father engagement: child
support orders, formal, informal and in-kind child support, and visitation
154
children of the great recession
Figure 6.2 Father Engagement
Percentage of Fathers
70
60
Child support orders
50
Formal support
40
Informal support
30
In-kind support
20
Visitation
10
0
1
3
5
(1999–2001) (2001–2003) (2003–2006)
9
(2007–2010)
Child’s Age-Year
Source: Authors’ calculations using Fragile Families and Child Wellbeing Study data circa
1998 to 2010.
in figure 6.2. All fathers who became nonresident by the time the child
was age nine are included in the sample. The proportion of nonresident
fathers with child support orders when the focal child was one year old was
12 percent; by the time the child was nine, the proportion of fathers with
court orders to provide child support had grown to 48 percent (see figure 6.2). Formal and informal child support payments show different trajectories. As court orders to provide formal child support increase through
the years, heterogeneity increases in the nonresident sample; also, conditional on having a court order, the proportion of fathers who pay formal
child support decreases from 67 percent when the child was one year old
to 57 percent at age nine. However, the absolute number of fathers who
paid formal child support increased through the years. When the child was
one, 53 percent of nonresident fathers provided informal child support;
by age nine, only 29 percent did so. In-kind child support showed a similar
trajectory, dropping from 63 percent to 53 percent. Last, the proportion
of fathers who visited their children at least once a month during the year
was 63 percent in the child’s first year, but 50 percent by age nine.
Large disparities in the provision of formal child support are observed
among nonresident fathers of different educational backgrounds. Among
nonresident fathers with a child support order, those who dropped out of
high school provided an average yearly payment of $273; fathers with a
high school diploma provided $464, college dropouts granted $775, and
college graduates provided $1,430. As shown in figure 6.3, differences by
education in the proportion of nonresident fathers who pay any formal
nonresident father involvement155
Figure 6.3 Child Support and Visitation
Days
Percentage of Fathers
100
80
Formal
7.5
Informal
7.0
60
In-kind
6.5
6.0
40
20
8.0
5.5
Less than
high school
High
school
Some
college
College +
5.0
Visited at least
once a month
Visitation days
per month
(right axis)
Source: Authors’ calculations using Fragile Families and Child Wellbeing Study data circa
1998 to 2010.
child support or provide any in-kind support are also large; differences in
informal cash support and visitation are comparatively smaller.
In sum, then, these results show increasing prevalence of nonresident
fatherhood and formal child support orders over time, but a decrease in formal child support payments conditional on having an order and a decrease
in informal cash and in-kind support and visitation. In the next section,
we examine whether and to what extent the Great Recession altered these
trends.
THE GREAT RECESSION, CHILD SUPPORT,
AND VISITATION IN FRAGILE FAMILIES
As in previous chapters, to estimate the effects of business cycles on formal
child support, informal child support, and visitation, we take advantage
of the vast differences in unemployment rates experienced by our respondents over time (when their children were one, three, five, and nine years
old). Both economic booms and busts are captured in the data. In this
section, we first examine the relationship between the local unemployment
rate and child support outcomes and then the relationship between the
local unemployment rate and visitation outcomes. Both analyses are net of
nonresident fathers’ demographic characteristics.23 We use random rather
than fixed-effects models and predict child support and visitation outcomes
given a change in the unemployment rate from 5 percent to 10 percent,
156
children of the great recession
Figure 6.4 Formal Child Support per Year
$4,000
Amount Paid
3,500
−20%†
−13%*
−26% **
3,000
2,500
UR 5 percent
2,000
1,500
UR 10 percent
−11%
1,000
500
0
All
Less than
high school
High
school
Some
college +
Source: Authors’ calculations using Fragile Families and Child Wellbeing Study data circa
1998 to 2010.
***p < .001; **p < .01; *p < .05; †p < .1
which is approximately the size of the change experienced by families during the Great Recession (for detail on the methodology and results, see the
appendix). We also consider the impacts of the recession by fathers’ educational attainment. Because so few college-educated fathers separated or
divorced and became nonresident fathers, the college-educated and some
postsecondary education groups are combined.
As observed in figures 6.4, 6.5, 6.6, 6.7, and 6.8, formal and informal cash
support decline substantially—around 20 percent—but in-kind support
and visitation are hardly affected by a deep recession. Only the relationship
between formal child support and unemployment is statistically significant at
conventional levels. Table 6.A1 shows an average decrease of $105.8 in formal child support per one percentage point increase in the unemployment
rate. High school dropouts show the least adverse effects of the recession in
formal support, but bigger percentage drops in informal support. Although
the drops in informal support in figure 6.6 are estimated with a great deal
of error, they are all negative and fairly substantial. By way of contrast, the
changes in in-kind support (figure 6.6) and visitation are much smaller and
number of days of visiting (conditional on any visiting) are actually positive.
We also examine differences by race-ethnicity and find no significant
differences (see table 6.A4).
Capturing the Effect of Unemployment
A number of additional analyses test the links between the deterioration of
the economy and child support and visitation. In model 2, we add to our
core model two terms to capture increases and decreases in the unemploy-
nonresident father involvement157
Figure 6.5 Informal Child Support per Year
$1,000
−16%
−21%
Amount Paid
800
−13%
−16%
600
UR 5 percent
400
UR 10 percent
200
0
All
Less than
high school
High
school
Some
college +
Source: Authors’ calculations using Fragile Families and Child Wellbeing Study data circa
1998 to 2010.
ment rate. In model 3, we include a mother’s and a father’s self-reported
employment. In model 4, we add an interaction term with the unemployment rate and the year nine wave of data collection to test whether the
association between the unemployment rate and outcomes differed during
the recession.
Effects of Increases and Decreases in the Unemployment Rate
The stress associated with the anticipation of economic adversity might
be an additional pathway that affects fathers’ decisions to provide child
Figure 6.6 In-Kind Child Support per Year
Percentage of Fathers
100
80
60
−10%
0%
+1%
−4%
UR 5 percent
UR 10 percent
40
20
0
All
Less than
high school
High
school
Some
college +
Source: Authors’ calculations using Fragile Families and Child Wellbeing Study data circa
1998 to 2010.
158
children of the great recession
Figure 6.7 Visitation Days per Month
Number of Days
6
+10%
5
+11%
+7%
+11%
4
UR 5 percent
3
UR 10 percent
2
1
0
All
Less than
high school
High
school
Some
college +
Source: Authors’ calculations using Fragile Families and Child Wellbeing Study data circa
1998 to 2010.
support or visit their children. Results are shown in table 6.A2 (model 2).
Overall, they suggest that an accelerated unemployment rate or “things
getting worse” does not affect nonresident father engagement. These
results suggest that stress has no impact on a father’s decision to provide
child support or to visit the child.
Individual Unemployment
Another potential pathway through which a deteriorated economy could
lead to changes in nonresident father involvement is a mother’s or father’s
Figure 6.8 Share of Nonresident Fathers Visiting Their Children
Percent of Fathers
100
80
60
0%
+1%
−2%
+3%
UR 5 percent
UR 10 percent
40
20
0
All
Less than
high school
High
school
Some
college +
Source: Authors’ calculations using Fragile Families and Child Wellbeing Study data circa
1998 to 2010.
nonresident father involvement159
self-reported employment. Results are shown in table 6.A3 (model 3). As
expected, we find that mothers’ individual-level unemployment does not
affect child support or visitation outcomes. However, we do find a large
and significant effect of father’s self-reported unemployment on all of the
engagement outcomes. All estimates are larger in magnitude in models
that do not include fixed effects.
Unemployment Rate During the Great Recession
During the Great Recession, unemployment rates rose consistently across
the country. Thus we estimate the interaction of the unemployment rate
and year nine that captures the period of interest (see table 6.A3, model 4).
Evidence indicates that the effects of the unemployment in the last wave
do not differ particularly from those of prior years.
Discussion
By the middle of the twentieth century, the average earnings of men without a college degree had been declining for fifteen years, nonmarital births
were common, and the legal and administrative mechanisms for enforcing
the social expectation that nonresident fathers support their children were
firmly in place. Trends in nonresident father visitation also ran counter
to the long-held view that nonresident fathers were uninvolved in the
lives of their children. Doubling of the unemployment rate during the
Great Recession of 2007 to 2009 might well have reversed these trends.
Unemployment reduces nonresident fathers’ ability to provide financial
support, either formally or informally. To the extent that financial support in either form declined, visitation would also decrease because fathers
might be reluctant to visit children for whom they were providing less
financial support or mothers might retaliate by blocking fathers’ access to
their children.
This chapter examines associations between lagged unemployment
rates on the one hand and father involvement on the other. We use these
estimated associations to predict how much our measures of father involvement (formal, informal, and in-kind child support and visitation) would
change following an increase in the unemployment rate from 5 percent
to 10 percent, about the size of the change during the Great Recession,
which occurred about seven years after our sample children were born.
Our predictions show that such an increase in unemployment would be
associated with an average 20 percent decline in formal and informal cash
support, but with little change in in-kind support or visitation. Further,
only the relationship between formal child support and unemployment
was statistically significant at conventional levels and this association was
lower among fathers with less education. What is more, it was the change
160
children of the great recession
in the level of unemployment rather than the stress associated with the
perception that things were getting worse that was significantly and negatively associated with formal child support payments. By contrast, all
of our measures of father involvement were negatively and significantly
associated with the fathers’ self-reported employment status.
Our findings suggest that the association between unemployment
and formal child support during the 2007–2009 recession was no different from that in previous recessions. Like previous studies, our models control for demographic characteristics, which are associated with
men’s earnings. Earnings typically decline when the unemployment rate
rises. Perhaps as a result few prior studies find a significant association
between unemployment and formal child support. Still the large variation in unemployment rates that occurred in our sample, which included
both the relatively mild recession 2001 recession and the much more
severe Great Recession, could explain why our results differ from previous studies.
APPENDIX
We use data from waves 1 through 5 of the Fragile Families and Child
Wellbeing Study. In all of our analyses, we pool the data (N ~16,400) and
use waves 2 through 5 for our key independent and dependent variables
(covariates are measured at baseline, wave 1, survey).
We use multiple imputation techniques to obtain information on the
missing parents. We believe fathers and mothers are not missing completely at random, but we are confident that by controlling for socioeconomic and demographic characteristics in our imputation models our
estimates are unbiased.
Measures
We use three measures to track child support outcomes.
Formal child support assesses the monetary value of child support
paid by fathers who had a legal child support order. To determine
it, the question is, “How much of the legally agreed child support
has father actually paid over the last year?” Amounts are deflated
to constant dollars from 1999.
Informal child support is determined one of two ways. Of mothers
who have an informal agreement, the question is, “How much
informal support have you received since informal agreement
was reached?” Of mothers who do not, the question is about
anything received from father in past year. Amounts are deflated
to constant dollars from 1999.
nonresident father involvement161
To measure in-kind child support, the question is how often father
bought articles for his child such as toys, medicines, clothes,
food, and school-related items. Answers are binary: 1 indicating
purchases and 0 indicating none.
We use two measures to track visitation outcomes.
Visitation in the last month (at least once). To assess more nuanced
changes, the question is, “During the past thirty days, has the
biological father seen child?” The answer was coded 1 for yes
and 0 for no.
Days of visitation in the last month. To explore the changes in visitation at the intensive margin (continuous response), the question
is, “During the past thirty days, how many days has father seen
child?”
Key Independent Variable
The main unemployment rate used in this study is based on father’s area
of residence to reflect his relevant labor market conditions. This rate is a
seasonally adjusted average of the twelve months before the month of the
interview at the metropolitan area of residence at the time of the interview
in the first wave (baseline). First, we use the unemployment rate at baseline
interview and not current city of residence to avoid problems of endogeneity with moving decisions. Second, a twelve-month average specification
is based on the timing in which the child support and visitation outcomes
were framed. Mothers were asked about father’s child support compliance
“since last month” at the time of the interview; if a month could not be
provided, respondents were asked about a father’s child support compliance in the past year. The visitation questions required information from
previous years. Thus, specifying a lagged average of the unemployment
rate allows us to analyze the effects of changes in the fathers’ labor market
conditions on child support payments for the full sample and for visitation
outcomes.
Control Variables
Based on previous literature, we include a number of father and child
characteristics as well as mothers’ preferences as covariates. Along with
a continuous measure of father’s age at the birth, we include several
dummy variables to measure father’s other demographic characteristics:
race-ethnicity (non-Hispanic black, non-Hispanic white, Hispanic, or
other race, with the omitted category being non-Hispanic white), and
162
children of the great recession
immigrant status (foreign born = 1). We also include a number of variables that proxy father’s commitment to the child: parental relationship
status at child birth (married, cohabiting or single), mother’s preference
on father’s involvement in child’s upbringing (yes, no), father’s hospital
visitation at childbirth (yes, no). Last, we include controls that reflect
father’s earnings capacity: father’s employment status and father’s incarceration record at baseline.
Method
To estimate the association between changes in the unemployment rate
and child support–visitation, we use a pooled linear probability model
for the dichotomous outcomes (in-kind child support and extensive
margin visitation) and an ordinary least squares model (OLS) for continuous outcomes (yearly formal and informal child support and monthly
visitation). Our outcomes of interest are the three measures of child
support (formal, informal, and in-kind) and two measures of visitation. Formal and informal child support are measured in yearly dollar
amounts; in-kind child support is measured as a binary outcome where
a value of 1 indicates that the father provides any in-kind child support
and 0 that he provides none. Visitation is measured in two ways: as a
binary outcome and as a continuous number of days in a month. Our
key independent variable of interest is the aggregate unemployment rate
at the father’s metropolitan area of residence. We include control variables for the mothers’ preferences and fathers’ characteristics described
earlier. We also add time and city fixed effects to account for unmeasured,
geographic, and time-specific factors that are constant over time. This
method removes a potentially large source of omitted variable bias over
time and location. We do not use individual fixed effects. Information
on area of residence was missing for 30 percent of the fathers in our
sample. More than likely, the missing information comes from the most
disadvantaged fathers, who probably were hit hardest by the recession.
Leaving these individuals out of our analyses would result in a downward biased estimate of the effect of the treatment (unemployment)
on child support compliance. To avoid such an outcome, we rely on
multiple imputation techniques to generate a complete dataset, which
takes into account uncertainty in the prediction of values of missing
variables (random noise) and allow for sampling error and hence population variation. To account for possibly unobserved heterogeneity among
the observations, most chapters in this volume use fixed-effects regressions, which unfortunately would not run with our imputed data. As
an alternative, we use random-effects (RE) regressions, which assume
that the individuals’ error term were not correlated with the predictors.
nonresident father involvement163
Under this assumption, our models retain the time-invariant variables.
The Hausman test on the non­imputed data tests whether the unique
errors (ui) are correlated with the regressors; the null hypothesis is they
are not. Results indicate that the null is rejected (Prob > chi2 = 0.3540)
and RE are better. We also use the Breusch-Pagan Lagrange Multiplier
to check whether a simple OLS is better than a RE regression. This test
indicates significant differences across individuals (panel effect), and an
RE regression would be better (Prob > chibar2 = 0.0000). The main drawback of this model is the possibility of omitted variable bias. We face two
types of possible biases: a selection bias from missing fathers (on nonimputed data) or an omitted variable bias when using random effects. We
believe the latter is lesser of the two evils, especially in view of the fact that
the local unemployment rate is exogenous to the individual.
As a robustness check, we use a two-step Heckman selection correction
model to account for a possible selection bias arising from fathers’ transitions into nonresidency at each wave. We do not find significant evidence
of selection into nonresidence on models that measure informal or in-kind
child support, or any of the visitation outcomes. We find evidence only of
positive selection into nonresidency in the case of formal child support. In
this case, results from the fixed-effects model and the selection-corrected
model are almost identical. Thus our results are based on our randomeffects model.
To account for missing data, we use in all models a multiple imputation (MI) technique that creates an algorithm consisting of chained
iterations. In this analysis, M was set to 40, based on the optimization
of largest fraction of missing information that show the goodness of fit
of an MI model.
To predict the effect of the Great Recession, we estimate the predicted
probability of full child support compliance and visitation at wave 5
(the Great Recession wave) assuming the unemployment rate at father’s
metro­politan area of residence was 5 percent and compare that with the
predictions when the unemployment rate was 10 percent. Controls for
father’s age, race-ethnicity, father’s immigrant status, mother’s preferences on father’s involvement in child’s upbringing, whether the father
visited mother at the hospital at childbirth, father’s employment status at
baseline, father’s incarceration status at baseline, the gender and age of
the focal child, and year and city fixed effects are included.
We conduct these analyses stratifying by fathers’ education. Because the
sample size for college-educated fathers is too small, we combine them with
the some postsecondary education group, yielding three distinct groups:
less than high school, high school, and greater than high school education (see table 6.A2). Analyses stratified by race-ethnicity are reported in
table 6.A4.
Unemployment rate
Education
High school
diploma
or GED
Some college
College or more
Father’s age
Race-ethnicity
Black
Hispanic
Other
Immigrant
Father working at
baseline
Father incarcerated
at baseline
Father visited mother
at hospital at birth
Mother wants father
involved in child’s
life
Child is a boy
Child’s age
Interview year
2000
2001
2002
2003
2004
1,102.462†
834.838†
1,069.775*
1,227.786*
1,365.337**
78.808
-1.776
222.048
287.043**
-125.858
(214.017)
(155.537)
(165.671)
(159.538)
(167.743)
(72.509)
(8.754)
(147.820)
(89.341)
(279.072)
(123.905)
(162.010)
(267.756)
(134.539)
(82.863)
(106.701)
(214.283)
(5.427)
189.925†
422.031*
-1.371
52.626
49.924
85.379
27.267
205.791*
(89.787)
(27.841)
52.036
-28.374
Informal Support ($)
(611.478)
39.588
(503.882)
1.519
(517.190) -239.732
(522.906) -143.417
(523.584)
-82.673
(146.169)
(19.527)
116.076
26.792
(610.280)
-102.146
(374.574)
(248.830)
(320.514)
(551.623)
(301.317)
(169.155)
-1,034.714***
-318.475
-894.627
-22.040
603.068***
416.382
(205.984)
(482.163)
(11.394)
671.730**
3,100.268***
42.093***
(190.662)
(177.230)
334.324†
413.793*
(42.779)
-105.786*
Formal Support ($)
Table 6.A1 Full Regression Results, Child Support and Visitation
-0.007
-0.086†
-0.197***
-0.212***
-0.181***
0.004
0.001
0.168***
0.221***
0.080
0.009
-0.061*
0.028
0.034
0.083***
0.061***
0.035
-0.001
0.023
-0.000
(0.061)
(0.044)
(0.050)
(0.049)
(0.049)
(0.013)
(0.002)
(0.029)
(0.016)
(0.050)
(0.023)
(0.027)
(0.047)
(0.025)
(0.015)
(0.018)
(0.040)
(0.001)
(0.016)
(0.008)
In-Kind Support (%)
-0.027
-0.100*
-0.179***
-0.206***
-0.189***
0.016
0.001
0.132***
0.228***
0.054
0.001
-0.068*
-0.013
0.045†
0.067***
0.041*
0.037
0.000
0.034*
0.000
(0.072)
(0.051)
(0.054)
(0.052)
(0.051)
(0.013)
(0.002)
(0.031)
(0.016)
(0.052)
(0.025)
(0.027)
(0.042)
(0.025)
(0.016)
(0.020)
(0.041)
(0.001)
(0.016)
(0.006)
Visited Child at
Least Once in a
Month (%)
0.159
-1.269
-3.205**
-4.039***
-3.947***
0.123
0.019
1.863**
3.499***
0.028
0.958*
-0.206
-0.340
0.440
0.712*
0.715†
0.714
0.025
0.400
0.088
(1.428)
(0.979)
(1.023)
(0.998)
(1.010)
(0.276)
(0.033)
(0.588)
(0.352)
(1.046)
(0.483)
(0.560)
(0.842)
(0.505)
(0.315)
(0.401)
(0.847)
(0.021)
(0.322)
(0.102)
Visitation Days
in a Month
77.934
-193.327
-10.339
254.297
-145.142
-85.087
-36.411
-133.686
-181.168
-116.062
273.718
305.899
-323.019
87.072
50.755
34.456
-156.824
-339.945
138.471
4,068
1,678
1,411.739**
2,310.491**
1,520.125*
1,692.102**
1,798.178**
2,097.808**
(402.413)
(350.910)
(348.506)
(361.561)
(372.341)
(487.812)
(424.041)
(347.903)
(329.456)
(456.762)
(509.043)
(505.603)
(444.805)
(535.984)
(456.268)
(557.841)
(524.408)
(439.780)
(1,074.857)
59.886
169.426
32.650
296.350†
68.065
-148.804
294.915†
117.281
-126.829
307.224*
219.914
157.174
264.978
104.238
0.978
-246.391
-120.042
-93.464
-35.449
9,206
3,353
(520.231)
-55.321
(889.802)
366.420
(630.981)
120.470
(530.494)
-84.422
(568.166) -151.810
(645.335) -148.872
(146.896)
(134.627)
(130.840)
(152.260)
(136.579)
(210.997)
(155.320)
(138.939)
(146.668)
(155.411)
(152.454)
(190.258)
(192.258)
(183.857)
(187.441)
(254.021)
(217.213)
(208.587)
(380.128)
(160.673)
(342.080)
(303.254)
(178.650)
(191.796)
(245.967)
0.016
0.067*
0.058†
0.069†
0.038
0.024
0.075†
0.095**
0.072†
0.059
0.057
0.029
0.052
0.046
0.014
0.001
0.069
0.056
0.327**
9,206
3,353
-0.202***
-0.142
-0.144*
-0.244***
-0.298***
-0.305***
(0.036)
(0.032)
(0.033)
(0.039)
(0.034)
(0.054)
(0.039)
(0.033)
(0.039)
(0.037)
(0.038)
(0.047)
(0.047)
(0.047)
(0.049)
(0.068)
(0.055)
(0.054)
(0.102)
(0.047)
(0.094)
(0.069)
(0.052)
(0.060)
(0.073)
0.013
0.090*
0.081*
0.111**
0.068†
0.046
0.081*
0.126***
0.084*
0.076*
0.084†
0.043
0.088†
0.043
0.026
0.066
0.137*
0.101†
0.009
8,965
3,157
-0.215***
-0.139
-0.113
-0.246***
-0.297***
-0.319***
(0.036)
(0.036)
(0.038)
(0.036)
(0.036)
(0.037)
(0.040)
(0.038)
(0.036)
(0.038)
(0.043)
(0.053)
(0.051)
(0.050)
(0.052)
(0.053)
(0.058)
(0.052)
(0.052)
(0.051)
(0.099)
(0.071)
(0.055)
(0.062)
(0.073)
0.257
1.411†
1.926*
1.975**
1.807*
0.652
1.695*
2.401**
1.514*
1.534†
1.222
0.975
1.374
0.734
0.614
1.182
0.890
1.563
0.388
8,965
3,157
-4.402***
-4.299*
-4.266**
-5.550***
-6.318***
-6.835***
(0.792)
(0.755)
(0.786)
(0.760)
(0.789)
(0.786)
(0.833)
(0.825)
(0.762)
(0.822)
(0.865)
(1.108)
(1.105)
(1.049)
(1.102)
(1.132)
(1.218)
(1.126)
(1.112)
(1.018)
(1.890)
(1.397)
(1.046)
(1.112)
(1.369)
Source: Authors’ calculations using Fragile Families and Child Wellbeing Study data circa 1998 to 2010.
Note: Standard errors in parentheses. Covariates are measured at the baseline survey (except year) and are clustered at city and individual level. Model 1 includes level unemployment rate. The model with individual random effects is clustered at city level.
***p < .001; **p < .01; *p < .05; †p < .1
2005
2006
2007
2008
2009
2010
City
Austin
Baltimore
Detroit
Newark
Philadelphia
Richmond
Corpus Christi
Indianapolis
Milwaukee
New York
San Jose
Boston
Nashville
Chicago
Jacksonville
Toledo
San Antonio
Pittsburgh
Norfolk
Constant
Observations
Number of
individuals
166
children of the great recession
Table 6.A2 Coefficients and Standard Errors, Rate of Change,
Father Involvement
Formal child support
Unemployment rate
(model 1)
Unemployment rate
(model 2)
Increasing unemployment
Decreasing unemployment
Observations
Number of individuals
Informal child support
Unemployment rate
(model 1)
Unemployment rate
(model 2)
Increasing unemployment
rate
Decreasing unemployment
rate
Observations
Number of individuals
In-kind child support
Unemployment rate
(model 1)
Unemployment rate
(model 2)
Increasing unemployment
rate
Decreasing unemployment
rate
All
Less than
High
School
High
School
Some
College and
College +
-105.786*
(42.779)
-101.144*
(44.039)
-2.356
(2.378)
-7.309
(11.825)
4,068
1,678
-16.853
(29.482)
-22.849
(30.946)
-0.204
(1.451)
-6.349
(7.606)
3,395
1,285
-168.682**
(60.606)
-175.896**
(64.206)
-1.471
(3.893)
-15.518
(17.621)
1,612
683
-152.260†
(87.879)
-143.482
(93.080)
-0.780
(5.271)
6.423
(26.787)
869
370
-28.374
(27.841)
-10.017
(30.459)
-1.049
(1.504)
12.637
(7.895)
4,068
1,678
-29.800
(43.363)
-6.440
(44.910)
-1.710
(2.028)
13.392
(10.552)
3,395
1,285
-21.246
(42.686)
-8.119
(46.500)
-1.025
(2.407)
8.494
(11.803)
1,612
683
-24.683
(59.046)
-10.400
(60.059)
-0.395
(3.409)
11.895
(17.238)
869
370
-0.000
(0.008)
0.001
(0.008)
-0.000
(0.000)
0.000
(0.001)
0.001
(0.010)
0.002
(0.012)
-0.000
(0.000)
-0.000
(0.002)
-0.004
(0.010)
-0.002
(0.011)
-0.000
(0.000)
-0.000
(0.002)
0.002
(0.013)
0.005
(0.013)
-0.000
(0.001)
0.001
(0.003)
nonresident father involvement167
Table 6.A2 Continued
All
Observations
Number of individuals
Visited child at least once in
the last month
Unemployment rate
(model 1)
Unemployment rate
(model 2)
Increasing unemployment
rate
Decreasing unemployment
rate
Observations
Number of individuals
Visitation days in a month
Unemployment rate
(model 1)
Unemployment rate
(model 2)
Increasing unemployment
rate
Decreasing unemployment
rate
Observations
Number of individuals
Less than
High
School
High
School
Some
College and
College +
4,068
1,678
3,395
1,285
1,612
683
869
370
0.000
(0.006)
0.001
(0.006)
-0.001
(0.000)
-0.001
(0.002)
8,965
3,157
0.001
(0.009)
0.003
(0.009)
-0.001
(0.001)
-0.000
(0.003)
3,248
1,192
-0.002
(0.009)
-0.003
(0.009)
-0.000
(0.001)
-0.003
(0.002)
3,425
1,227
0.002
(0.011)
0.004
(0.011)
-0.001
(0.001)
0.000
(0.003)
1,829
708
0.088
(0.102)
0.112
(0.104)
-0.012*
(0.005)
-0.010
(0.025)
8,965
3,157
0.085
(0.169)
0.128
(0.173)
-0.015†
(0.008)
-0.001
(0.039)
3,248
1,192
0.097
(0.156)
0.091
(0.156)
-0.009
(0.008)
-0.036
(0.042)
3,425
1,227
0.067
(0.209)
0.117
(0.220)
-0.011
(0.011)
0.023
(0.059)
1,829
708
Source: Authors’ calculations using Fragile Families and Child Wellbeing Study data circa 1998 to
2010.
Note: Standard errors in parentheses. Model 1 includes the unemployment rate as a level. Model 2
includes unemployment rate as a level as well as rate of change in unemployment rate. SEs are clustered
at city level.
**p < 0.01, *p < 0.05, †p < 0.1
168
children of the great recession
Table 6.A3 Sensitivity of Coefficients, Child Support
and Visitation Outcomes
Models
Formal child support
Unemployment rate (model 1)
Unemployment rate (model 3)
Mother’s unemployment
Bio-social fathers not employed
Unemployment rate (model 4)
Unemployment rate * year nine
Informal child support
Unemployment rate (model 1)
Unemployment rate (model 3)
Mother’s unemployment
Bio-social father’s not employed
Unemployment rate (model 4)
Unemployment rate * year nine
In-kind child support
Unemployment rate (model 1)
Unemployment rate (model 3)
Mother’s unemployment
Bio-social fathers not employed
Unemployment rate (model 4)
Unemployment rate * year nine
Visited child at least once in the last month
Unemployment rate (model 1)
Unemployment rate (model 3)
Mother’s unemployment
Bio-social fathers not employed
Unemployment rate (model 4)
Unemployment rate * year nine
Visitation days in a month
Unemployment rate (model 1)
Unemployment rate (model 3)
Mother’s unemployment
Bio-social fathers not employed
Unemployment rate (model 4)
Unemployment rate * year nine
Coefficients
Standard
Errors
-105.786*
-87.790*
-189.256
-563.707***
-106.541†
-2.170
(42.779)
(42.582)
(115.192)
(109.329)
(54.912)
(56.114)
-28.374
-23.971
-8.955
-182.413**
-34.514
22.703
(27.841)
(27.809)
(73.953)
(66.977)
(31.493)
(36.307)
-0.000
0.001
-0.003
-0.066***
-0.005
0.012†
(0.008)
(0.008)
(0.014)
(0.013)
(0.009)
(0.007)
0.000
0.002
-0.005
-0.079***
-0.004
0.011†
(0.006)
(0.006)
(0.014)
(0.013)
(0.007)
(0.007)
0.088
0.120
-0.119
-1.334***
0.042
0.132
(0.102)
(0.102)
(0.254)
(0.262)
(0.122)
(0.123)
Source: Authors’ calculations using Fragile Families and Child Wellbeing Study data circa 1998 to
2010.
Note: Standard errors in parentheses. Model 3 includes unemployment rate and a measure of individual
unemployment. Model 4 includes unemployment rate and an interaction between unemployment rate
and year nine, when the Great Recession hit. SEs are clustered at city level.
***p < .001; **p < .01; *p < .05; †p < .1
nonresident father involvement169
Table 6.A4 Coefficients and Standard Errors, Model 1,
Child Support and Visitation Outcomes
Formal child support
Unemployment rate
Informal child support
Unemployment rate
In-kind child support
Unemployment rate
Visited child at least once
in the last month
Unemployment rate
Visitation days in a month
Unemployment rate
Black
Hispanic
White
-81.399†
(47.229)
-114.297**
(42.379)
-108.447
(153.350)
-36.303
(32.780)
-31.333
(54.459)
-3.935
(78.532)
-0.002
(0.008)
-0.022
(0.012)
0.001
(0.016)
0.002
(0.007)
-0.005
(0.009)
0.006
(0.015)
0.077
(0.131)
0.009
(0.183)
0.247
(0.247)
Source: Authors’ calculations using Fragile Families and Child Wellbeing Study data circa 1998 to
2010.
Note: Standard errors in parentheses. Model 1 includes level unemployment rate; results include individual random-effects and time and city fixed effects. SEs are clustered at city level.
**p < .01; *p < .05; †p < .1
NOTES
1. Amato, Meyers, and Emory 2009; Cheadle, Amato, and King 2010.
2. Beller and Graham 1986.
3. Henry 1999; Hatcher and Lieberman 2003.
4. Paul and Moser 2009.
5. Bowman 1990; Bowman and Forman 1997; McAdoo 1993.
6. Anderson 2000.
7. Achatz and McAllum 1994; Furstenberg and Hughes 1995; Johnson,
Levine, and Doolittle 1999.
8. Edin and Lein 1997; Anderson 1993.
9. Furstenberg et al. 1983; Mott 1990; Lerman 1993.
10. Weis and Willis 1985; Graham and Beller 2002.
11. Anderson 1993; Del Boca and Ribero 2001; Furstenberg 1988. Although a
reciprocal relationship may also exist between visitation and child support,
this is much more likely in the case of informal support and supported by
some empirical evidence (Nepomnyaschy 2007).
12. Seltzer, McLanahan, and Hanson 1998.
170
children of the great recession
13. Cancian and Meyer 2006.
14. Bartfield and Meyer 2003; Meyer, Ha, and Hu 2008; Ha et al. 2008.
15. Wu 2011.
16. Ha, Cancian, and Meyer 2011; Huang 2010; Huang and Edwards 2009;
Nepomnyaschy and Garfinkel 2010; Sorensen and Hill 2004; Meyer, Ha,
and Hu 2008; Bartfield and Meyer 2003.
17. Meyer, Ha, and Hu 2008.
18. Ha et al. 2008; Wu 2011.
19. Mincy, Miller, and De la Cruz Toledo 2016.
20. Nepomnyaschy and Garfinkel 2010.
21. Mott 1990; Lerman 1993; Danziger and Radin 1990; Coley and ChaseLandsdale 1999.
22. Veum 1993; Seltzer, McLanahan, and Hanson 1998.
23. Specifically, we control for fathers’ age, race-ethnicity, relationship status at
birth, immigrant status, the gender of the focal child, father’s employment at
baseline, incarceration at baseline, father’s visitation at baseline and mother’s
preference for father involvement and year and city fixed effects (for why we
used random rather than fixed-effects estimates, see the appendix).
REFERENCES
Achatz, Mary, and Crystal A. MacAllum. 1994. Young Unwed Fathers: Report
from the Field. Philadelphia, Pa.: Public/Private Ventures.
Amato, Paul R., Catherine E. Meyers, and Robert E. Emery. 2009. “Changes
in Nonresident Father-Child Contact From 1976 to 2002.” Family Relations
58(1): 41–53.
Anderson, Elijah. 1993. “Sex Codes and Family Life Among Poor Inner-City
Youth.” In The Ghetto Underclass: Social Science Perspectives, edited by William
Julius Wilson. Newberry Park, Calif.: Sage Publications.
———. 2000. Code of the Street: Decency, Violence, and the Moral Life of the Inner
City. New York: W. W. Norton.
Bartfeld, Judi, and Daniel R. Meyer. 2003. “Child Support Compliance Among
Discretionary and Nondiscretionary Obligors.” Social Service Review 77(3):
347–72.
Beller, Andrea H., and John W. Graham. 1986. “Child Support Awards: Differ­
entials and Trends by Race and Marital Status.” Demography 23(2): 231–45.
Bowman, Philip J. 1990. “Coping With Provider Role Strain: Adaptive Cultural
Resources Among Black Husband-Fathers.” Journal of Black Psychology 16(2):
1–21.
Bowman, Philip J., and Tyrone A. Forman. 1997. “Instrumental and Expressive
Family Roles Among African American Fathers.” In Family Life in Black
America, edited by Robert J. Taylor, James S. Jackson, and Linda M. Chatters.
New York: Sage Publications.
Cancian, Maria, and Daniel R. Meyer. 2006. “Alternative Approaches to Child
Support Policy in the Context of Multiple-Partner Fertility.” Report to the
Wisconsin Department of Workforce Development.
nonresident father involvement171
Cheadle, Jacob E., Paul R. Amato, and Valarie King. 2010. “Patterns of Non­
resident Father Contact.” Demography 47(1): 205–25.
Coley, Rebekah L., and P. Lindsay Chase-Lansdale. 1999. “Stability and Change
in Paternal Involvement Among Urban African American Fathers.” Journal of
Family Psychology 13(3): 416–35.
Danziger, Sandra K., and Norma Radin. 1990. “Absent Does Not Equal
Uninvolved: Predictors of Fathering in Teen Mother Families.” Journal of
Marriage and the Family 52(3): 636–42.
Del Boca, Daniela, and Rocio Ribero. 2001. “The Effect of Child-Support
Policies on Visitations and Transfers.” American Economic Review 91(2):
130–34.
Edin, Kathryn, and Laura Lein. Making Ends Meet: How Single Mothers Survive
Welfare and Low-Wage Work. New York: Russell Sage Foundation.
Furstenberg, Frank F., Jr. 1988. “Good Dads—Bad Dads: Two Faces of
Fatherhood.” In The Changing American Family and Public Policy, edited by
Andrew J. Cherlin. Washington, D.C.: Urban Institute Press.
Furstenberg, Frank F., Jr., and Mary Elizabeth Hughes. 1995. “Social Capital
and Successful Development Among At-Risk Youth.” Journal of Marriage and
Family 57(3): 580–92.
Furstenberg, Frank F., Jr., Christine W. Nord, James L. Peterson, and Nicholas
Zill. 1983. “The Life Course of Children of Divorce: Marital Disruption and
Parental Contact.” American Sociological Review 48(5): 656–68.
Graham, John W., and Andrea H. Beller. 2002. “Nonresident Fathers and Their
Children: Child Support and Visitation from an Economic Perspective.” In
Handbook of Father Involvement: Multidisciplinary Perspectives, edited by
Natasha J. Cabrera and Catherine S. Tamis-LeMonda. London: Routledge.
Ha, Yoonsook, Maria Cancian, and Daniel R. Meyer. 2011. “The Regularity
of Child Support and Its Contribution to the Regularity of Income.” Social
Service Review 85(3): 401–19.
Ha, Yoonsook, Maria Cancian, Daniel R. Meyer, and Eunhee Han. 2008. “Factors
Associated with Nonpayment of Child Support.” Madison: University of
Wisconsin–Madison, Institute for Research on Poverty.
Hatcher, Daniel L., and Hannah Lieberman. 2003. “Breaking the Cycle of Defeat
for ‘Deadbroke’ Noncustodial Parents Through Advocacy on Child Support
Issues.” Clearinghouse Review 37(1–2)(May–June): 5–21.
Henry, Ronald K. 1999. “Child Support at a Crossroads: When the Real World
Intrudes Upon Academics and Advocates.” Family Law Quarterly 33(1):
235–64.
Huang, Chien-Chung. 2010. “Trends in Child Support from 1994 to 2004:
Does Child Support Enforcement Work?” Journal of Policy Practice 9(1):
36–53.
Huang, Chien-Chung, and Richard L. Edwards. 2009. “The Relationship
Between State Efforts and Child Support Performance.” Children and Youth
Services Review 31(2): 243–48.
Johnson, Earl S., Ann Levine, and Fred Doolittle. 1999. Fathers’ Fair Share:
Helping Poor Men Manage Child Support and Fatherhood. New York: Russell
Sage Foundation.
Lerman, Robert I. 1993. “A National Profile of Young Unwed Fathers.” In Young
Unwed Fathers: Changing Roles and Emerging Policies, edited by Robert I.
Lerman and Theodora J. Ooms. Philadelphia, Pa.: Temple University Press.
172
children of the great recession
McAdoo, John L. 1993. “The Roles of African American Fathers: An Ecological
Perspective.” Families in Society 74(1): 28–35.
Meyer, Daniel R., Yoonsook Ha, and Mei-Chen Hu. 2008. “Do High Child
Support Orders Discourage Child Support Payments?” Social Service Review
82(1): 93–118.
Mincy, Ronald, Daniel P. Miller, and Elia De la Cruz Toledo. Forthcoming.
“Child Support Compliance During Economic Downturns.” Children and
Youth Services Review.
Mott, Frank L. 1990. “When Is a Father Really Gone? Paternal-Child Contact in
Father-Absent Homes.” Demography 27(4): 499–517.
Nepomnyaschy, Lenna. 2007. “Child Support and Father-Child Contact: Testing
Reciprocal Pathways.” Demography 44(1): 93–112.
Nepomnyaschy, Lenna, and Irwin Garfinkel. 2010. “Child Support Enforcement
and Fathers’ Contributions to Their Nonmarital Children.” Social Service
Review 84(3): 341–80.
Paul, Karsten I., and Klaus Moser. 2009. “Unemployment Impairs Mental
Health: Meta-Analyses.” Journal of Vocational Behavior 74(3): 264–82.
Seltzer, Judith A., Sara S. McLanahan, and Thomas L. Hanson. 1998. “Will Child
Support Enforcement Increase Father-Child Contact and Parental Conflict
After Separation.” In Fathers Under Fire: The Revolution in Child Support
Enforcement, edited by Irwin Garfinkel, Sara S. McLanahan, Daniel R. Meyer,
and Judith A. Seltzer. New York: Russell Sage Foundation.
Sorensen, Elaine, and Ariel Hill. 2004. “Single Mothers and Their Child-Support
Receipt: How Well Is Child-Support Enforcement Doing?” Journal of Human
Resources 39(1): 135–54.
Veum, Jonathan R. 1993. “The Relationship Between Child Support and
Visitation: Evidence from Longitudinal Data.” Social Science Research 22(3):
229–44.
Weiss, Yoram, and Robert J. Willis. 1985. “Children as Collective Goods and
Divorce Settlements.” Journal of Labor Economics 3(3): 268–92.
Wu, Chi-Fang. 2011. “Child Support in an Economic Downturn: Changes in
Earnings, Child Support Orders, and Payments.” Working paper. Madison:
University of Wisconsin.
Chapter 7
Mothers’ and Fathers’ Parenting
William Schneider, Jane Waldfogel,
and Jeanne Brooks-Gunn
I
n previous chapters, we see that recessions take an economic toll on families. They also lead to reductions in parents’ health, relationship quality,
and contributions from nonresident fathers. In this chapter, we turn to the
question of whether recessions also affect the experiences of children in
their homes, as measured by the parenting provided by their mothers and
fathers when the children are one, three, five, and nine years of age.
We ask how a large change in the unemployment rate, one similar to that
of the Great Recession, affects mothers’ and fathers’ parenting. We look
at three aspects of mothers’ parenting—use of harsh parenting, expression
of warmth toward the child, and participation in cognitively stimulating
activities with the child. Fathers were asked about use of harsh parenting.
Most of our parenting measures were first observed when children were
three years old (spanking, an indicator of harsh parenting, is an exception
and was first measured at age one).
An important consideration for our chapter is that we analyze the parenting of children who are experiencing dramatic developmental changes
during the study and that interactions between parents and children change
a great deal as children age. Parents seek activities and disciplinary strategies that are appropriate for their children’s age as well as their cognitive,
emotional, and social capacities. The amount of time parents spend with
school-age children, and the types of games and activities in which they
engage, are different from the amount of time and kinds of activities parents and young children might do together. For example, spanking is much
more common with younger than older children, given the increases in
self-regulation that occur over the childhood years. Expressions of warmth
tend to decline as children mature. Cognitively stimulating activities that
parents engage in with a three- or five-year-old child, like playing with
blocks or telling stories, are replaced by activities like helping with homework, talking about current events in the child’s life, and watching television together by the time a child is nine years old. A nine-year-old child
might arrive home from school, and together parent and child might work
on the child’s homework or discuss the events of the day. A three-yearold, in contrast, may have spent the day in daycare or with the parent, and
174
children of the great recession
their interactions may focus more on care and age-appropriate play. Thus
the specific parenting activities that parents are asked about in the Fragile
Families Study vary according to their appropriateness to the age of child.
We analyze both mothers’ and fathers’ parenting. Mothers and fathers
exhibit similar parenting behaviors, even though the frequency with which
they engage in certain behaviors sometimes differs. Past studies indicate
that parents from different socioeconomic backgrounds also differ in the
amount of specific behaviors exhibited.1 Parents with more education
are likely to spend more time doing things like reading to their children
than their less-educated peers.2 Thus, as with other chapters in this volume,
we use mothers’ education as an indicator of families’ socioeconomic status
and ask whether more and less-educated parents respond to recessions in
different ways.
ECONOMIC HARDSHIP, UNCERTAINTY, AND PARENTING
Perhaps the most famous study of how parenting is affected by economic conditions is Glen H. Elder Jr.’s study of families during the Great
Depression.3 Elder, and in later work with his colleague Rand Conger,
found that individual-level unemployment and job loss was associated
with increased harsh parenting and more conflict between mothers and
fathers.4 In their studies, and in replications in different contexts, Elder
and Conger found that changes in parenting that result from economic
hardship and uncertainty have negative effects for child well-being.5
These results are consistent with a body of research linking poverty and
economic hardship to a range of parenting practices that may adversely
affect children. In particular, individual-level experiences of poverty and
economic hardship have been shown to be associated with increased punitive parenting behaviors as well as less warmth in parenting.6
How is parenting likely to be affected by a big recession? A number of
recent studies have used macroeconomic measures to assess the effect of
the Great Recession on parenting. These studies provide ample and growing evidence that the Great Recession was associated with increases in
harsh parenting. Worsening economic conditions (measured by increases
in local unemployment rates, foreclosure rates, or state-level mass layoffs) and declining consumer confidence (measured by changes in the
national Consumer Sentiment Index) have been linked to indicators of
harsh parenting such as increased physical and psychological aggression or
increased reports of possible child maltreatment.7 However, increases in
the unemployment rate during the Great Recession have not been related
to more reports to or investigations by Child Protective Service agencies.8
Evidence on other types of parenting is more limited. Studies of prior
recessions and the 1980s Iowa Farm Crisis indicate that individual-level
measures of economic hardship and uncertainty are associated with
reduced parenting warmth, consistent with Elder’s findings.9 This chapter
mothers’ and fathers’ parenting175
builds on this body of research by focusing primarily on macroeconomic
factors, not only on individual-level experiences of economic downturns
(job losses). In this way, our results are less likely than results from prior
studies to be driven by other individual differences between parents that
could affect both their likelihood of experiencing unemployment or other
economic shocks and their parenting.
TRENDS IN PARENTING: CHILDREN AGES THREE TO NINE
We begin by looking at some basic trends in parenting when children were
three, five, and nine years old. We focus on three aspects of parenting: harshness (spanking, physical aggression, and psychological aggression), warmth,
and activities with children. The questions on harshness were drawn from
the Conflict Tactics Scale, a well-established battery of questions designed
to gauge parents’ physical and psychological aggression toward their children.10 Both mothers and fathers were asked a series of questions about
their own and each other’s harsh parenting practices. A specific question on
spanking was also asked when children were one, three, five, and nine years
old. Child maltreatment is often thought of as occurring on a continuum
with maltreatment on one end and more widely accepted parental disciplinary practices on the other.11 The more frequently a parent uses harsh parenting, the greater the risk for child maltreatment. Parents were also asked
about the kinds of activities they did with their children. These questions
ranged from activities like reading books together and playing outside, to
watching television or playing video games, and varied depending on the
age of the child. Last, mothers were observed interacting with their child
in the home. The number of warm interactions, like using terms of endearment or cuddling, between mother and child were recorded.12 The sample
size for the warmth measure is somewhat smaller than those for harsh and
cognitively stimulating activities given that the former is observed in the
home rather than being based on parental self-report.
For each of our measures of parenting behavior we concentrate on
behaviors that occur frequently. We focus on high-frequency parenting
behaviors for two reasons. First, high-frequency harsh parenting is itself
a particular risk factor for children and is associated with problem behaviors.13 Second, we expect that deep recessions, like the Great Recession,
would be more likely to move parents toward increasing or decreasing
their current parenting behaviors, rather than exhibiting new ones.
Theory and empirical evidence would lead us to hypothesize that harsh
parenting, and particularly high-frequency harsh parenting, would decline
over time as children age. This might be for a number of reasons. First, as
children mature, they are better able to regulate their own behavior, and
parents are able to use other forms of discipline to influence them. Second,
as children move from early childhood to middle childhood and early adolescence they gain greater autonomy and parents perform less monitoring
176
children of the great recession
Mean
Figure 7.1 High-Frequency Maternal Spanking by Education
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
College +
Some college
High school
Less than
high school
1
3
5
(1999–2001) (2001–2003) (2003–2006)
9
(2007–2010)
Child’s Age-Year
Source: Authors’ calculations based on data from the Fragile Families and Child Wellbeing
Study.
and have fewer opportunities to administer corporal punishment or harsh
parenting. Third, as children develop physically, corporal punishment and
physically aggressive parenting become more difficult and are generally
used less frequently. Warmth is also expected to decline as children mature.
The trend in cognitively stimulating activities is more difficult to predict,
given that the activities measured are different at the various ages.
Mothers’ Parenting
As expected, high-frequency spanking and physically aggressive parenting both decrease as children age, while psychologically aggressive parenting is fairly consistently used across childhood (figures 7.1 through 7.3).
Although we do not have information about physical and psychological
aggression until children are three years old, we do have data about spanking at age one. High-frequency spanking is quite common at age one (about
20 percent of mothers exhibiting this behavior), and percentages are similar
at age three and five. High-frequency spanking is most common among
the high school–educated mothers at age one, about 25 percent spanking
frequently, and among the less than high school educated at age nine, about
20 percent spanking frequently. In contrast, psychologically aggressive parenting is most common among the college educated, about half of that
group reporting using psychologically aggressive parenting frequently.
The next set of figures show trends in mothers’ warmth and activities with
children. We might reasonably expect high-frequency maternal warmth to
mothers’ and fathers’ parenting177
Figure 7.2 High-Frequency Maternal Physical Aggression by Education
1.0
0.9
College +
0.8
Mean
0.7
Some college
0.6
0.5
High school
0.4
0.3
Less than
high school
0.2
0.1
0
1
3
5
(1999–2001) (2001–2003) (2003–2006)
9
(2007–2010)
Child’s Age-Year
Source: Authors’ calculations based on data from the Fragile Families and Child Wellbeing
Study.
Mean
Figure 7.3 High-Frequency Maternal Psychological Aggression
by Education
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
College +
Some college
High school
Less than
high school
1
3
5
(1999–2001) (2001–2003) (2003–2006)
9
(2007–2010)
Child’s Age-Year
Source: Authors’ calculations based on data from the Fragile Families and Child Wellbeing
Study.
178
children of the great recession
Mean
Figure 7.4 High-Frequency Maternal Warmth by Education
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
College +
Some college
High school
Less than
high school
1
3
5
(1999–2001) (2001–2003) (2003–2006)
9
(2007–2010)
Child’s Age-Year
Source: Authors’ calculations based on data from the Fragile Families and Child Wellbeing
Study.
change over time as the children in our study move from being three-yearolds dependent on their parents and greatly in need of warm parenting,
to nine-year-olds at the beginning of greater individuation. Figure 7.4
demonstrates that observed high-frequency maternal warmth declines as
children age. It also demonstrates a clear education gradient in maternal
warmth, the more-educated mothers showing more warmth at all ages.
Last, we turn to trends in maternal activities with children. The types of
activities that parents do with children change over time as children age.
Playing with blocks is an age-appropriate activity for a three-year-old; by age
nine, however, children and parents are engaged in more complicated and
involved interactions. For this reason, questions about parenting activities
were designed to be developmentally appropriate, meaning that parents
were asked about different activities when children were three, five, and
nine years old. In addition, parents understandably spend less time with
their nine-year-old child who attends school during the day, than they do
with a three-year-old who may be at home with a parent most of the time.
Figure 7.5 demonstrates a decline in the frequency of parenting activities
as children age. Mothers with at least some college education are more
involved in parenting activities up to age five, and step away from those
activities from age five to nine more rapidly than less-educated mothers do.
Fathers’ Parenting
We also examine trends in fathers’ harsh parenting over time. For spanking,
we draw on information about fathers who recently had contact with their
child, whether they lived in the home or not. For high-frequency physical
mothers’ and fathers’ parenting179
Mean
Figure 7.5 High-Frequency Maternal Parenting Activities by Education
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
College +
Some college
High school
Less than
high school
1
3
5
(1999–2001) (2001–2003) (2003–2006)
9
(2007–2010)
Child’s Age-Year
Source: Authors’ calculations based on data from the Fragile Families and Child Wellbeing
Study.
and psychological aggression we are restricted to fathers who lived in the
house with the child. Figure 7.6 shows trends in high-frequency spanking by fathers beginning when children are age one. Approximately 10 to
20 percent of fathers frequently spanked their child when the child was
one year old. This percentage increases sharply—to roughly 30 percent to
70 percent—when children are three and five years old, and then falls to
Mean
Figure 7.6 High-Frequency Paternal Spanking by Education
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
College +
Some college
High school
Less than
high school
1
3
5
(1999–2001) (2001–2003) (2003–2006)
9
(2007–2010)
Child’s Age-Year
Source: Authors’ calculations based on data from the Fragile Families and Child Wellbeing
Study.
180
children of the great recession
Mean
Figure 7.7 High-Frequency Paternal Physical Aggression by Education
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
College +
Some college
High school
Less than
high school
1
3
5
(1999–2001) (2001–2003) (2003–2006)
9
(2007–2010)
Child’s Age-Year
Source: Authors’ calculations based on data from the Fragile Families and Child Wellbeing
Study.
near zero at age nine. Thus, as with mothers, fathers spank less frequently
when children are age nine, but much more frequently than mothers at
ages three and five.
As well as spanking, we also examine trends in physically and psychologically aggressive parenting among fathers (see figures 7.7 and 7.8).
Overall, physically aggressive parenting is quite low across all ages and
generally decreases after age three. In contrast, psychologically aggressive parenting is much more stable over time; fathers with at least some
college education use psychological aggression more often than their lesseducated counterparts.
LOCAL UNEMPLOYMENT RATES AND PARENTING
As in previous chapters, we use pooled data from surveys conducted up to
age nine to estimate the effects of unemployment on parenting, with and
without individual fixed effects. As mentioned, we have data at ages one,
three, five, and nine for spanking, and at ages three, five, and nine for our
other measures. The fixed-effects model does a better job of accounting for
unobservable differences between parents and is therefore our preferred
model. The models control for a host of demographic characteristics,
including the mother’s age, race-ethnicity, parents’ relationship status at
birth, immigrant status, whether the mother grew up with both parents,
calendar year, and family’s city of residence. We use the results from our
models to simulate the difference in parenting when the unemployment
rate is 5 percent and 10 percent, which is approximately the size of the
mothers’ and fathers’ parenting181
Mean
Figure 7.8 High-Frequency Paternal Psychological Aggression
by Education
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
College +
Some college
High school
Less than
high school
1
3
5
(1999–2001) (2001–2003) (2003–2006)
9
(2007–2010)
Child’s Age-Year
Source: Authors’ calculations based on data from the Fragile Families and Child Wellbeing
Study.
increase brought on by the Great Recession. These estimates are presented
in figures 7.9 to 7.13. Table 7.A1 presents results for high-frequency
maternal spanking, with and without individual fixed effects. Table 7.A2
presents the results for the five maternal parenting measures, and for each
maternal education group.
Mothers’ Parenting
Figures 7.9 through 7.13 display the estimated effects of a big increase in
the unemployment rate on mothers’ parenting. As shown in figure 7.9,
higher unemployment is predicted to decrease the likelihood of frequent
spanking by 44 percent. When mothers are disaggregated by education,
this reduction is largest (at 53 percent) and statistically significant for
mothers with less than a high school education.
Higher unemployment is also predicted to decrease high-frequency
physical aggression, by about 52 percent in the overall sample (figure 7.10).
When mothers are analyzed separately by level of education, the reduction is marginally statistically significant for mothers with less than a high
school education (a 55 percent reduction) and for those with a high school
education (a 40 percent reduction). The predicted reduction for the collegeeducated group is significant and large, but given the relatively small sample
size for this group, this result may not be reliable.
Results for psychological aggression (figure 7.11) are mostly not significant. We do find, however, that higher unemployment is predicted to
182
children of the great recession
Percent
Figure 7.9 High-Frequency Maternal Spanking by Unemployment Rate
80
70
60
50
40
30
20
10
0
UR 5 percent
–44%
–53%
–22%
–24%
All
Less than
high school
High
school*
Some
college
UR 10 percent
0%
College +
Source: Authors’ calculations based on data from the Fragile Families and Child Wellbeing
Study.
Note: Chow tests show that the coefficient for unemployment for high school is different
from the coefficient for unemployment for the less than high school group. The coefficients for the all and college or more categories were significant in the individual fixed
effects models.
*p < .05
decrease the likelihood of frequent psychological aggression by 30 percent among mothers with some college education and by a nonsignificant
18 percent among mothers with a college degree or more. The pattern is
different for mothers with less education, for whom higher unemployment
is associated with a nonsignificant increase in the frequency of psychological
aggression.
Percent
Figure 7.10 High-Frequency Maternal Physical Aggression
by Unemployment Rate
80
70
60
50
40
30
20
10
0
UR 5 percent
–52%
–55%
–40%
–4%
All
Less than
high school
High
school
Some
college
–95%
UR 10 percent
College +
Source: Authors’ calculations based on data from the Fragile Families and Child Wellbeing
Study.
Note: No significant differences in the effect of unemployment among subgroups. The
coefficients for the All, less than high school, high school, and college or more categories were significant in the individual fixed-effects models.
mothers’ and fathers’ parenting183
Percent
Figure 7.11 High-Frequency Maternal Psychological Aggression
by Unemployment Rate
80
70
60
50
40
30
20
10
0
+18%
–4%
+8%
–30%
–18%
UR 5 percent
UR 10 percent
All
Less than
high school
High
school
Some
college**
College +
Source: Authors’ calculations based on data from the Fragile Families and Child Wellbeing
Study.
Note: Chow tests show that the coefficient for unemployment for some college is different
from the coefficient for unemployment for the less than high shool group. The coefficient
for some college was significant in the individual fixed-effects models.
**p < .01
Finally, we do not find any significant effects of higher unemployment
rates on mothers’ high-frequency warmth (figure 7.12) either for mothers
overall or for mothers analyzed separately by education level. Figure 7.13
depicts the effects of higher unemployment rates on mothers’ high-frequency
parenting activities; again, we do not find any significant effects. In additional analyses that looked at white, black, and Hispanic mothers separately,
Figure 7.12 High-Frequency Maternal Warmth by Unemployment Rate
–12%
80
70
Percent
60
–2%
50
+5%
–3%
+2%
UR 5 percent
40
UR 10 percent
30
20
10
0
All
Less than
high school
High
school
Some
college
College +
Source: Authors’ calculations based on data from the Fragile Families and Child Wellbeing
Study.
Note: No significant differences in the effect of unemployment among subgroups.
184
children of the great recession
Percent
Figure 7.13 High-Frequency Maternal Parenting Activities
by Unemployment Rate
80
70
60
50
40
30
20
10
0
+1%
0%
–3%
+12%
–8%
UR 5 percent
UR 10 percent
All
Less than
high school
High
school
Some
college
College +
Source: Authors’ calculations based on data from the Fragile Families and Child Wellbeing
Study.
Note: No significant differences in the effect of unemployment among subgroups.
we found that high unemployment was associated with decreases in warmth
among black and white mothers and increases in warmth among Hispanic
mothers (see table 7.A4). These effects offset one another, resulting in a null
effect for the combined sample.
In sum, the results suggest that high unemployment rates reduce the
likelihood of high-frequency spanking and physically aggressive parenting among all mothers. They also increase the likelihood of maternal
warmth among Hispanic mothers. In contrast, high unemployment rates
reduce maternal warmth among black and white mothers. Effects of local
un­employment rates on parenting activities are not significant.
In addition to looking at the effects of high unemployment rates, we also
explored whether rapidly changing unemployment rates might affect parenting behavior. As noted in the introduction to this volume, the meaning
of a given level of unemployment may differ greatly depending on whether
it represents the status quo, an improvement, or a worsening of economic
conditions. Separate work drawing on related theories finds that declines
in consumer confidence were associated with increases in high-frequency
spanking during the Great Recession.14
To examine this possibility, we estimate a model that uses the rapidity of
increases and decreases in the unemployment rates—as well as the level of the
unemployment rate—to predict each of our parenting outcomes (tables 7.A2
and 7.A8). The fixed-effects models indicate that rapidly decreasing unemployment rates were associated with increases in physical and psychological
aggression (but not spanking), whereas rapidly increasing unemployment
rates were associated with increases in warmth and increases in activities.
mothers’ and fathers’ parenting185
Figure 7.14 High-Frequency Paternal Spanking by Unemployment Rate
60
50
Percent
40
–31%
–30%
–29%
–21%
–47%
30
UR 5 percent
20
UR 10 percent
10
0
All
Less than
high school
High
school
Some
college
College +
Source: Authors’ calculations based on data from the Fragile Families and Child Wellbeing
Study.
Note: No significant differences in the effect of unemployment among subgroups.
The results for rapidly increasing unemployment rates are very sensitive to
model specification, however. When we estimate odds ratios rather than
a linear probability model, we find a positive relationship between rapidly
increasing unemployment and physical aggression. These positive associations between rapidly increasing unemployment and harsh parenting are
consistent with our previous results for consumer confidence.
Fathers’ Parenting
We next turn to the predicted effects of a big recession on fathers’ parenting. Consistent with the results for mothers, we find that a high unemployment rate is predicted to reduce high-frequency spanking among fathers,
although these estimates are not statistically significant (figure 7.14 and
table 7.A5). Figure 7.15 illustrates the effects of a big recession on fathers’
high-frequency physical aggression. Again, results are similar to those found
for mothers. High unemployment is predicted to decrease the frequency of
fathers’ physical aggression; these results are marginally significant. (Again,
the sample size for the college-educated group warrants caution in interpreting that estimate). The effects of a big recession on fathers’ psychological aggression are shown in figure 7.16. Results, though not statistically
significant, vary considerably by parental education: high unemployment
is predicted to increase the likelihood among fathers with some college
education or more but to decrease it among fathers with less education
(table 7.A6). Adding controls for the father’s unemployment did not alter
the overall pattern of results (table 7.A7). Thus, results for fathers, though
186
children of the great recession
Figure 7.15 High-Frequency Paternal Physical Aggression
by Unemployment Rate
60
Percent
50
UR 5 percent
40
30
20
–46%
All
10
0
UR 10 percent
–33%
–39%
–43%
–91%
Less than
high school
High
school
Some
college
College +
Source: Authors’ calculations based on data from the Fragile Families and Child Wellbeing
Study.
Note: No significant differences in the effect of unemployment among subgroups. The
coefficient for all was significant in the individual fixed-effects model.
similar to those for mothers (high unemployment being related to less
harsh parenting), are not significant for the total sample.
ADDITIONAL ESTIMATES
Although many of the effects in the data were small or not statistically
significant, that a big recession reduces frequent spanking and frequent
physically aggressive parenting—especially among mothers—is surprising.
Figure 7.16 High-Frequency Paternal Psychological Aggression
by Unemployment Rate
60
+25%
Percent
50
40
–3%
–28%
+43%
–35%
UR 5 percent
30
UR 10 percent
20
10
0
All
Less than
high school
High
school
Some
college
College +*
Source: Authors’ calculations based on data from the Fragile Families and Child Wellbeing
Study.
Note: Chow tests show that the coefficient for unemployment for college or more is
different from the coefficient for unemployment for the less than high school group.
*p < .05
mothers’ and fathers’ parenting187
We therefore carry out several further sets of estimates for parenting outcomes. First we estimate models that controlled for parents’ unemployment (tables 7.A3 and 7.A7). The estimates from these models do not
change the basic story.
Next, we estimate models that include an interaction term that allows
the effect of unemployment rates to vary by whether the data were drawn
from the year nine survey, which for most families was in the midst of the
Great Recession. Our rationale for these analyses is that by pooling the data
over several years, we may have missed effects that were specific to the
Great Recession. The results based on these models (tables 7.A3 and 7.A7)
suggest that in fact, the effects of the unemployment rate on parenting
did vary depending on whether the family was interviewed in year nine.
Among mothers, we find that the associations between high unemployment rates and less frequent spanking and less frequent physical aggression were weaker during the Great Recession, though still in the same
direction. The pattern is similar for fathers.
These results suggest that the effects of a high unemployment rate may
well differ depending on the state of the overall economy or on the age
of the child. Overall, a higher local unemployment rate is—surprisingly—
associated with less harsh parenting by both mothers and fathers. However,
this reduction is much smaller during the year nine interviews. Because
these interviews were taken at the height of the Great Recession, it may
be that unemployment rates have a different effect when the overall economy is poor (and when unemployment rates are very high). It may also be,
however, that the effects of the unemployment rate differ when children
are older. As noted earlier, harsh parenting becomes less frequent when
children are age nine; so perhaps then, parenting can be tipped toward
greater harshness by an external factor such as high unemployment rates.
Given the design of our data, it is difficult to distinguish between those
two interpretations for the patterns at year nine. To further explore this
puzzle, we run our models again separately for children of different ages
and find that effects of high unemployment do seem to differ by child age,
the strongest effects occurring at age three.
Finally, to better understand the patterns in the data, we reestimate all
our models, stratifying our sample by the mothers’ race-ethnicity and marital status. We focus on the former models here, because results from these
models point to considerable heterogeneity. (Full results for both sets of
models are in tables 7.A4 and 7.A8). First, looking at spanking, only the
coefficient for Hispanics is significant, though the direction of the signs are
similar for the other two ethnic groups. Both Hispanic and black mothers
show a similar pattern for frequent physical aggression, but only black
mothers have a significant negative coefficient for psychological aggression. The most cautious interpretation is that black and Hispanic mothers
may be more aggressive toward their children when the unemployment
188
children of the great recession
rate is lower. No evidence was seen for white mothers for either measure
of aggression. Second, for high-frequency maternal warmth, the signs are
different for Hispanic than for white and black mothers (all three coefficients at p < 0.10). Hispanic mothers are more likely to exhibit frequent
warmth when local unemployment rates are high, and black and white
mothers are less likely to do so. Although these effects are only marginally
significant, they do point to a distinctive pattern for Hispanic mothers.
Why might higher unemployment rates be linked with more positive changes in parenting among Hispanic mothers? Two possible inter­
pretations come to mind. One is that maternal employment may be more
stressful in Hispanic families, and thus reductions in employment associated with higher unemployment rates might actually lead to improved
parenting. If this is the case, then controlling for the mothers’ unemployment should reduce the estimated effect of the local unemployment rate
on her parenting. However, in additional models in which we add such
controls, this is not the case. A second possibility is that when unemployment rates are higher, the most stressed Hispanic mothers are likely to
leave the sample (perhaps because they return to their countries of origin), and thus the improved parenting reflects the select nature of the
sample for Hispanics at age nine. Some evidence for this conjecture is in
fact seen in our data; the attrition rate between age five and nine is higher
for Hispanic mothers (17 percent) than for white and black mothers
(13 percent each).
CONCLUSION
Overall, our results challenge some conventional wisdom about the effects
of a big increase in unemployment rates on parenting. Previous research
might suggest that a big recession would lead to harsher parenting. Our
results suggest that overall, when unemployment rates are higher, both
mothers and fathers are less likely to engage in very frequent spanking or
physically aggressive parenting. At the same time, the evidence supports
the thesis that when rates of unemployment increase or decrease rapidly,
maternal parenting behavior is affected. When rates drop rapidly, physical
and psychological aggression are higher. When rates increase rapidly, harsh
parenting, warmth, and physical activities with the child increase. These
trends are seen when maternal and paternal unemployment are controlled
and, save for harsh parenting and rapidly increasing unemployment, are
seen in both the models with and without fixed effects. What is intriguing
is that these trends are seen for rapidly changing employment rates, suggesting that uncertainty in general might influence maternal aggression
toward the child as well as other parenting behaviors.
Our results also point to some intriguing variation by race-ethnicity,
though again we are unable to draw firm conclusions as to why. Clearly,
however, higher local unemployment rates seem to have a different effect
mothers’ and fathers’ parenting189
on Hispanic mothers than they do for other mothers, the latter effects
being beneficial.
APPENDIX
Measures
High-frequency maternal and paternal physical and psychological aggression. Mothers were asked a series of questions drawn from the Conflict
Tactics Scale for Parent and Child. This scale is designed to assess physically
and psychologically aggressive parenting behaviors and includes questions
about how often mothers spank, pinch, or hit their child, or use psychologically aggressive parenting, such as calling their child names, yelling,
cursing, or threatening, among other indicators. We recode these scales so
that high-frequency physically aggressive behavior is defined as aggressive
behavior that occurred eleven or more times in the previous year (physical aggression, mean = 0.21; SD = 0.40; psychological aggression, mean
= 0.50; SD = 0.50). We draw on a separate question about spanking and
examine it separately as well as part of the larger physical aggression scale
(spanking, mean = 0.12; SD = 0.32). In addition to reporting on their
own behaviors, mothers also reported on the same set of questions about
resident father’s aggressive parenting and spanking (physical aggression,
mean = 0.10, SD = 0.30; psychological aggression, mean = 0.34; SD =
0.48; spanking, mean = 0.22; SD = 0.41).
High-frequency maternal warmth. Interviewers visited a subsample of
mothers in their homes and recorded a number of observations about
mother-child interactions, including whether mothers spoke to child,
used terms of endearment, or cuddled child, among other items. These
questions are combined to create a dichotomous variable where highfrequency maternal warmth is defined as performing six or seven of the
seven items in the scale (mean = 0.56; SD = 0.50).
High-frequency maternal parenting activities. Mothers were asked how
many times in the past month or week they had performed a series of activities with their child, including activities such as playing outside or inside,
or watching TV together, among other activities. Each of these variables
is recoded so that 0 equals having not performed an activity in the past
month and 1 equals having performed an activity at least one to two times
in the past month. We sum these variables and create a bivariate measure
where high-frequency parenting activities are defined as performing six or
seven of the seven items in the scale (mean = 0.74; SD = 0.44).
Key Independent Variable
For each analysis, the unemployment rate is constructed using a measure of
the average unemployment rate in the sample city over the twelve months
prior to the interview.
190
children of the great recession
Key Moderating Variables
We study differences in the trajectories over time, and in the effects of
the Great Recession, on parenting stratified by maternal-paternal education at baseline. Parent’s education is coded as less than a high school
diploma or the completion of a GED, a high school diploma, some college or an associate’s degree or technical degree, or a bachelor’s degree
or greater.
Control Variables
We include a number of covariates in our models all measured at the
first survey wave (baseline). These include: mother’s or father’s age at
the birth, immigrant status (foreign born), number of children in the
household, a measure of whether the mother or father was living with
both biological parents at age fifteen, as well as city (twenty dummies
for each sample city) and survey year fixed effects (twelve calendar year
dummies).
Method
The figures that plot the trajectories of each outcome measured over time
present the mean levels of each outcome at each survey wave. All means
are weighted with the wave-specific city-weights to be representative of
births in the twenty study cities; the sample is restricted to parents who are
interviewed in all survey waves.
To study the effects of the Great Recession, we conduct linear probability
models for binary outcomes and ordinary least squares regression analyses
using the pooled data (waves 2 through 5 or 3 through 5, depending on
outcome). We use linear probability models for ease of interpretation but
logit models provide similar results (available on request). The standard
errors are clustered at both the city and individual level to account for
within city and within person clustering–nonindependence. Analyses are
conducted for all mothers and fathers and separately for mothers or fathers
with less than high school, high school only, some college, or college
degree or greater. We estimate pooled models and also a parallel set of
models with mother or father fixed effects.
To predict the effects of the Great Recession, we estimate the predicted
probability of each outcome when the unemployment rate is set at 5 percent, a rate typical of the period prior to the recession, and compare these
predictions with when the unemployment rate is set to 10 percent, a rate
typical of the Great Recession. We predict different probabilities for each
level of mothers’ or fathers’ education.
mothers’ and fathers’ parenting191
Table 7.A1 Full Regression Results, Maternal Parenting
High Frequency Maternal Spankinga
With Individual Fixed
Effects
Mothers
Unemployment rate
Education
Less than high school
High school
Some college
Relationship status
Married
Cohabiting
Mother’s age
Race-ethnicity
Black
Hispanic
Other
Immigrant
Children in household
Lived with both parents at
age fifteen
Interview year
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
-0.01**
0.11*
0.11**
0.21***
0.20***
0.11*
0.60***
0.28*
-0.08†
-0.01
0.04
0.09
(-2.64)
(2.18)
(2.74)
(4.71)
(4.50)
(2.54)
(3.54)
(2.21)
(-1.72)
(-0.31)
(0.93)
(1.22)
Without Individual
Fixed Effects
-0.02***
(-3.45)
0.01
0.04*
0.03*
(0.59)
(2.23)
(2.18)
0.03*
-0.00
-0.00***
(2.47)
(-0.03)
(-5.41)
-0.01
-0.04*
0.04
-0.05***
-0.01***
-0.01
(-0.55)
(-2.54)
(1.27)
(-3.81)
(-3.30)
(-0.96)
0.03
0.02
0.07†
0.05†
-0.03
0.01
0.04
-0.16***
-0.14***
-0.11***
-0.07†
(1.02)
(0.89)
(1.91)
(1.67)
(-1.09)
(0.31)
(0.41)
(-15.6)
(-5.70)
(-3.58)
(-1.72)
(Table continues on p. 192.)
192
children of the great recession
Table 7.A1 Continued
High Frequency Maternal Spankinga
With Individual Fixed
Effects
City
Austin
Baltimore
Detroit
Newark
Philadelphia
Richmond
Corpus Christi
Indianapolis
Milwaukee
New York
San Jose
Boston
Nashville
Chicago
Jacksonville
Toledo
San Antonio
Pittsburgh
Norfolk
Constant
Observations
Number of individuals
0.12**
13,3369
4,502
(2.66)
Without Individual
Fixed Effects
-0.02***
-0.07***
0.05**
-0.03†
-0.06**
-0.01
-0.02
0.01
-0.05***
-0.07***
-0.01
-0.07***
0.05***
-0.01
-0.02
0.00
-0.05**
-0.05**
-0.00
0.40***
13,369
4,604
(-3.47)
(-4.32)
(2.86)
(-1.67)
(-3.14)
(-0.74)
(-1.51)
(0.60)
(-3.34)
(-4.86)
(-0.50)
(-4.80)
(3.41)
(-0.39)
(-1.34)
(0.01)
(-2.75)
(-2.71)
(-0.14)
(11.10)
Source: Authors’ calculations based on data from the Fragile Families and Child Wellbeing Study.
Note: Z-stats in parentheses. Covariates are measured at the baseline survey (except year) and are clustered at city and individual level. Model 1 includes the level unemployment rate. The model without
individual fixed effects is clustered at city and individual level.
a
Includes years one through nine.
***p < .001; **p < .01; *p < .05; †p < .1
High
School
Mothers
High-frequency maternal spanking (z-stat)a
Unemployment
-0.01**
-0.02†
-0.01
rate (model 1)
(-2.64)
(-1.77)
(-0.70)
Unemployment
-0.02*
-0.02
-0.01
rate (model 2)
(-2.15)
(-1.63)
(-0.62)
Increasing
-0.00
-0.00
-0.00
unemployment
(-0.15)
(-0.33)
(-1.29)
rate
Decreasing
0.00
0.00
0.00
unemployment
(0.69)
(0.55)
(0.63)
rate
Observations
133,369
4,458
4,098
Number of
4,502
1,554
1,356
individuals
High-frequency maternal physical aggression (z-stat)b
Unemployment
-0.02***
-0.02†
-0.02†
rate (model 1)
(-3.32)
(-1.88)
(-1.68)
-0.02†
Unemployment
-0.02***
-0.02†
rate (model 2)
(-3.32)
(-1.88)
(-1.68)
Increasing
0.00
0.00
0.00
unemployment
(0.78)
(0.54)
(0.69)
rate
All
Less than
High
School
-0.05**
(-3.08)
-0.03
(-1.45)
0.00
(0.38)
0.01
(1.35)
1,445
484
-0.03†
(-1.74)
-0.03†
(-1.74)
0.00
(0.20)
-0.00
(-0.71)
3,368
1,108
-0.02
(-1.45)
-0.02
(-1.45)
0.00
(0.02)
College +
-0.01
(-0.75)
-0.01
(-0.65)
0.00
(1.04)
Some
College
With Individual Fixed Effects
-0.03***
(-5.78)
-0.03***
(-5.90)
0.00*
(2.31)
133,369
4,604
0.00*
(2.13)
-0.02***
(-3.45)
-0.02***
(-3.91)
0.00*
(2.02)
All
-0.05***
(-3.73)
-0.02†
(-1.79)
0.00†
(1.81)
4,458
1,593
-0.00*
(-2.44)
-0.01
(-1.57)
-0.01
(-1.60)
0.00
(1.02)
Less than
High
School
-0.05*
(-2.46)
-0.03***
(-4.66)
0.00
(0.95)
4,089
1,391
0.00
(0.91)
-0.01*
(-2.08)
-0.01*
(-1.95)
0.00
(0.63)
High
School
-0.04
(-1.20)
-0.03
(-1.45)
0.00
(1.56)
1,445
496
0.00
(-0.15)
-0.04*
(-2.12)
-0.04*
(-2.10)
0.00
(0.83)
College +
(Table continues on p. 194.)
-0.05**
(-2.91)
-0.02†
(-1.80)
0.00
(0.20)
3,368
1,124
-0.00
(-0.69)
-0.01
(-1.03)
-0.01
(-1.12)
0.00
(0.39)
Some
College
Without Individual Fixed Effects
Table 7.A2 Coefficients and Standard Errors, Rate of Change in Unemployment for Maternal Parenting Outcomes
Less than
High
School
High
School
0.01
(1.42)
973
421
-0.02
(-0.71)
-0.01
(-0.55)
-0.00
(-0.50)
0.01
(0.98)
979
421
-0.02
(-0.55)
-0.03
(-0.84)
2,308
994
-0.04**
(-2.61)
-0.03*
(-2.53)
-0.00
(-0.54)
0.00
(0.02)
2,321
995
-0.00
(-0.21)
-0.01
(-0.32)
College +
0.00
(0.55)
Some
College
With Individual Fixed Effects
Decreasing
0.00*
0.00
0.00
unemployment
(2.10)
(1.22)
(1.12)
rate
Observations
9,080
3,001
2,798
Number of
4,068
1,398
1,255
individuals
High-frequency maternal psychological aggression (z-stat)b
Unemployment
-0.00
0.01
0.02†
rate (model 1)
(-0.48)
(0.60)
(1.70)
Unemployment
-0.00
0.01
0.02†
rate (model 2)
(-0.12)
(0.91)
(1.81)
Increasing
0.00
0.00
0.00
unemployment
(0.45)
(1.03)
(0.34)
rate
Decreasing
0.00**
0.01*
0.00
unemployment
(2.56)
(2.03)
(1.60)
rate
Observations
9,143
3, 030
2,813
Number of
4,071
1,399
1,256
individuals
High-frequency maternal warmth (z-stat)b,c
Unemployment
-0.00
0.00
0.01
rate (model 1)
(-0.10)
(0.15)
(0.36)
Unemployment
-0.00
0.01
0.01
rate (model 2)
(-0.15)
(0.29)
(0.25)
All
Table 7.A2 Continued
0.01*
(2.43)
0.00†
(1.68)
-0.00
(-0.05)
0.00
(0.01)
-0.01
(-0.18)
0.01
(0.44)
3, 030
1,399
-0.01
(-0.42)
0.01
(0.77)
0.00
(1.27)
-0.01†
(-1.77)
-0.01
(-1.61)
0.00
(1.57)
9,143
4,071
3,001
1,398
0.00
(1.77)
†
9,080
4,068
0.00
(1.17)
All
Less than
High
School
0.04
(1.05)
0.02
(0.96)
2,813
1,256
0.00
(1.31)
-0.01
(-0.39)
0.00
(0.34)
0.00
(1.38)
2,798
1,255
0.00
(0.53)
High
School
-0.03
(-0.57)
-0.02
(-0.90)
2,321
995
-0.00
(-0.89)
-0.06**
(-3.07)
-0.04**
(-3.11)
-0.00
(-0.40)
2,308
994
-0.00
(-0.36)
Some
College
-0.02
(-0.43)
-0.04
(-1.06)
979
421
-0.00
(-0.11)
0.02
(0.47)
-0.02
(-1.01)
0.00
(0.46)
973
421
0.00
(0.92)
College +
Without Individual Fixed Effects
0.00
(0.33)
6,743
3,550
-0.00
(-0.02)
-0.00
(-0.08)
0.00
(1.00)
-0.01
(-1.43)
674
350
-0.01
(-0.68)
-0.01
(-0.43)
-0.00
(-0.16)
0.00
(0.97)
1,268
486
0.00
(0.10)
1,701
887
0.02
(1.61)
0.02
(1.33)
0.00†
(1.83)
-0.00
(-0.83)
2,821
1,086
11,237
4,425
-0.00
(-0.50)
0.00†
(1.67)
-0.00
(-0.43)
0.00†
(1.84)
3,683
1,506
-0.00
(-0.15)
0.02
(1.26)
0.00
(0.02)
0.00
(0.78)
2,256
1,203
0.00
(1.23)
0.00
(1.18)
3,465
1,347
-0.00
(-0.55)
0.02
(1.32)
-0.01
(-0.45)
0.00
(0.26)
2,112
1,110
-0.00
(-0.52)
0.00†
(1.67)
2,821
1,086
-0.00
(-0.61)
0.05*
(2.35)
0.01
(0.69)
0.00*
(2.07)
1,701
887
0.00
(0.37)
0.00*
(2.39)
1,268
486
0.00
(0.90)
0.02
(0.46)
-0.01
(-0.31)
-0.00
(-0.93)
674
350
-0.00
(-0.40)
-0.00
(-0.24)
Source: Authors’ calculations based on data from the Fragile Families and Child Wellbeing Study.
Note: Z-stats in parentheses. Covariates are measured at the baseline survey (except year) and are clustered at city and individual level. The model without individual fixed effects is
clustered at city and individual level. Estimates from linear probability models.
a
Includes years one through nine.
b
Includes years three through nine.
c
A subsample of families received in-home visits and were assessed for warmth.
***p < .001; **p < .01; *p < .05; †p < .1
Increasing
0.00**
0.00†
0.00
rate of
(2.74)
(1.78)
(1.51)
unemployment
Decreasing
0.00
0.00
-0.00
rate of
(0.12)
(1.09)
(-0.61)
unemployment
Observations
6,743
2,256
2,112
Number of
3,550
1,203
1,110
individuals
High-frequency maternal parenting activities (z-stat)b
Unemployment
0.00
-0.00
-0.01
rate (model 1)
(0.24)
-0.01
(-0.44)
Unemployment
0.00
-0.00
-0.01
rate (model 2)
(0.10)
(-0.01)
(-0.53)
Increasing
0.00
0.00
0.00
rate of
(1.62)
(1.39)
(0.23)
unemployment
Decreasing
-0.00
0.00
-0.00
rate of
(-0.50)
(0.46)
(-0.70)
unemployment
Observations
11,237
3,683
3,465
Number of
4,425
1,506
1,347
individuals
196
children of the great recession
Table 7.A3 Sensitivity of Coefficients, Maternal Parenting Outcomes
With Individual Fixed
Effects
Mothers
High-frequency maternal spanking (z-stat)a
Unemployment rate (model 1)
-0.01**
(-2.64)
Unemployment rate (model 3)
-0.03*
(-2.11)
Mother’s unemployment
0.02
(0.68)
(-1.64)
Bio-social fathers not employed
-0.05†
Unemployment rate (model 4)
-0.03***
(-3.76)
Unemployment rate * year nine
0.02**
(2.68)
High-frequency maternal physical aggression (z-stat)b
Unemployment rate (model 1)
-0.02***
(-3.63)
Unemployment rate (model 3)
-0.06**
(-2.57)
Mother’s unemployment
0.03
(0.93)
Bio-social fathers not employed
0.00
(0.05)
Unemployment rate (model 4)
-0.04***
(-4.92)
Unemployment rate * year nine
0.03***
(3.32)
High-frequency maternal psychological aggression (z-stat)b
Unemployment rate (model 1)
-0.00
(-0.48)
Unemployment rate (model 3)
-0.04
(-1.55)
Mother’s unemployment
0.03
(0.90)
Bio-social fathers not employed
-0.02
(-0.47)
Unemployment rate (model 4)
-0.01
(-1.30)
Unemployment rate * year nine
0.01
(1.35)
High-frequency maternal warmth (z-stat)b,c
Unemployment rate (model 1)
-0.00
(-0.10)
Unemployment rate (model 3)
0.04
(0.95)
Mother’s unemployment
-0.07
(-1.39)
Bio-social fathers not employed
0.03
(0.57)
Unemployment rate (model 4)
0.00
(0.04)
Unemployment rate * year nine
-0.00
(-0.14)
High-frequency maternal parenting activities (z-stat)b
Unemployment rate (model 1)
0.00
(0.24)
Unemployment rate (model 3)
0.01
(0.38)
Mother’s unemployment
0.01
(0.52)
Bio-social fathers not employed
0.01
(0.43)
Unemployment rate (model 4)
0.02**
(2.74)
Unemployment rate * year nine
-0.03***
(-3.62)
Without Individual
Fixed Effects
-0.02***
-0.02***
-0.01
-0.03
-0.03***
0.02***
(-3.45)
(-3.85)
(-0.65)
(-1.60)
(-3.62)
(3.36)
-0.03***
-0.05*
-0.01
0.02
-0.05***
0.03***
(-5.78)
(-2.26)
(-0.80)
1.49
(-4.95)
(3.66)
-0.01†
-0.01
0.01
0.01
-0.02†
0.01
(-1.77)
(-0.51)
0.37
(0.44)
(-1.67)
(1.27)
-0.00
0.02
-0.08***
-0.02
-0.00
0.00
(-0.05)
(0.40)
(-3.39)
(-0.72)
(-0.11)
(0.15)
-0.00
-0.00
0.02
-0.02
0.03*
-0.03***
(-0.02)
(-0.24)
(1.15)
(-1.09)
(1.98)
(-5.01)
Source: Authors’ calculations based on data from the Fragile Families and Child Wellbeing Study.
Note: Z-stats in parentheses. Covariates are measured at the baseline survey (except year) and are
clustered at city and individual level. The model without individual fixed effects is clustered at city and
individual level. Estimates from linear probability model.
a
Includes years one through nine.
b
Includes years three through nine.
c
A subsample of families received in-home visits and were assessed for warmth.
***p < .001; **p < .01; *p < .05; †p < .1
0.00
(0.09)
-0.04†
(-1.78)
-0.02†
(-1.64)
0.00
(0.27)
0.04†
(1.94)
-0.01
(-0.75)
-0.02†
(-1.82)
-0.02†
(-1.95)
-0.04*
(-2.37)
-0.01
(-0.71)
0.01
(0.67)
-0.01
(-0.69)
-0.02†
(-1.62)
-0.02
(-1.24)
White
-0.01
(-0.60)
Hispanic
Source: Authors’ calculations based on data from the Fragile Families and Child Wellbeing Study.
Note: Z-stats in parentheses. The model includes individual fixed effects. Estimates from linear probability models.
a
Includes years one through nine.
b
Includes years three through nine.
c
A subsample of families received in-home visits and were assessed for warmth.
*p < 0.05; †p < 0.1
High-frequency parenting activities (z-stat)
Unemployment rate
High-frequency maternal warmth (z-stat)b,c
Unemployment rate
High-frequency maternal psychological aggression (z-stat)b
Unemployment rate
High-frequency maternal physical aggression (z-stat)b
Unemployment rate
Mothers
High-frequency maternal spanking (z-stat)a
Unemployment rate
Black
-0.01
(-0.40)
-0.03
(-1.39)
-0.00
(-0.21)
-0.03*
(-2.43)
-0.01
(-1.27)
Married
0.01
(1.06)
0.02
(1.32)
-0.00
(-0.05)
-0.02
(-1.54)
-0.02†
(-0.65)
Cohabiting
With Individual Fixed Effects
Table 7.A4 Coefficients and Standard Errors, Maternal Parenting Outcomes by Subgroup
-0.01
(-0.50)
-0.01
(-0.33)
-0.01
(-0.60)
-0.03*
(-2.35)
-0.01
(-0.56)
Single
198
children of the great recession
Table 7.A5 Full Regression Results, Paternal Parenting
High-Frequency Paternal Spankinga
With Individual Fixed
Effects
Fathers
Unemployment rate
-0.01
Education
Less than high school
High school
Some college
Relationship status
Married
Cohabiting
Fathers’ age
Race-ethnicity
Black
Hispanic
Other
Immigrant
Children in household
Lived with both parents at age nineteen
Interview Year
2000
—
2001
0.27***
2002
0.30***
2003
0.24***
2004
0.23***
2005
0.17**
2006
0.09
2007
-0.01
2008
-0.04
2009
-0.06
2010
—
(-1.63)
(—)
(3.88)
(5.04)
(4.46)
(3.82)
(2.95)
(0.89)
(-0.09)
(-0.69)
(-1.21)
(—)
Without Individual
Fixed Effects
-0.01†
(-1.75)
0.05*
0.08***
0.08***
(2.25)
(3.45)
(4.07)
0.09***
0.05***
-0.00**
(7.07)
(3.84)
(-2.85)
-0.04*
-0.06*
-0.03
-0.06*
-0.01†
-0.03*
(-2.00)
(-2.08)
(-0.91)
(-2.16)
(-1.95)
(-2.39)
—
0.17**
0.21***
0.15***
0.13**
0.07†
—
-0.12†
-0.12*
-0.15**
-0.12*
(—)
(3.00)
(4.63)
(3.47)
(2.62)
(1.74)
(—)
(-1.94)
(-2.41)
(-2.92)
(-2.05)
mothers’ and fathers’ parenting199
Table 7.A5 Continued
High-Frequency Paternal Spankinga
With Individual Fixed
Effects
City
Austin
Baltimore
Detroit
Newark
Philadelphia
Richmond
Corpus Christi
Indianapolis
Milwaukee
New York
San Jose
Boston
Nashville
Chicago
Jacksonville
Toledo
San Antonio
Pittsburgh
Norfolk
Constant
Observations
Number of individuals
0.14†
9,142
3,220
(1.77)
9,142
3,220
Without Individual
Fixed Effects
0.09***
0.02
0.11***
0.01
0.03*
0.09***
0.13***
0.11***
0.08***
0.04***
0.12***
-0.01
0.23***
0.06***
0.11***
0.11***
0.15***
0.16***
0.13***
0.16**
(12.10)
(0.82)
(6.03)
(1.01)
(2.22)
(5.66)
(10.10)
(9.50)
(9.01)
(4.08)
(8.96)
(-0.68)
(20.20)
(5.59)
(9.24)
(5.47)
(10.50)
(10.90)
(8.86)
(2.73)
Source: Authors’ calculations based on data from the Fragile Families and Child Wellbeing Study.
Note: Z-stats in parentheses. Covariates are measured at the baseline survey (except year) and are
clustered at city and individual level. Model 1 includes level unemployment rate. The model without
individual fixed effects is clustered at city and individual level.
a
Years one through nine.
***p < .001; **p < .01; *p < .05; †p < .1
High
School
Fathers
High-frequency paternal spanking (z-stat)a
Unemployment
-0.01
-0.01
-0.02
rate (model 1)
(-1.63)
(-1.10)
(-1.48)
-0.01
-0.02†
Unemployment
-0.01†
rate (model 2)
(-1.71)
(-0.98)
(-1.67)
Increasing
-0.00
-0.00
0.00
unemployment
(-0.64)
(-1.06)
(0.19)
rate
Decreasing
-0.00
0.00
-0.00
unemployment
(-0.85)
(0.41)
(-1.38)
rate
Observations
9,142
2,203
1,774
Number
3,220
1,071
829
of individuals
High-frequency paternal physical aggression (z-stat)b
Unemployment
-0.01†
-0.01
-0.01
rate (model 1)
(-1.66)
(-0.67)
(-0.67)
-0.01
-0.01
Unemployment
-0.01†
rate (model 2)
(-1.68)
(-0.70)
(-0.60)
Increasing
-0.00
-0.00
-0.00
unemployment
(-0.63)
(-0.25)
(-0.94)
rate
All
Less than
High
School
-0.01
(-0.59)
-0.01
(-0.41)
0.00*
(2.03)
0.00
(0.89)
890
364
-0.02
(-1.17)
-0.02
(-1.10)
-0.00
(-0.15)
-0.00
(-0.94)
1,455
645
-0.01
(-0.82)
-0.01
(-0.79)
0.00
(0.41)
College +
0.00
(0.16)
-0.00
(-0.01)
-0.00
(-1.26)
Some
College
With Individual Fixed Effects
-0.02***
(-4.04)
-0.02***
(-3.97)
0.00
(0.61)
9,142
3,220
-0.00
(-1.12)
-0.01†
(-1.75)
-0.01†
(-1.84)
0.00
(0.16)
All
-0.02**
(-2.66)
-0.02**
(-2.61)
0.00
(0.33)
2,203
1,071
0.00
(0.29)
-0.01
(-1.35)
-0.01
(-1.23)
-0.00*
(-2.00)
Less than
High
School
1,455
645
-0.01
(-0.69)
-0.01
(-0.77)
0.00
(0.00)
-0.02†
(-1.67)
-0.02†
(-1.63)
0.00
(0.57)
-0.01*
(-1.96)
-0.01
(-0.39)
-0.01
(-0.71)
-0.00
(-0.93)
Some
College
1,774
829
-0.00*
(-2.37)
-0.02†
(-1.69)
-0.02*
(-2.00)
0.00
(1.39)
High
School
Without Individual Fixed Effects
Table 7.A6 Coefficients and Standard Errors, Rate of Change in Unemployment for Paternal Parenting Outcomes
-0.02
(-1.15)
-0.01
(-0.90)
0.00
(0.82)
890
364
0.00
(0.71)
-0.01
(-1.18)
0.01
(-0.64)
0.00**
(2.74)
College +
0.00
(0.28)
692
321
0.05
(1.54)
0.06†
(1.77)
0.00
(0.20)
0.01
(1.37)
692
321
-0.00
(-0.13)
1,158
603
0.02
(1.16)
0.02
(1.07)
-0.00
(-1.12)
0.00
(0.12)
1,158
603
4,919
2,659
1,736
993
1,333
742
0.01
(1.08)
0.01†
(1.86)
0.00
(1.15)
-0.03
(-1.28)
-0.03
(-1.20)
-0.00
(-0.01)
-0.03***
(-3.40)
-0.03**
(-3.04)
0.00
(1.35)
-0.01
(-1.03)
-0.01
(-0.86)
0.00
(0.66)
1,333
742
0.00
(0.67)
1,736
993
-0.00
(-0.23)
4,919
2,659
-0.00
(-0.31)
1,158
603
-0.00
(-0.085)
0.02
(1.52)
0.02
(1.28)
-0.00
(-0.90)
1,158
603
-0.00
(-1.37)
692
321
0.01†
(1.87)
0.06
(1.60)
0.06†
(1.84)
0.00
(0.50)
692
321
0.00
(0.35)
Source: Authors’ calculations based on data from the Fragile Families and Child Wellbeing Study.
Note: Z-stats in parentheses. Covariates are measured at the baseline survey (except year) and are clustered at city and individual level. The model without individual fixed effects
is clustered at city and individual level. Estimates from linear probability models.
a
Includes years one through nine.
b
Includes years three through nine.
***p < .001; **p < .01; *p < .05; †p < .1
Decreasing
-0.00
-0.00
0.00
unemployment
(-0.18)
(-0.29)
(0.05)
rate
Observations
4,919
1,736
1,333
Number
2,659
993
742
of individuals
High-frequency paternal psychological aggression (z-stat)b
Unemployment
-0.00
-0.02
-0.03
rate (model 1)
(-0.14)
(-1.09)
(-1.18)
Unemployment
0.00
-0.02
-0.02
rate (model 2)
(0.03)
(-1.03)
(-0.96)
Increasing
-0.00
0.00
-0.00
unemployment
(-0.90)
(0.34)
(-1.14)
rate
Decreasing
0.00
0.00
0.01
unemployment
(1.43)
(0.53)
(1.57)
rate
Observations
4,919
1,736
1,333
Number
2,659
993
742
of individuals
(-1.63)
(-0.38)
(-0.52)
(1.47)
(-2.88)
(2.44)
(-1.66)
(-1.12)
(-0.16)
(-0.38)
(-2.96)
(2.52)
(-0.14)
(-0.76)
(0.07)
(2.87)
(-0.76)
(0.07)
-0.01
-0.01
-0.02
0.05
-0.03**
0.02*
-0.01†
-0.03
-0.01
-0.02
-0.03**
0.02*
-0.00
-0.03
0.01
0.18**
-0.03
0.01
-0.01
-0.04
-0.00
0.07*
-0.04
-0.00
-0.02***
-0.03*
-0.01
0.05*
-0.04***
0.03***
-0.01†
-0.01
-0.01
0.05*
-0.03**
0.03*
(-1.03)
(-1.15)
(-0.05)
(2.37)
(-1.15)
(-0.05)
(-4.04)
(-2.26)
(-0.52)
(2.16)
(-5.05)
(3.73)
(-1.75)
(-1.50)
(-0.44)
(-2.22)
(-2.67)
(2.28)
Without Individual
Fixed Effects
Source: Authors’ calculations based on data from the Fragile Families and Child Wellbeing Study.
Note: Z-stats in parentheses. Covariates are measured at the baseline survey (except year) and are clustered at city and individual
level. The model without individual fixed effects is clustered at city and individual level. Estimates from linear probability models.
a
Includes years one through nine.
b
Includes years three through nine.
***p < .001; **p < .01; *p < .05; †p < .1
Fathers
High-frequency paternal spanking (z-stat)a
Unemployment rate (model 1)
Unemployment rate (model 3)
Mother’s unemployment
Bio-social fathers not employed
Unemployment rate (model 4)
Unemployment rate * year nine
High-frequency paternal physical aggression (z-stat)b
Unemployment rate (model 1)
Unemployment rate (model 3)
Mother’s unemployment
Bio-social father’s not employed
Unemployment rate (model 4)
Unemployment rate * year nine
High-frequency paternal psychological aggression (z-stat)b
Unemployment rate (model 1)
Unemployment rate (model 3)
Mother’s unemployment
Bio-social father’s not employed
Unemployment rate (model 4)
Unemployment rate * year nine
With Individual
Fixed Effects
Table 7.A7 Sensitivity of Coefficients, Paternal Parenting Outcomes
0.01
(0.75)
0.00
(0.04)
0.00
(0.15)
-0.03*
(-2.19)
-0.01
(-0.50)
0.01
(0.25)
Hispanic
0.01
(0.53)
-0.01
(-0.99)
-0.00
(-0.03)
White
0.02
(1.12)
-0.01
(-1.28)
-0.01
(-0.54)
Married
—
(—)
—
(—)
-0.01
(-0.55)
-0.02
(-1.01)
Single
-0.01
(-0.77)
-0.02†
(-1.82)
Cohabiting
Source: Authors’ calculations based on data from the Fragile Families and Child Wellbeing Study.
Note: Z-stats in parentheses. The model includes individual fixed effects and is clustered at city and individual level. Estimates from linear probability
models.
a
Includes years one through nine.
b
Includes years three through nine.
*p < .05; †p < .1
High-frequency paternal psychological aggression (z-stat)b
Unemployment rate
High-frequency paternal physical aggression (z-stat)b
Unemployment rate
Fathers
High-frequency paternal spanking (z-stat)a
Unemployment rate
Black
Table 7.A8 Coefficients and Standard Errors, Paternal Parenting Outcomes by Subgroups
204
children of the great recession
NOTES
1.Brooks-Gunn and Markman 2005; Lareau 2003.
2. Brooks-Gunn and Markman 2005; Raikes et al. 2006.
3. Elder 1974.
4. Elder and Conger 2000.
5. See, for example, Leinonen, Solantaus, and Punamki 2002.
6. On punative, McLoyd 1990; on warmth, Klebanov, Brooks-Gunn, and
Duncan 1994.
7. Brooks-Gunn, Schneider, and Waldfogel 2013; Huang et al. 2011; Lee et al.
2013; Lindo, Hanson, and Schaller 2013.
8. Berger et al. 2013.
9. On prior recessions, Leinonen, Solantaus, and Punamaki 2002; on Iowa,
Conger and Elder 1994; see also Elder 1974.
10. Straus et al. 1998.
11. Gershoff 2002.
12. Caldwell and Bradley 2001.
13. MacKenzie et al. 2014.
14.Brooks-Gunn, Schneider, and Waldfogel 2013.
REFERENCES
Berger, Lawrence M., Sarah A. Font, Kristen S. Slack, and Jane Waldfogel. 2013.
“Income and Child Maltreatment: Evidence from the Earned Income Tax
Credit.” Paper presented at the Association for Public Policy Analysis and
Management annual conference. Washington, D.C. (November 8, 2013).
Brooks-Gunn, Jeanne, and Lisa Markman. 2005. “The Contribution of Parenting
to Ethnic and Racial Gaps in School Readiness.” The Future of Children 15(1):
139–69.
Brooks-Gunn, Jeanne, William Schneider, and Jane Waldfogel. 2013. “The
Great Recession and the Risk for Child Maltreatment.” Child Abuse & Neglect
37(10): 721–29.
Caldwell, Bettye M., and Robert H. Bradley. 2001. HOME Inventory Admin­
istration Manual, 3rd ed. Little Rock: University of Arkansas at Little Rock.
Conger, Rand D., and Glen H. Elder Jr. 1994. “Families in Troubled Times: The
Iowa Youth and Families Project.” In Families in Troubled Times: Adapting to
Change in Rural America, edited by Rand D. Conger and Glen H. Elder Jr.
Hillsdale, N.J.: Aldine.
Elder, Glen H., Jr. 1974. Children of the Great Depression: Social Changes in Life
Experience. Boulder, Colo.: Westview Press.
Elder, Glen H., Jr., and Rand D. Conger. 2000. Children of the Land: Adversity
and Success in Rural America. Chicago: University of Chicago Press.
mothers’ and fathers’ parenting205
Gershoff, Elizabeth T. 2002. “Corporal Punishment by Parents and Associated
Child Behaviors and Experiences: A Meta-Analytic and Theoretical Review.”
Psychological Bulletin 128(4): 539–79.
Huang, Mary I., Mary Ann O’Riordan, Ellen Fitzenrider, Lolita McDavid,
Alan R. Cohen, and Shenandoah Robinson. 2011. “Increased Incidence
of Nonaccidental Head Trauma in Infants Associated with the Economic
Recession.” Journal of Neurosurgery: Pediatrics 8(2): 171–76.
Klebanov, Pamela Kato, Jeanne Brooks-Gunn, and Greg J. Duncan. 1994. “Does
Neighborhood and Family Poverty Affect Mothers’ Parenting, Mental Health,
and Social Support?” Journal of Marriage and the Family 56(2): 441–55.
Lareau, Annette. 2003. Unequal Childhoods: Class, Race, and Family Life.
Berkeley: University of California Press.
Lee, Dohoon, Jeanne Brooks-Gunn, Sara S. McLanahan, Daniel Notterman,
and Irwin Garfinkel. 2013. “The Great Recession, Genetic Sensitivity, and
Maternal Harsh Parenting.” Proceedings of the National Academy of Sciences
110(34): 13780–84.
Leinonen, Jenni A., Tytti S. Solantaus, and Raija-Leena Punamki. 2002. “The
Specific Mediating Paths Between Economic Hardship and the Quality of
Parenting.” International Journal of Behavioral Development 26(5): 423–35.
Lindo, Jason M., Jessamyn Schaller, and Benjamin Hansen. 2013. “Economic
Conditions and Child Abuse.” NBER working paper no. 18994. Cambridge,
Mass.: National Bureau of Economic Research.
MacKenzie, Michael J., Eric Nicklas, Jeanne Brooks-Gunn, and Jane Waldfogel.
2014. “Repeated Exposure to High-Frequency Spanking and Child
Externalizing Behavior Across the First Decade: A Moderating Role for
Cumulative Risk.” Child Abuse & Neglect 38(12): 1895–901.
McLanahan, Sara S., Irwin Garfinkel, Ronald Mincy, and Elizabeth Donahue.
2010. “Introducing the Issue.” The Future of Children 20(2): 3–17.
McLanahan, Sara S., and Gary Sandefur. 1994. Growing Up with a Single Parent:
What Hurts, What Helps. Cambridge, Mass.: Harvard University Press.
McLoyd, Vonnie C. 1990. “The Impact of Economic Hardship on Black
Families and Children: Psychological Distress, Parenting, and Socioemotional
Development.” Child Development 61(2): 311–46.
Pleck, Joshua H. 2007. “Why Could Father Involvement Benefit Children?
Theoretical Perspectives.” Applied Development Science 11(4): 196–202.
Raikes, Helen, Barbara A. Pan, Gayle Luze, Catherine Tamis-LeMonda, Jeanne
Brooks-Gunn, Jill Constatine, Louisa B. Tarullo, H. Abigail Raikes, and Eileen
T. Rodriguez. 2006. “Mother-Child Book Reading in Low-Income Families:
Correlates and Outcomes During the First Three Years of Life.” Child
Development 77(4): 924–53.
Straus, Murray A., Sherry L. Hamby, David Finkelhor, David W. Moore, and
Desmond Runyan. 1998. “Identification of Child Maltreatment with the
Parent Child Conflict Tactics Scales: Development and Psychometric Data for
a National Sample of American Parents.” Journal of Child Abuse & Neglect
22(4): 249–70.
Chapter 8
Child Well-Being
William Schneider, Jane Waldfogel,
and Jeanne Brooks-Gunn
I
n this final chapter, we ask how a large change in the unemployment
rate, one similar to that during the Great Recession, affects children’s
well-being. To address this question, we examine child well-being at ages
three, five, and nine, drawing on data about children’s behavior problems (assessed through maternal report of externalizing and internalizing
behavior problems), language development (measured by asking children
to define words), and physical health (assessed by children’s weight for
height and age, translated into the rate of overweight or obesity).
The other chapters in this volume consider outcomes beginning at age
one. This chapter begins at age three. We do so because measuring aspects
of children’s behavior and cognitive development at very young ages in
the same metric as used when measuring older children’s capacities is
difficult.
An important consideration for this chapter, as with that on parenting,
is the developmental trajectory of children from age three to nine. That
the children in our study were first evaluated at age three is important
for a number of reasons. First, it has the distinct advantage of providing
information on the child and family’s background prior to school entry.
Second, by measuring development over time, we are able to evaluate
the influence of a variety of child, family, and other factors on child wellbeing—most importantly, the macroeconomic conditions generated by
recessions.
Because children’s abilities change over time, the questions they were
asked and the tools used to evaluate them sometimes change as well. Our
three measures were chosen because they can be assessed when children
were three, five, and nine years old. Child behavior problems are measured
by asking mothers about common behaviors at each age. Some of the
behaviors are constant over the three age groups and others are specific
to a particular age group. For example, when children are three years
old, a question used to help evaluate child behavior focuses on children’s
interactions with their parents, but when children are nine years old, the
survey also includes questions about the ways in which they interact with
their peers. A standard measure of receptive vocabulary is used at all
child well-being207
three ages—the Peabody Picture Vocabulary Test (PPVT). The words
included in the PPVT become progressively more difficult, such that
a nine-year-old will almost always identify the easiest words in the test
and a three-year-old will rarely correctly identify the words known by
most nine-year-olds. Scores on such tests are standardized by age, so that
comparisons are made within age groups (raw scores could also be used;
unlike standardized scores, the raw scores increase with age).1 Scores are
standardized to have a mean of 100 and a standard deviation of 15. Weight
and height are measured in the same way at all three ages. Because both
increase with age, our measures of overweight and obesity are standardized by age as well as height and gender (using the growth charts that
pediatricians commonly use).
HIGHER UNEMPLOYMENT RATES AND CHILD WELL-BEING
A wide range of research has demonstrated both direct effects of economic hardship on child well-being and as indirect effects through altered
parenting practices.2 Economic hardship and poverty have negative and
long-term effects on child development. Children who live in poverty,
particularly early in their lives, are more likely to leave school early, have
lower scores on cognitive ability tests, and have more emotional and
behavioral problems.3 Poverty affects children in multiple ways. Material
deprivation such as unsafe housing or lack of nutritious food is one.4
Neighborhood poverty and instability have also been shown to negatively
affect children.5 Children may also be affected through changes in parenting. Parents experiencing poverty and economic hardship are more likely
to use harsh and aggressive parenting practices, be depressed, and use less
sensitive parenting with their children.6 In addition, boys and girls may
be differently affected by economic hardship, boys more likely to act out
and increase risky behavior and girls more likely to become withdrawn or
less likely to be affected in general.7 The family stress model documents
the pathways through which economic hardship and uncertainty increase
harsh and inconsistent parenting, resulting in increased child behavior
problems.8
In contrast, we focus here on the direct effect on child well-being of the
local unemployment rate in the year before the time the children were seen
and changes in the local unemployment rate. Consequently, we examine
economic shocks that occurred community-wide, rather than only to certain families. Individual-level unemployment of mothers and fathers is
also examined in some of our models.
A limited body of research has begun to investigate the effects that
the Great Recession may have had on children’s problem behaviors. One
study using Fragile Families data found that increased uncertainty during
the Great Recession, as measured by the national consumer sentiment
208
children of the great recession
index, was associated with increased problem behaviors among nine-yearold boys, but not among girls of the same age.9 A study of low-income
families in Michigan during the Great Recession found that both the local
unemployment rate and more subjective measures of economic hardship were associated with increased problem behaviors among children.10
Research on macroeconomic changes and children’s cognitive functioning is rare. One study reports that mass job layoffs have negative effects
on children’s school achievement.11 To date, no studies have investigated
associations between the Great Recession and children’s language outcomes or health. Little is known about recessions and child weight, specifically being overweight or obese. Research has found that poverty and
economic hardship are associated with adults’ decreased access to healthy
and nutritious food.12 However, some evidence indicates that economic
downturns are associated with better adult health and health behaviors.13
Whether associations will be found for children’s health is an open question.
TRENDS IN CHILD WELL-BEING
We are interested in several aspects of child well-being. In children’s
behavior, we focus on two types of problems, internalizing and externalizing behaviors as reported by the child’s mother. Our measures are drawn
from a well-known battery of questions researchers use to assess children’s
behavior problems.14 Internalizing behavior is made up of a series of
questions designed to assess whether children are anxious or depressed
or withdrawn, or have other somatic complaints. Externalizing behavior
is made up of a series of questions designed to assess whether children are
being aggressive or breaking rules. Children’s externalizing and internalizing behaviors in and of themselves are predictive of lower academic
achievement, worse school adjustment, more substance use, and more
juvenile delinquency.15
To assess children’s cognitive development, we draw on a measure that
tests children’s receptive vocabulary, the PPVT. Respondents are presented
with a series of cards, each with four pictures on it, and asked to point to
the picture when a specific word or phrase is said. The receptive language
cards become more difficult, and the test has been standardized by age and
is appropriate for children as young as three years as well as for adults.16
Finally, to assess children’s health, we use data on children’s height,
weight, age, and gender to calculate their body mass index (BMI). We
then use the BMI to determine whether children are overweight or obese.
Following standard procedures, we categorize children in the 85th percentile of BMI or above as being overweight or obese.17
We begin by documenting the trends over time in child well-being
outcomes between ages three and nine, and how these differ by mother’s
education—our key measure of family disadvantage. Figures 8.1 and 8.2
child well-being209
Figure 8.1 Child Internalizing Behavior Problems
14
College +
Mean
12
10
Some college
8
High school
6
Less than
high school
4
2
1
3
5
(1999–2001) (2001–2003) (2003–2006)
9
(2007–2010)
Child’s Age-Year
Source: Authors’ calculations based on data from the Fragile Families and Child Wellbeing
Study.
Note: Thirty-two items in scale, range of 0 to 64.
show that both internalizing and externalizing problem behaviors decline
between age three and nine, as is expected from what is known about
child behavior. Declines are more pronounced for externalizing behavior
between five and nine years of age than for internalizing behavior. Children
of mothers with less education show somewhat higher levels of internalizing behaviors at younger ages, declines over child age being steeper for
Figure 8.2 Child Externalizing Behavior Problems
14
12
College +
Mean
10
Some college
8
High school
6
Less than
high school
4
2
1
3
5
(1999–2001) (2001–2003) (2003–2006)
9
(2007–2010)
Child’s Age-Year
Source: Authors’ calculations based on data from the Fragile Families and Child Wellbeing
Study.
Note: Thirty items in scale, range of 0 to 70.
210
children of the great recession
Figure 8.3 Child PPVT Scores
Mean
120
110
College +
100
Some college
90
High school
80
Less than
high school
70
60
1
3
5
(1999–2001) (2001–2003) (2003–2006)
9
(2007–2010)
Child’s Age-Year
Source: Authors’ calculations based on data from the Fragile Families and Child Wellbeing
Study.
Note: Thirty items in scale, range of 0 to 70. PPVT = Peabody Picture Vocabulary Test;
standardized to mean of 100 and standard deviation of 15.
children with less-educated mothers than for those with more-educated
mothers. Children of less-educated mothers show consistently higher average externalizing behaviors than their counterparts and these differences
are statistically significant.
Children’s scores on the PPVT (figure 8.3) are standardized by child age
and are therefore generally stable over time. As expected, unstandardized
scores (not shown) do increase as children age. The gradient by maternal
education in these scores is clear: children of more-educated mothers have
much higher average scores than their counterparts, and these differences
are also statistically significant. These results are consistent with previous
literature.
Finally, figure 8.4 shows rates of overweight or obesity. Interestingly,
our data show very little difference in average overweight-obesity rates
either by child age or by maternal education. Indeed, approximately
20 percent of children are overweight or obese regardless of their age or
their mothers’ education level.
LOCAL UNEMPLOYMENT RATES AND CHILD WELL-BEING
Next, we estimate the effects of the unemployment rate on child wellbeing to quantify those of a deep recession. In these models, we have
combined data from the surveys when children were three, five, and nine
years old to examine associations between the local unemployment rate
and children’s internalizing and externalizing behaviors, PPVT score, and
child well-being211
Percent
Figure 8.4 Child Overweight-Obese
0.50
0.45
0.40
0.35
0.30
0.25
0.20
0.15
0.10
0.05
0
College +
Some college
High school
Less than
high school
1
3
5
(1999–2001) (2001–2003) (2003–2006)
9
(2007–2010)
Child’s Age-Year
Source: Authors’ calculations based on data from the Fragile Families and Child Wellbeing
Study.
Note: The college and less than high school categories overlap. Overweight-obese is
based on children’s BMI, which is calculated using child age, gender, height, and weight.
Children with a BMI at or above the 85th percentile are defined as overweight-obese.
overweight-obesity. As in previous chapters, our models control for a host
of demographic characteristics, including the mother’s age, race-ethnicity,
relationship status at birth, immigrant status, whether the mother grew
up with both parents, survey year, and family’s city of residence. And,
as in previous chapters, we estimate both pooled regression models and
models that include individual fixed effects. The latter better account for
unobservable differences between children and are thus our preferred
models.
The estimates for the individual fixed-effects models presented in
table 8.A2 are used to predict what the well-being of children would be
given an increase in the unemployment rate from 5 percent to 10 percent,
which is approximately the size of the increase brought on by the Great
Recession. The predicted effects of a deep recession—both overall and
by mothers’ education level—are presented in figures 8.5 through 8.8.
However, the regression estimates on which these predictions are based
are not statistically significant: we do not in fact find statistically significant
effects of the unemployment rate in our fixed-effects models for the children
overall or for any of the maternal education subgroups.
Figure 8.5 displays the predicted effects of a deep recession on mothers’
report of their children’s internalizing behaviors, for all children and for
mothers’ education subgroups. A deep recession is not predicted to have
any effect on children’s internalizing behavior in the overall sample, or
in the maternal education subgroups. Figure 8.6 illustrates the predicted
effects on children’s externalizing behaviors. Overall, a deep recession is
212
children of the great recession
Figure 8.5 Child Internalizing Behaviors
14
12
Mean
10
8
6
–16%
0%
UR 5 percent
–14%
0%
0%
Some
college
College +
UR 10 percent
4
2
All
Less than
high school
High
school
Source: Authors’ calculations based on data from the Fragile Families and Child Wellbeing
Study.
Note: No significant differences in the effect of unemployment between subgroups. No
significant associations in individual fixed-effects models.
predicted to decrease children’s externalizing behavior score by about
1 point, or 9 percent, which is not significant. No effects are seen by maternal
education subgroup.
The predicted effects of a deep recession on children’s PPVT scores
are shown in figure 8.7. No overall effect is found for the total sample
or for any of the maternal education subgroups. However, the children
Figure 8.6 Child Externalizing Behaviors
14
12
–9%
0%
–9%
Mean
10
–10%
+11%
8
UR 5 percent
UR 10 percent
6
4
2
All
Less than
high school
High
school
Some
college
College +
Source: Authors’ calculations based on data from the Fragile Families and Child Wellbeing
Study.
Note: No significant differences in the effect of unemployment between subgroups. No
significant associations in individual fixed-effects models.
child well-being213
Mean
Figure 8.7 Child PPVT Scores, Unemployment Rates
120
110
100
90
80
70
60
+1%
0%
–2%
–3%
+1%
UR 5 percent
UR 10 percent
All
Less than
high school
High
school
Some
college
College +
Source: Authors’ calculations based on data from the Fragile Families and Child Wellbeing
Study.
Note: Chow tests show that the coefficient for unemployment for high school is different
from the coefficient for unemployment for the less than high school group. No significant
associations in individual fixed effects models.
of less-educated mothers are predicted to experience a slight decline (not
significant) in PPVT scores, and their counterparts a slight increase.
Figure 8.8 depicts the predicted effects of a deep recession on children
being overweight or obese. The figure shows an overall limited effect of a
deep recession on children being overweight or obese, and significant variability depending on the mothers’ level of education. (The figure shows
large effects among children of college-educated mothers, but the small
sample size for this subgroup limits the reliability of the estimate.)
Percent
Figure 8.8 Child Overweight-Obese, Unemployment Rates
50
45
40
35
30
25
20
15
10
5
0
UR 5 percent
+6%
All
+20%
–50%
–19%
–85%
Less than
high school
High
school
Some
college
College +
UR 10 percent
Source: Authors’ calculations based on data from the Fragile Families and Child Wellbeing
Study.
Note: No significant differences in the effect of unemployment between subgroups. No
significant associations in individual fixed-effects models.
214
children of the great recession
ADDITIONAL ESTIMATES
These results point to generally small or neutral effects of a deep recession,
some predicted to move in a positive direction (see tables 8.A1 through 8.A4).
That the results are not stronger and more negative may be somewhat surprising given research indicating that events such as mass layoffs, farm crises for
individual families, high parental unemployment, and economic uncertainty
can directly and adversely affect child well-being.18 One explanation for this
difference may be that children may be more affected by the uncertainty
associated with times of great economic upheaval than by the unemployment rate itself.19 The meaning of a given level of unemployment may differ
greatly depending on whether it represents the status quo, an improvement,
or a worsening. Separate work drawing on related theories demonstrates that
changes in national consumer confidence are associated with increases in
nine-year-olds’ behavior problems during the Great Recession.20
To examine the possible effects of economic uncertainty on child wellbeing, we estimate a supplemental model that uses increases and decreases
in the unemployment rate—as well as the level of the unemployment
rate—to predict each of the four outcomes. This spline model allows us
to test the idea that rapid changes in the unemployment rate, such as one
might expect to occur during a deep recession, are what affect child wellbeing, rather than the unemployment rate. These results are summarized
in table 8.A3. To estimate the effects of a deep recession, we use the fixedeffects estimates from the spline model to predict the effects of a rapid
increase in the unemployment rate, such as what occurred during the Great
Recession (table 8.A3 includes estimates without individual fixed effects).
These supplemental results do not provide any evidence that a rapid
increase or decrease in the unemployment rate affects children’s internalizing behavior. A rapid increase, however, does affect their externalizing
behavior. In the sample overall, and among children with less-educated
mothers (a high school diploma or less), an increase is predicted to increase
externalizing behaviors.
With regard to receptive language competency, a decreasing rate of
unemployment was significantly associated with a lower PPVT score in the
overall sample. This effect was also significant for mothers with less than a
high school education. No effects were seen for being overweight or obese.
Are the results from our main models altered if we also control for the
mother’s or the father’s unemployment? The association between rapid
increases in unemployment and higher externalizing behavior is still
significant, as is that between rapid decreases in unemployment and lower
PPVT scores.
We would also like to know whether the results from our main models
change if we estimate models isolating the effect of the unemployment
child well-being215
rate during the Great Recession when the children were nine years old.
In these analyses, we include a variable interacting the year nine survey,
when the Great Recession was under way, with the unemployment rate
(see table 8.A3). Significant interactions would tell us whether the links
between the unemployment rate and child well-being differed when children were age three and five versus age nine. Wave-age interactions were
found for internalizing and externalizing behaviors. Higher unemployment
during the Great Recession was associated with decreases in children’s
internalizing and externalizing behaviors. We can only speculate as to why.
It may be that children reacted to the strain of the Great Recession by
improving their behavior, or that their parents reacted by reporting fewer
behavior problems. It may also be a function of child age—it may be that
nine-year-olds (the group surveyed during the Great Recession) react
differently than younger children. We find no interactions for receptive
language or overweight-obesity.
Finally, we would like to know whether the results vary by mothers’
race-ethnicity or marital status at baseline. Although largely not significant, these estimates indicate considerable heterogeneity by race-ethnicity.
Most strikingly, we find that higher unemployment rates are associated with
significantly decreased internalizing behaviors and increased PPVT scores
among children of white mothers, but not with significant effects among
black or Hispanic mothers. We can only speculate as to what might explain
these results. One possibility is that when unemployment rates are higher,
white mothers are less likely to be employed, and that this in turn has beneficial effects for their children’s well-being. To test this hypothesis, we add a
control for the mother’s unemployment but do not find results to support it.
CONCLUSIONS
Our results indicate that a deep recession, such as the Great Recession,
might have some effects on children’s well-being, but results vary considerably by outcome, empirical specification, and subgroup. Research
on recessions and economic hardship would lead us to hypothesize that
child behavior problems would increase as the economic climate worsens.
However, we do not find this to be true. We do find some evidence that
rapidly increasing unemployment rates are associated with higher levels of
child externalizing behaviors. We posit that the uncertainty associated with
large increases in unemployment might be accounting for these findings.21
Such an interpretation would be in line with our findings about increases in
uncertainty as measured by the national consumer sentiment index being
associated with higher levels of externalizing behavior in nine-year-old boys
during the Great Recession.22 The analyses of nine-year-olds included selfreport measures of early juvenile delinquency in addition to maternal report
of externalizing behavior (because, by age nine, children can be effectively
216
children of the great recession
asked about their behavior), lending credence to the notion that the effect is
not accounted for only by maternal perceptions of her child’s behavior.
Little research addresses the effect of recessions on children’s language
and cognitive ability. Indeed, it is perhaps unclear how one might expect
recessions to affect children’s language development. If parents are severely
adversely affected by an economic downturn, it may be that parent-child
interactions are so changed that children’s language development would be
affected. However, it may also be that language development already under
way among children between the ages of three and nine (as in our study) is
less susceptible to sudden shocks than it would be among younger children.
Alternatively, higher unemployment rates might be related to mothers’
unemployment, which in turn might lead to positive effects on children’s
language development, if having mothers in the household increases the
learning and supportive resources available to children. Overall, we find little
support for the proposition that higher unemployment rates affect children’s
language development, though we do find some that higher unemployment
rates are associated with increased PPVT scores among children of white
mothers. One important caveat here is that our research design would not
capture any longer-term effects on children’s language development.
We also examine one child health outcome—overweight-obesity. The
literature investigating the effect of economic hardships and macro­
economic shocks on adult health is growing, but little looks specifically at
children. In general, this research provides mixed results: that economic
hardship may be associated with both better and worse health. We find
no evidence that unemployment rates are associated with child obesityoverweight. Again, however, we caution that our analysis focuses on
short-term effects.
It is difficult in this analysis to disentangle the influence of child age
from other forces. Indeed, the trends discussed earlier in this chapter
show the powerful influence of age in child development, such as the
dramatic decline in externalizing behaviors. Children at the year nine
survey are both older and being exposed to higher unemployment rates.
This patterning in the data could yield a spurious association between
higher unemployment rates and lower levels of externalizing behaviors.
To test this possibility, we reestimate our models controlling for child age
in months, or child age in half-year increments. These additional models
still suggest an association between higher unemployment rates and fewer
externalizing behaviors.
In many ways, our results on the unemployment rate at the local level
are not surprising. Children are powerfully influenced by their parents and
the parenting they receive. Results from chapter 7 indicate that higher
unemployment rates are generally associated with decreased rather than
increased harsh parenting. It stands to reason that the unemployment
rate, then, would not be associated with increases in behavior problems.
child well-being217
However, we also note the intriguing results indicating that rapidly
increasing unemployment rates may be associated with higher levels
of child externalizing behaviors as well as possibly higher frequencies of
harsh parenting. This finding may point to the effect of uncertainty on
children’s social-emotional well-being. Clearly, it will be important to
look at the longer-term effects of the Great Recession as these children
mature.
APPENDIX
Measures
Internalizing behavior problems. At each wave, mothers answered a series
of questions about their children’s internalizing behaviors, or innerfocused behaviors. The questions were drawn from three subscales of the
Achenbach Child Behavioral Check List (CBCL), which focuses on children’s anxious-depressed or withdrawn-depressed behaviors and somatic
complaints and includes thirty-two items.23 To be developmentally appropriate, the questions vary depending on the age of the child. We sum these
items creating a scale ranging from 0 to 64 (mean = 6.15, SD = 5.47).
Externalizing behavior problems. At each wave, mothers were asked a series
of questions about their children’s externalizing behaviors, or outwardfocused behaviors. The questions were drawn from the aggression and rulebreaking subscales of the CBCL.24 To be developmentally appropriate, the
questions again vary depending on the age of the child. We sum these items
creating a scale ranging from 0 to 70 (mean = 10.69, SD = 8.18).
Peabody Picture Vocabulary Test. The PPVT was administered to a sub­
sample of children who received in-home visits as part of the Fragile Families
Study. The test was administered by a home visitor to the focal child and is
a well-established measure of children’s receptive vocabulary, verbal ability,
and scholastic aptitude. As with the other measures in this chapter, the
content varies depending on the age of the child.
Overweight-obese. Children who received an in-home assessment also
had their height and weight measurements recorded. Using this infor­
mation and that about gender and birth date, we create indicators of
children’s body mass index. We rely on standard measures from the Center
for Disease Prevention and Control classifying children with BMI at or
above the 85th percentile as overweight or obese.
Key Independent Variable
For each analysis, the unemployment rate is constructed using a measure of
the average unemployment rate in the sample city over the twelve months
before the interview.
218
children of the great recession
Key Moderating Variables
We study differences in the trajectories over time, and in the effects of
the Great Recession, on child well-being stratified by maternal education at baseline. Mother’s education is coded as less than a high school
degree or the completion of a GED, a high school diploma, some college or an associate’s or technical degree, or a bachelor’s degree or
greater.
Control Variables
We include a number of covariates in our models, all measured at the first
survey wave (baseline). These include mother’s age at the birth, immigrant
status (foreign born), number of children in the household, a measure of
whether the mother was living with both biological parents at age fifteen,
as well as city (twenty dummies for each sample city) and survey year fixed
effects (twelve calendar year dummies).
Method
The figures that plot the trajectories of each outcome measured over time
present the mean levels of each outcome at each survey wave. All means are
weighted with the wave-specific city-weights to be representative of births
in the twenty study cities; the sample is restricted to parents interviewed in
all survey waves.
To study the effects of the Great Recession, we conduct linear probability models for our binary outcome and ordinary least squares regression analyses for continuous outcomes using the pooled data (years three
through five). We use linear probability models for ease of interpretation
but logit models provide very similar results (available on request). The standard errors are clustered at both the city and individual level to account for
within city and within person clustering–nonindependence. Analyses are
conducted for all children and separately for children with mothers with
less than high school, high school only, some college, or college degree or
greater. We estimated pooled models and also a parallel set of models with
child fixed effects.
To predict the effects of the Great Recession, we estimate the predicted
probability of each outcome when the unemployment rate is set at 5 percent, a rate typical of the period before the recession, and compare these
predictions with when the unemployment rate is set to 10 percent, a rate
typical of the Great Recession. We predict different probabilities for each
level of mother’s education.
child well-being219
Table 8.A1 Full Regression Results, Child Well-Being
Internalizinga
With Individual
Fixed Effects
Unemployment rate
Education
Less than high school
High school
Some college
Relationship status
Married
Cohabiting
Mother’s age
Race-ethnicity
Black
Hispanic
Other
Immigrant
Children in household
Lived with both parents
at age fifteen
Interview year
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
-0.11
—
—
0.49
-0.30
-2.78***
-2.97***
-1.68
—
-3.14***
-3.12***
—
(-1.39)
(—)
(—)
(0.65)
(-0.43)
(-3.70)
(-4.11)
(-0.91)
(—)
(-4.05)
(-4.55)
(—)
Without Individual
Fixed Effects
-0.09
(-0.88)
1.79***
-0.36
-1.19
(7.00)
(-1.85)
(-7.05)
-1.45
-0.3
-0.02
(-6.1)
(-1.9)
(-1.54)
-0.17
0.55*
0.75*
0.67
0.19***
-0.16
(-1.15)
(2.4)
(2.08)
(2.41)
(3.69)
(-1.79)
—
—
3.17***
2.74***
0.06
-0.03
(—)
(—)
(4.4)
(3.79)
(0.08)
(-0.04)
0.49
-0.45
-0.25
1.22
(0.52)
(-0.6)
(-0.28)
(0.9)
220
children of the great recession
Table 8.A1 Continued
Internalizinga
City
Austin
Baltimore
Detroit
Newark
Philadelphia
Richmond
Corpus Christi
Indianapolis
Milwaukee
New York
San Jose
Boston
Nashville
Chicago
Jacksonville
Toledo
San Antonio
Pittsburgh
Norfolk
Constant
Observations
Number of individuals
With Individual
Fixed Effects
Without Individual
Fixed Effects
8,297
3,861
1.16***
(8.66)
-0.15
(-0.46)
0.52
(1.33)
1.08**
(3.15)
0.73*
(2.24)
0.59
(1.62)
0.67
(1.58)
1.24**
(2.86)
1.20**
(2.73)
0.86*
(2.02)
0.61
(1.31)
0.32
(0.74)
0.54
(1.2)
0.28
(0.63)
-0.08
(-0.17)
(1.73)
0.79†
2.15***
(4.9)
1.48***
(3.39)
0.35
(0.73)
6.17***
(6.95)
8,297
4,487
Source: Authors’ calculations based on data from the Fragile Families and Child Wellbeing Study.
Note: Z-stats in parentheses. Covariates are measured at the baseline survey (except year) and are
clustered at city and individual level. Model 1 includes level unemployment rate. The model without
individual fixed effects is clustered at city and individual level.
a
Coefficients from OLS regressions; years three through nine only.
***p < .001; **p < .01; *p < .05; †p < .1
Internalizing-coefficients from OLS (z-stat)
Unemployment rate (model 1)
-0.11
(-1.39)
Unemployment rate (model 2)
-0.10
(-1.25)
Increasing unemployment rate
0.01
(1.00)
Decreasing unemployment rate
0.01
(0.84)
Observations
8,297
Number of individuals
3,861
Externalizing-coefficients from OLS (z-stat)a,c
Unemployment rate (model 1)
-0.08
(-0.80)
Unemployment rate (model 2)
-0.06
(-0.58)
Increasing rate of unemployment
0.02***
(3.30)
Decreasing rate of unemployment
0.03
(1.45)
Observations
8,320
Number of individuals
3,862
a,c
All
-0.02
(-0.12)
-0.03
(-0.19)
0.01
(1.20)
0.01
(-0.21)
2,135
960
-0.17
(-1.00)
-0.17
(-1.01)
0.02
(1.57)
0.01
(0.25)
2,150
960
-0.18
(-1.00)
-0.21
(-1.15)
0.03**
(2.92)
0.01
(0.29)
2,634
1,207
0.05
(0.24)
0.14
(0.65)
0.02†
(1.67)
0.08†
(1.77)
2,643
1,293
Some
College
-0.19
(-1.40)
-0.20
(1.48)
0.01
(0.65)
-0.02
(-0.57)
2,603
1,205
High
School
-0.18
(-1.00)
-0.13
(-0.69)
0.01
(0.41)
0.07†
(1.80)
2,667
1,295
Less than
High
School
With Individual Fixed Effects
0.11
(0.42)
0.08
(0.32)
-0.01
(-0.32)
-0.01
(-0.26)
893
402
0.02
(0.10)
0.04
(0.20)
-0.01
(-0.47)
0.03
(0.78)
892
401
College +
-0.04
(-0.45)
-0.02
(-0.17)
0.02***
(2.64)
0.02†
(1.62)
8,320
4,487
-0.09
(-0.88)
-0.09
(-0.77)
0.01
(1.02)
0.02
(1.50)
8,297
4,487
All
-0.47*
(-2.24)
-0.46**
(-2.58)
0.02*
(2.04)
-0.01
(-0.28)
2,634
1,368
0.23†
(1.74)
0.36†
(1.90)
0.04**
(2.68)
0.10**
(2.91)
2,643
1,542
0.07
(0.52)
0.06
(0.52)
0.01
(0.57)
-0.02
(-0.54)
2,150
1,090
0.06
(0.66)
0.05
(0.58)
0.01
(0.69)
-0.02
(-1.37)
2,135
1,090
Some
College
-0.07
(-0.32)
-0.06
(-0.26)
0.01
(0.39)
-0.01
(-0.10)
893
487
-0.06
(-0.36)
0.03
(0.13)
0.02
(1.01)
0.05
(0.72)
892
487
College +
(Table continues on p. 222.)
-0.32**
(-2.72)
-0.32**
(-2.68)
0.00
(0.32)
0.00
(-0.12)
2,603
1,368
High
School
-0.09
(-0.45)
-0.03
(-0.13)
0.01
(0.81)
0.08*
(2.45)
2,667
1,542
Less than
High
School
Without Individual Fixed Effects
Table 8.A2 Coefficients and Standard Errors, Rate of Change in Unemployment, Child Well-Being Outcomes
a,c
All
High
School
-0.01
(-1.35)
-0.01
(-1.32)
-0.00
(-0.50)
0.00
(1.37)
7,698
4,487
-0.03
(-1.37)
-0.03
(-1.33)
-0.00
(-0.36)
0.00
(0.20)
743
364
0.01
(0.69)
0.01
(0.68)
-0.00
(-0.32)
-0.00
(-0.18)
1,916
916
-0.04
(-1.48)
-0.04
(-1.49)
0.00
(0.67)
0.00
(-0.27)
2,542
1,368
-0.03**
(-2.71)
-0.03**
(-2.69)
-0.00
(-1.34)
-0.00
(-0.08)
2,396
1,368
0.01†
(1.71)
0.02†
(1.93)
0.00
(0.64)
0.00†
(1.90)
2,643
1,542
High
School
-0.05
(-1.57)
-0.05†
(-1.71)
0.00
(1.02)
-0.01
(-1.38)
2,683
1,542
Less than
High
School
0.00
(0.01)
0.00
(0.03)
-0.00
(-0.41)
0.00
(0.18)
1,916
1,090
-0.01
(-0.28)
-0.01
(-0.54)
-0.00
(-0.52)
-0.01
(-1.50)
2,023
1,090
Some
College
Without Individual Fixed Effects
-0.03
(-1.50)
-0.02
(-1.34)
0.00
(0.28)
0.00
(0.70)
743
487
-0.70
(-1.48)
-0.07
(-1.23)
0.00
(0.93)
-0.01
(-0.64)
809
487
College +
Source: Authors’ calculations based on data from the Fragile Families and Child Wellbeing Study.
Note: Z-stats in parentheses. Covariates are measured at the baseline survey (except year) and are clustered at city and individual level. Model includes level unemployment rate.
The model without individual fixed effects is clustered at city and individual level.
a
Coefficients from OLS regressions.
b
Coefficients from linear probability models.
c
Years three through nine only.
***p < .001; **p < .01; *p < .05; †p < .1
-0.04*
(-2.05)
-0.04*
(-2.39)
0.00
(0.50)
-0.01
(-1.41)
8,057
4,487
All
0.01
(0.13)
-0.00
(-0.02)
0.00
(0.21)
-0.01
(-0.95)
809
399
College +
0.04
(1.59)
0.33
(1.45)
-0.00
(-0.89)
-0.01
(-1.51)
2,023
951
Some
College
With Individual Fixed Effects
Less than
High
School
-0.00
-0.02
-0.03
(-0.09)
(-0.70) (-1.39)
Unemployment rate (model 2)
-0.01
-0.20
-0.03
(-0.39)
(-0.90) (-1.47)
Increasing rate of unemployment
-0.00
0.00
0.00
(-0.30)
(0.16)
(1.00)
Decreasing rate of unemployment
-0.01**
-0.01*
-0.00
(-2.88)
(-2.05) (-0.21)
Observations
8,057
2,683
2,542
Number of individuals
3,893
1,327
1,216
Overweight/obese-coefficients from linear probability models (z-stat)b,c
Unemployment rate (model 1)
-0.00
0.01
-0.02
(-0.37)
(0.79) (-1.49)
Unemployment rate (model 2)
-0.00
0.01
-0.02
(-0.28)
(0.98) (-1.51)
Increasing rate of unemployment
0.00
0.00
0.00
(0.15)
(0.35)
(0.20)
Decreasing rate of unemployment
0.00
0.00
-0.00
(0.71)
(1.37) (-0.27)
Observations
7,698
2,643
2,396
Number of individuals
3,740
1,291
1,169
PPVT-coefficients from OLS (z-stat)
Unemployment rate (model 1)
Table 8.A2 Continued
child well-being223
Table 8.A3 Sensitivity of Coefficients, Child Well-Being Outcomes
With Individual
Fixed Effects
Without Individual
Fixed Effects
Internalizing coefficients from OLS (z-stat)a,c
Unemployment rate (model 1)
-0.11
(-1.39)
-0.09
Unemployment rate (model 3)
0.00
(0.00)
-0.18
Mother’s unemployment
-0.08
(0.25)
0.60**
Bio-social fathers not employed
-0.12
(-0.39)
0.18
Unemployment rate (model 4)
0.06
(0.50)
-0.03
Unemployment rate * year nine
-0.22*
(-2.04)
-0.09
Externalizing coefficients from OLS (z-stat)a,c
Unemployment rate (model 1)
-0.08
(-0.80)
-0.04
Unemployment rate (model 3)
-0.12
(-0.28)
-0.31
Mother’s unemployment
0.05*
(2.05)
0.91†
Bio-social father’s not employed
0.08
(1.50)
0.33
Unemployment rate (model 4)
0.11
(0.78)
0.14
(-1.84)
-0.23*
Unemployment rate * year nine
-0.24 †
PPVT coefficients from OLS (z-stat)a,c
Unemployment rate (model 1)
-0.00
(-0.09)
-0.04*
Unemployment rate (model 3)
-0.06
(-1.17)
-0.03
Mother’s unemployment
-0.01
(-0.14)
-0.20***
Bio-social fathers not employed
0.07
(1.10)
-0.07*
Unemployment rate (model 4)
-0.01
(-0.25)
-0.05†
Unemployment rate * year 9
0.00
(0.26)
0.02
Overweight-obese coefficients from linear probability models (z-stat)b,c
Unemployment rate (model 1)
-0.00
(-0.37)
-0.01
Unemployment rate (model 3)
-0.00
(-0.13)
-0.00
Mother’s unemployment
-0.01
(-0.35)
0.10**
Bio-social fathers not employed
0.02
(1.23)
0.04
Unemployment rate (model 4)
0.01
(0.44)
0.00
Unemployment rate * year nine
-0.01
(-0.95)
-0.01
(-0.88)
(-0.97)
(2.83)
(0.91)
(-0.19)
(-1.10)
(-0.45)
(-0.75)
(1.89)
(0.86)
(0.90)
(-1.97)
(-2.05)
(-0.66)
(-4.80)
(-2.26)
(-1.75)
(0.87)
(-1.35)
(-0.15)
(2.79)
(1.29)
(0.15)
(-1.44)
Source: Authors’ calculations based on data from the Fragile Families and Child Wellbeing Study.
Note: Z-stats in parentheses. Covariates are measured at the baseline survey (except year) and are
clustered at city and individual level. Model includes level unemployment rate. The model without
individual fixed effects is clustered at city and individual level.
a
Coefficients from OLS regressions.
b
Coefficients from linear probability models.
c
Years three through nine only.
***p < .001; **p < .01; *p < .05; †p < .1
-0.18
(-1.02)
0.05
(0.22)
-0.02
(-0.80)
-0.02
(-1.17)
-0.28
(-1.61)
-0.01
(-0.70)
0.01
(0.42)
Hispanic
0.06
(0.41)
Black
Source: Authors’ calculations based on data from the Fragile Families and Child Wellbeing Study.
Note: Z-stats in parentheses The model includes individual fixed effects.
a
Coefficients from OLS regressions.
b
Coefficients from linear probability models.
c
Years three through nine only.
**p < .01; *p < .05
Internalizing coefficients from OLS (z-stat)
Unemployment rate
Externalizing coefficients from OLS (z-stat)a,c
Unemployment rate
PPVT coefficients from OLS (z-stat)a,c
Unemployment rate
Overweight-obese coefficients from linear probability models (z-stat)b,c
Unemployment rate
a,c
-0.02
(-1.27)
0.08**
(2.96)
-0.22
(-1.20)
-0.30*
(-2.13)
White
Table 8.A4 Coefficients and Standard Errors, Child Well-Being Outcomes by Subgroup
-0.01
(-0.87)
-0.01
(-0.38)
-0.16
(-0.88)
-0.05
(-0.31)
Married
0.01
(0.74)
-0.02
(-1.17)
-0.15
(-0.90)
-0.06
(-0.44)
Cohabiting
-0.01
(-0.73)
0.01
(0.70)
0.09
(0.49)
-0.13
(-0.90)
Single
child well-being225
NOTES
1. Craigie, Brooks-Gunn, and Waldfogel, 2010.
2. Conger, Conger, and Martin 2010; Elder 1974; McLoyd 1990, 1998.
3. Brooks-Gunn and Duncan 1997; McLoyd 1998.
4. Gershoff et al. 2007.
5. Leventhal and Brooks-Gunn 2000.
6. McLeod and Shanahan 1993; Conger and Elder 1994; Yeung, Linver, and
Brooks-Gunn 2002.
7. Not all research finds gender effects, however (Mistry et al. 2008).
8. Elder 1974.
9. Schneider, Waldfogel, and Brooks-Gunn 2015.
10. Leininger and Kalil 2014.
11. Ananat et al. 2011.
12. Anderson and Butcher 2006; Currie 2008.
13. Ruhm 2005.
14. Achenbach and Rescorla 2001.
15. On academic achievement, Masten et al. 2005; on school adjustment, Aunola,
Stattin, and Nurmi 2000; on substance use, King, Iacono, and McGue 2004;
on juvenile delinquency, Nagin and Tremblay 2003.
16. Dunn and Dunn 1981.
17. Troiano and Flegal 1998.
18. Ananat et al. 2011; Conger and Elder 1994; Leininger and Kalil 2014.
19. Kalil 2013; Gassman-Pines, Gibson-Davis, and Ananant 2015.
20. Schneider, Waldfogel, and Brooks-Gunn 2015.
21. See Lee et al. 2013.
22. Schneider, Waldfogel, and Brooks-Gunn 2015.
23. Achenbach and Rescorla 2001.
24. Ibid.; for a discussion of child behaviors, see also Craigie, Brooks-Gunn, and
Waldfogel 2010.
REFERENCES
Achenbach, Thomas M., and Leslie Rescorla. 2001. Manual for the ASEBA SchoolAge Forms & Profiles: An Integrated System of Multi-informant Assessment.
Burlington: University of Vermont, Research Center for Children, Youth, and
Families.
Ananat, Elizabeth O., Anna Gassman-Pines, Dania V. Francis, and Christina
M. Gibson-Davis. 2011. “Children Left Behind: The Effects of Statewide
226
children of the great recession
Job Losses on Student Achievement.” NBER working paper no. 17104.
Cambridge, Mass.: National Bureau of Economic Research.
Anderson, Patricia M., and Kristin F. Butcher. 2006. “Childhood Obesity: Trends
and Potential Causes.” Future of Children 16(1): 19–45.
Aunola, Kasia, Hakan Stattin, and Jari-Erik Nurmi. 2000. “Parenting Styles and
Adolescents’ Achievement Strategies.” Journal of Adolescences 23(2): 205–22.
Brooks-Gunn, Jeanne, and Greg J. Duncan. 1997. “The Effects of Poverty on
Children.” The Future of Children 7(2): 55–71.
Conger, Rand D., Katherine J. Conger, and Monica J. Martin. 2010. “Socio­
economic Status, Family Processes, and Individual Development.” Journal of
Marriage and Family 72(3): 685–704.
Conger, Rand D., and Glen H. Elder Jr. 1994. “Families in Troubled Times: The
Iowa Youth and Families Project.” In Families in Troubled Times: Adapting to
Change in Rural America, edited by Rand D. Conger and Glen H. Elder Jr.
Hillsdale, N.J.: Aldine.
Craigie, Terry-Ann, Jeanne Brooks-Gunn, and Jane Waldfogel. 2010. “Family
Structure, Family Stability and Early Child Wellbeing.” CRCW working paper
no. WP10–14-FF. Princeton, N.J.: Princeton University.
Currie, Janet. 2008. “Healthy, Wealthy, and Wise: Socioeconomic Status, Poor
Health in Childhood, and Human Capital Development.” NBER working
paper no. 13987. Cambridge, Mass.: National Bureau of Economic Research.
Dunn, Lloyd M., and Douglas M. Dunn. 1981. Peabody Picture Vocabulary Test,
rev. ed. Circle Pines, Mn.: American Guidance Service.
Elder, Glen H., Jr. 1974. Children of the Great Depression: Social Changes in Life
Experience. Boulder, Colo.: Westview Press.
Gassman-Pines, Anna, Christina M. Gibson-Davis, and Elizabeth O. Ananant.
2015. “How Economic Downturns Affect Children’s Development: An
Interdisciplinary Perspective on Pathways of Influence.” Child Development
Perspectives 9(4): 233–38.
Gershoff, Elizabeth T., J. Lawrence Aber, C. Cybele Raver, and Mary Clare
Lennon. 2007. “Income Is Not Enough: Incorporating Material Hardship
into Models of Income Associations with Parenting and Child Development.”
Child Development 78(1): 70–95.
Kalil, Ariel. 2013. “The Effects of the Great Recession on Child Development.”
Annals of the American Academy of Political Science 650(1): 232–49.
King, Serena M., William G. Iacono, and Matt McGue. 2004. “Childhood
Externalizing and Internalizing Psychopathology in the Prediction of Early
Substance Use.” Addiction 99(12): 1548–59.
Lee, Dohoon, Jeanne Brooks-Gunn, Sara S. McLanahan, Daniel Notterman,
and Irwin Garfinkel. 2013. “The Great Recession, Genetic Sensitivity, and
Maternal Harsh Parenting.” Proceedings of the National Academy of Science
110(34): 13780–84.
Leininger, Lindsey J., and Ariel Kalil. 2014. “Economic Strain and Children’s
Behavior in the Aftermath of the Great Recession.” Journal of Marriage and
Family 76(5): 998–1010.
Leventhal, Tama, and Jeanne Brooks-Gunn. 2000. “The Neighborhoods They
Live In: The Effects of Neighborhood on Child and Adolescent Outcomes.”
Psychological Bulletin 126(2): 309–37.
Masten, Ann S., Glenn I. Roisman, Jeffery D. Long, Keith B. Burt, Jelena
Obradovic, Jennifer R. Riley, Kristen Boelcke-Stennes, and Auke Telgen. 2005.
child well-being227
“Developmental Cascades: Linking Academic Achievement and Externalizing
and Internalizing Symptoms over 20 years.” Developmental Psychology 41(5):
733–46.
McLeod, Jane D., and Michael J. Shanahan. 1993. “Poverty, Parenting, and
Children’s Mental Health.” American Sociological Review 58(3): 351–66.
McLoyd, Vonnie C. 1990. “The Impact of Economic Hardship on Black
Families and Children: Psychological Distress, Parenting, and Socioemotional
Development.” Child Development 61(2): 311–46.
———. 1998. “Socioeconomic Disadvantage and Child Development.” American
Psychologist 53(2): 185–204.
Mistry, Rashmita S., Edward D. Lowe, Aprile D. Benner, and Nina Chien. 2008.
“Expanding the Family Economic Stress Model: Insights from a MixedMethods Approach.” Journal of Marriage and Family 70(1): 196–209.
Nagin, Daniel, and Richard E. Tremblay. 2003. “Trajectories of Boy’s Physical
Aggression, Opposition, and Hyperactivity on the Path to Physically Violent
and Nonviolent Juvenile Delinquency.” Child Development 70(5): 1181–96.
Ruhm, Christopher J. 2005. “Healthy Living in Hard Times.” Journal of Health
Economics 24(2): 341–63.
Schneider, William, Jane Waldfogel, and Jeanne Brooks-Gunn. 2015. “The Great
Recession and Behavior Problems in 9-Year Old Children.” Developmental
Psychology 51(11): 1615–29.
Troiano, Richard P., and Katherine M. Flegal. 1998. “Overweight Children
and Adolescents: Description, Epidemiology, and Demographics.” Pediatrics
101(3, pt 2): 497–504.
Yeung, Jean W., Mariam R. Linver, and Jeanne Brooks-Gunn. 2002. “How
Money Matters for Young Children’s Development: Parental Investment and
Family Processes.” Child Development 73(6): 1861–79.
Index
Boldface numbers refer to figures and tables.
American Community Survey (ACS), 23
American Recovery and Reinvestment
Act of 2009 (ARRA), 60–61
Anderson, Elijah, 150
Bitler, Marianne, 61
Brooks-Gunn, Jeanne, 18–20
Canada, 25
Charles, Kerwin, 103
children: born to unwed mothers, 3; maltreatment of, 175; parenting of (see
parenting of mothers and fathers)
children’s well-being, 19–20, 206–7,
215–17; additional estimates,
214–15; economic uncertainty and,
214; externalizing behavior problems by child age-year and education, 209; externalizing behavior
problems by unemployment rate and
education, 212; internalizing behavior problems by child age-year and
education, 209; internalizing behavior problems by unemployment rate
and education, 212; local unemployment rates and, 207, 210–13; measures of, 217; method for the study
of, 218; overweight-obese by child
age-year and education, 211; overweight-obese by unemployment rate
and education, 213; PPVT scores
by child age-year and education,
210; PPVT scores by unemployment
rate and education, 213; previous
research on the effects of economic
hardship and the Great Recession,
207–8; trends in internalizing and
externalizing behaviors from ages
three to nine, 208–10; variables for
the study of, 217–18; well-being, full
regression results, 219–20; well-being
outcomes, sensitivity of co­efficients,
223; well-being outcomes and rate of
change in unemployment, co-efficients
and standard errors, 221–22; wellbeing outcomes by subgroup, coefficients and standard errors, 224
child support, 150–52. See also nonresident
father involvement
Conflict Tactics Scale, 175
Conger, Rand D., 2–3, 21, 174
consumer confidence index, 47
Current Population Survey (CPS): Child
Support Supplement, 152; sample
size, 23
Currie, Janet, 16, 90, 103
data: demographic changes, accounting
for, 3–4; Fragile Families and Child
Wellbeing Study as basic, 3 (see also
Fragile Families and Child Wellbeing
Study); limitations of, 23–24; procedure for looking at, 9–11
DeCicca, Philip, 103
disadvantaged families: data on, 7–8; disadvantage, measure of, 7–8; economic
well-being, impact of recessions on,
45; fragile families sample composition
by mother’s education, 8; policy implications of findings for, 26. See also
parents’ relationship status and quality
divorce rates, economic conditions and,
119–20
dot-com recession, 2, 5–6
double-dip recession, 5
doubling up, 15, 59–60, 64–68, 73
Duque, Valentina, 16, 90
Earned Income Tax Credit (EITC), 4,
14–15, 32, 46, 58–64, 67–69, 73
index229
economic insecurity: effects of the Great
Recession on, 44, 44–45; of families
with children from birth through age
nine, 38, 40; hardship rates of families by child age-year and mother’s
education, 39; material hardship, full
regression results, 48–49; as a measure of economic well-being, 32–33,
46; public and private transfers,
impact of, 72
economic well-being, 13–14; economic
outcomes, coefficients and standard
errors, 54; economic outcomes, sensitivity of coefficients, 53; economic
outcomes rate of change, coefficients and standard errors, 50–52;
employment by education and local
unemployment rate, 42; of families
with children from birth through age
nine, 35–41; Great Recession, impact
of, 33, 41–45; hardship by education and child age-year, 39; hardship
by education and local unemployment rate, 44; household income,
big gains and losses by education
and child age-year in, 38; household
income by education and child ageyear, 37; household income by raceethnicity, parents’ relationship status,
and child age-year, 40; income by
education and local unemployment
rate, 43; income by race-ethnicity,
parents’ relationship status, and
local unemployment rate, 43;
material hardship, full regression
results, 48–49; maternal employment by education and child ageyear, 35; measures of, 31–33, 46;
paternal employment by education
and child age-year, 36; poverty rate
by education and child age-year,
39; poverty rate by education and
local unemployment rate, 44; prior
research on recessions and, 34–35;
supplemental analyses of, 47
education: children’s well-being and,
209–13; economic insecurity and,
38–40, 44, 44–45; economic wellbeing and, 13–14, 33; effects of
Great Recession and, 20–21, 45;
employment and, 35–36, 42; health
and, 16, 90–103; income and,
36–37, 42, 43, 58; as a measure
of well-being, 10; of mothers as
measure of disadvantage, 7–8; non-
resident father involvement and, 18,
153–58; parenting of mothers and
fathers and, 18, 176–82, 185–86;
parents’ relationships and, 17,
121–26, 128–34; policy implications
of findings for, 25–26; poverty and,
38–39, 42–44, 44; public and private
transfers and, 15, 63–72
EITC. See Earned Income Tax Credit
Elder, Glen H., Jr., 1–3, 21, 174
employment: by education, 42; as a measure of economic well-being, 31–32,
46; for mothers and fathers with children from birth through age nine,
35–36. See also local unemployment
rates; unemployment and joblessness
ethnicity. See race and ethnicity
families: data on, richness of, 8; dis­
advantaged (see disadvantaged
families); economic being of,
with children from birth through
age nine, 35–41; effects of Great
Recession on, 20–22; instability in,
3; involvement in childrens’ lives of
nonresident fathers (see nonresident
father involvement); parenting (see
parenting of mothers and fathers);
parents’ relationship (see parents’
relationship status and quality)
family stress model: future research on, 24;
implications of findings for, 22–23; as
legacy of Elder study, 2; marital quality and economic crises, relationship
of, 120; natural experiments, testing
with, 2–3; parenting and economic
hardship, relationship of, 207
Farm Crisis Study, 2
fathers: binge drinking, by education and
child age-year, 94; binge drinking,
effects of a recession on, 100; drug
use, by education and child age-year,
95; drug use, effects of a recession
on, 101; employment, simulated
effects of the Great Recession on,
41–42; employment of by education level, 36; health outcomes and
behavior, effects of uncertainty on,
101; health outcomes and behavior
by child age-year, 90–96; health
outcomes and behavior during recessions, 96–100; health outcomes and
behavior during recessions, conclusions regarding, 102–4; health
outcomes and behavior of, 16,
230
children of the great recession
fathers (continued)
88–89; health outcomes by education, coefficients and standard errors,
110–12; health problems that limit
work, by education and child ageyear, 93; health problems that limit
work, effects of a recession on, 98;
health status, effects of a recession
on, 97; health status is fair or poor,
by education and child age-year,
92; involvement in childrens’ lives
of nonresident (see nonresident
father involvement); mothers’ supportiveness by education and local
unemployment rate, 130; parent­ing of, 19, 178–80, 185–86 (see also
parenting of mothers and fathers);
positive sample selection, impact
of, 146n16; relationship with bio
mother by education and local
unemployment rates, 132; reports
of bio mothers’ supportiveness
by education and child age-year,
124; reports of relationship with
bio mother by education and child
age-year, 126. See also parenting of
mothers and fathers; parents’ relationship status and quality
FFS. See Fragile Families and Child
Wellbeing Study
fixed-effects models, 11–12
food insecurity, 72
food stamps. See Supplemental Nutrition
Assistance Program (SNAP)
foreclosure rate, 47
fragile families. See disadvantaged families
Fragile Families and Child Wellbeing
Study (FFS): as basis for analyses, 3;
data from, 6–9; health, as source on,
90, 104; limitations of, 23–24;
material hardship, questions measuring, 32–33; nonresident fathers, as
source on, 153, 160; parenting of
mothers and fathers, as source on,
174; public and private financial transfers, as source on, 61; relationship status and quality, as source on, 120, 133
future research, 24
Garfinkel, Irwin, 13–15, 90, 103, 152
Great Depression Study, 1–2
Great Recession: economic well-being
and, 34–35, 41–45 (see also economic well-being); effects on families, 20–22; estimating the effect of
(see local unemployment rate); health
impact during, 102; natural experiments based on, 2; public and private
transfers, effect on, 67–70; public
and private transfers during the,
impact of, 70–72; severity of, 4–6;
unemployment, income, and length
of, 5–6
Harknett, Kristen, 17
Hatchett, Shirley, 120
health: behaviors, patterns of for parents
by child age-year, 93–96; behaviors,
patterns of for parents with increase
in unemployment rate, 99–100;
children’s, 208, 210, 211, 213, 216;
fathers’ binge drinking by education and child age-year, 94; fathers’
binge drinking by education and
local unemployment rate, 100; of
fathers by education and child ageyear, 92; of fathers by education and
local unemployment rate, 97; fathers’
drug use by education and child
age-year, 95; fathers’ drug use by
education and local unemployment
rate, 101; during Great Recession
years, 102; measures of, 104–5;
method used in study of, 105,
115; of mothers, impacts of the
2007 recession on, 90; mothers’
binge drinking by education and
child age-year, 94; mothers’ binge
drinking by education and local
unemployment rate, 99; of mothers
by education and child age-year, 91;
of mothers by education and local
unemployment rate, 97; mothers’
drug use by education and child
age-year, 95; mothers’ drug use by
education and local unemployment
rate, 100; outcomes, coefficients and
standard errors, 114; outcomes by
maternal education, coefficients and
standard errors, 108–9; outcomes by
paternal education, coefficients and
standard errors, 110–12; of parents,
16, 88–89; of parents, sensitivity of
coefficients, 113; of parents during recessions, conclusions regarding, 102–4; physical, full regression
results, 106–7; physical, patterns of
for parents by child age-year, 91–93;
physical, patterns of for parents with
increase in unemployment rate,
index231
96–99; problems of fathers that
limit work by education and local
unemployment rate, 98; problems
of mothers that limit work by education and local unemployment rate,
98; problems that limit fathers’ work
by education and child age-year, 93;
problems that limit mothers’ work
by education and child age-year, 92;
race-ethnicity, differential impacts
due to, 102; recessions and, prior
research on, 89–90; relationship status, differential impacts due to, 102;
uncertainty, effects of, 100–101;
unemployment of the individual and,
101–2; variables in analyses, 105
hedonic adaptation theory, 12
Hispanics. See race and ethnicity
housing: assistance, cash and in-kind, 58;
assistance, federally funded, 59–60,
62–64, 73; doubling up, 15, 59–60,
64–68, 73
Hoynes, Hilary, 61
income: big gains and losses by families
with children by child age-year, 38;
education level and, 36–37, 42,
43, 58; by education level and local
unemployment rate, 43; effects
of public and private transfers on
household, 71; of families by child
age-year, race-ethnicity,and parents’
relationship status, 40; of families
with children from birth through age
nine, 36–38; family, effects of the
Great Recession on, 42; household
as a measure of economic well-being,
32, 46; household by education and
child age-year, 37; median household
income index, 2000–2014, 6; the
recessions of the twenty-first century
and, 5–6, 34–35
Kuka, Elira, 61
local unemployment rates: children’s wellbeing and, 19–20, 207, 210–13; city
of birth vs. current residence, 27n18;
economic well-being and, 13–14,
41–42, 47; FFS data and, combining, 8–9; health outcomes/behaviors
and, 16, 96–100; individual-level
measures and aggregate rates, 27n23;
during interviewing periods, 9; nonresident father involvement and, 18,
155–60; parenting of mothers and
fathers and, 19, 180–88; parents’
relationship and, 17, 126–33; public
and private transfers and, 15, 67–70,
74–75; role in fixed-effects models
of, 12–13. See also unemployment
and joblessness
marriage, economic conditions and rates
of, 119
material hardship. See economic insecurity
McLanahan, Sara, 17
Medicaid, 4, 14–15, 58–64, 67–69, 73
method/methodology: children’s wellbeing, used in study of, 218; fixedeffects models, 11–13; health of
mothers and fathers, used in study
of, 105, 115; natural experiments to
address omitted variable bias, 2–3,
11, 22; nonresident father involvement, used in study of, 162–63; parenting of mothers and fathers, used
in study of, 190; parents’ relationship
status and quality, used in study of,
135–36
Mills, Bradford, 72
Mincy, Ronald, 17–18
mothers: binge drinking, effects of a recession on, 99; binge drinking by education and child age-year, 94; drug use,
effects of a recession on, 100; drug
use by education and child age-year,
95; educational attainment as measure of disadvantage, 7–8; education
of, policy implications of findings for,
25–26; effects of Great Recession on
relationship status, 126–29; employment, simulated effects of the Great
Recession on, 41–42; employment
by education level, 35; fathers’ supportiveness by education and local
unemployment rate, 130; health,
impacts of the 2007 recession on,
90; health outcomes and behavior,
effects of uncertainty on, 101; health
outcomes and behavior during recessions, 96–100; health outcomes and
behavior during time with children
from birth through age nine, 90–96;
health outcomes and behavior of, 16,
88–89; health outcomes and behavior of during recessions, conclusions
regarding, 102–4; health outcomes
by education, coefficients and standard errors, 108–9; health problems
232
children of the great recession
mothers (continued)
that limit work, effects of a recession
on, 98; health problems that limit
work by education and child ageyear, 92; health status, effects of a
recession on, 97; health status is fair
or poor by education and child ageyear, 91; labor force participation of,
3; marriage and marriage or cohabitation by local unemployment rate,
127; marriage bio father or new
partner by education and local
unemployment rate, 128; marriage
or cohabitation by education and
local unemployment rate, 128; marriage or cohabitation to bio fathers
or new partners by education and
child age-year, 122; marriage to bio
fathers or new partners by education
and child age-year, 122; married
to or cohabiting with father or new
partner, full regression results, 137–
38; new partners’ supportiveness by
education and local unemployment
rate, 131; parenting of, 18–19,
176–78, 181–85 (see also parenting
of mothers and fathers); relationship
status by child age-year, 121; relationship with bio father by education
and local unemployment rates, 132;
reports of fathers’ supportiveness by
education and child age-year, 123;
reports of new partners’ supportiveness by education and child age-year,
125; reports of relationship with
bio father by education and child
age-year, 125; sample composition,
education and, 8. See also parents’
relationship status and quality
Mykerezi, Elton, 72
natural experiments, 2–3, 11
Nepomnyaschy, Lenna, 152
nonresident father involvement, 17–18,
149, 159–60; child support and
visitation, coefficients and standard
errors, 169; child support and visitation, effects of the Great Recession
on, 155–59; child support and
visitation, full regression results,
164–65; child support and visitation,
sensitivity of coefficients, 168; child
support and visitation by education,
155; empirical evidence regarding,
151–52; father engagement by education and child age-year, 154; father
involvement rate of change, coefficients and standard errors, 166–67;
financial support and visitation,
impact of unemployment on, 150;
formal child support per year by education, 156; informal child support
per year by education, 157; in-kind
child support per year by education,
157; measures of child support outcomes, 160–61; measures of visitation
outcomes, 161; method for the study
of, 162–63; mother’s unemployment and, 159; nonresidence status
by education and child’s age-year,
153; share of nonresident fathers
visiting their children by education,
158; stress of economic adversity
and, 157–58; trends in child-support
orders, payments, and visitation during period from children’s birth to
age nine, 152–55; variables for the
study of, 161–62; visitation and
child support, reciprocal relationship
between, 169n11; visitation days per
month by education, 158
parenting of mothers and fathers, 18–19,
173–74, 188–89; economic hardship and, impact on children of, 207;
maternal parenting, full regression
results, 191–92; maternal parenting activities by education and child
age-year, 179; maternal parenting
activities by unemployment rate and
education, 184; maternal parenting
outcomes, sensitivity of coefficients,
196; maternal parenting outcomes
and rate of change in unemployment, coefficients and standard
errors, 193–95; maternal parenting
outcomes by subgroup, coefficients
and standard errors, 197; maternal
physical aggression by education and
child age-year, 177; maternal physical aggression by unemployment
rate and education, 182; maternal
psychological aggression by education and child age-year, 177; maternal psychological aggression by
unemployment rate and education,
183; maternal spanking by education
and child age-year, 176; maternal
index233
spanking by unemployment rate and
education, 182; maternal warmth by
education and child age-year, 178;
maternal warmth by unemployment
rate and education, 183; measures
of, 189; method used in study of,
190; paternal parenting, full regression results, 198–99; paternal
parenting outcomes, sensitivity of
coefficients, 202; paternal parenting outcomes and rate of change
in unemployment, coefficients and
standard errors, 200–201; paternal
parenting outcomes by subgroup,
coefficients and standard errors,
203; paternal physical aggression
by education and child age-year,
180; paternal physical aggression by
unemployment rate and education,
186; paternal psychological aggression by education and child age-year,
181; paternal psychological aggression by unemployment rate and
education, 186; paternal spanking by
education and child age-year, 179;
paternal spanking by unemployment
rate and education, 185; previous
research on, 174–75; trends in with
children ages three to nine, 175–80;
unemployment rates, impact of with
children ages three to nine, 180–88;
unemployment rates and, rapidly
changing, 184–85; variables in study
of, 189–90
parents’ relationship status and quality,
17, 118–19, 133–34; differential
effects of recessions by, 45; effects
of Great Recession on relationship
quality, 129–33; effects of Great
Recession on relationship status,
126–29; fathers’ reports of bio
mothers’ supportiveness by education and child age-year, 124; fathers’
reports of mothers’ supportiveness
by education and local unemployment rate, 130; fathers’ reports of
relationship with bio mother by
education and child age-year, 126;
fathers’ reports of relationship with
bio mother by education and local
unemployment rates, 132; health
effects of, 102; household income
by child age-year and, 40; income
loss by mothers during the Great
Recession and, 42; income loss of
mothers by local unemployment rate,
43; married to or cohabiting with
father or new partner, full regression
results, 137–38; measures of relationship quality, 134–35; measures of
relationship status, 134; method for
study of, 135–36; mothers’ marriage
and marriage or cohabitation by local
unemployment rate, 127; mothers’
marriage bio father or new partner
by education and local unemployment rate, 128; mothers’ marriage or
cohabitation by education and local
unemployment rate, 128; mothers marriage or cohabitation to bio
fathers or new partners by education
and child age-year, 122; mothers
marriage to bio fathers or new partners by education and child age-year,
122; mothers relationship status
by child age-year, 121; mothers’
reports of fathers’ supportiveness by
education and child age-year, 123;
mothers’ reports of fathers’ supportiveness by education and local
unemployment rate, 130; mothers’
reports of new partners’ supportiveness by education and child age-year,
125; mothers’ reports of new partners’ supportiveness by education
and local unemployment rate, 131;
mothers’ reports of relationship
with bio father by education
and child age-year, 125; mothers’
reports of relationship with bio
father by education and local
unemployment rates, 132; recessions and romantic relationships,
119–20; relationship outcomes,
coefficients and standard errors for
unemployment rate, 139–42, 145;
relationship outcomes, sensitivity
of unemployment rate coefficients,
143–44; supplemental analyses of,
136; trends during period from children’s birth to age nine, 120–26;
variables for study of, 135
Patterson, Richard, 120
Peabody Picture Vocabulary Test (PPVT),
207–8, 210, 212–15
Piketty, Thomas, 5
Pilkauskas, Natasha, 13–15, 103
policy implications, 25–26
234
children of the great recession
poverty: children, effect on, 207; of
families with children from birth
through age nine, 38–39; the
Great Recession and, 34–35,
42–44; as a measure of economic
well-being, 32, 46; mitigating
effects of public and private transfers on, 71; official threshold for,
32; rate by education and local
unemployment rate, 44; rates of
families with children from birth
through age nine by mother’s
education, 39
PPVT. See Peabody Picture Vocabulary
Test
private cash/financial transfers, 15,
59–60, 64–68, 72–73
provider role strain, 150
public and private transfers, 14–15,
58–60; coefficients and standard
errors, 84; doubling up by education and child’s age-year, 65; effects
of the Great Recession on, 67–70;
effects of transfers on household
income by education, 71; expanded
by the American Recovery and
Reinvestment Act of 2009, 60;
helping effects of, 70–72; measures of, 73–74; mitigating effects
of transfers on poverty by education, 71; previous research on the
Great Recession and, 61; private
assistance, average dollar value of
by education, 66; private financial
transfers and doubling up by local
unemployment rate and education,
68; private financial transfers by
education and child age-year, 65;
public assistance benefits, average
dollar value of, 64; public assistance
receipt by child age-year, 62; public assistance receipt by education,
63; public transfer receipt rates by
local unemployment rate and education, 69; rate of change for, coefficients and standard errors, 77–81;
received during child’s age one to
nine, 61–67; sensitivity of coefficients, 82–83; SNAP, full regression
results for, 75–76; supplemental
analyses, 74–75
race and ethnicity: children’s well-being
and unemployment rates, impact on
the relationship of, 215; differential
14371-09-Index-5thPgs.indd 234
effects of recessions by, 45; health
effects differentiated by, 102; household income by child age-year and,
40; income loss during the Great
Recession of mothers by, 42; income
loss of mothers by local unemployment rate, 43; parenting practices
and, 183–84, 187–89; policy implications of findings for, 26
recession(s): definition of, 4; double-dip,
unemployment during, 5; economic
well- being and, 34–35 (see also economic well-being); health and (see
health); romantic relationships and,
119–20 (see also parents’ relationship
status and quality). See also dot-com
recession; Great Recession
Saez, Emmanuel, 5
Schneider, Daniel, 17
Schneider, William, 18–20
Seltzer, Judith, 152
SNAP. See Supplemental Nutrition
Assistance Program
SSI. See Supplemental Security Income
State Children’s Health Insurance
Program, 4
Supplemental Nutrition Assistance
Program (SNAP), 4, 14–15, 32, 46,
58–64, 67–69, 73, 75–76
Supplemental Security Income (SSI),
14–15, 58–60, 62–64, 68, 73
TANF. See Temporary Assistance for
Needy Families
Temporary Assistance for Needy Families
(TANF), 14–15, 58–60, 62–64,
67–69, 73
Toledo, Elia De la Cruz, 17–18
unemployment and joblessness: child
support compliance and, 151–52;
the dot-com recession and, 5; effects
of during the Great Recession,
21–22; fathers’ financial support for
and visitation of children, impact
on, 150; future research on, 24; the
Great Recession and, 4–6, 34–35;
health effects of, 101–2; informal
child support compliance and, 152;
local unemployment rates (see local
unemployment rates); nonresident
fathers’ visitation of children and,
152; unemployment rate, 2000–
2014, 6
8/5/16 12:17 PM
Unemployment Insurance (UI), 14–15,
58–64, 73
Veum, Jonathan, 152
Waldfogel, Jane, 18–20
welfare state programs: policy implications of findings for, 25; safety
net, better-developed, 4; safety-
index235
net transfers, 14. See also public
and private transfers
women: joblessness during the Great
Recession, 4–5; labor force participation of, 3; as mothers (see mothers).
See also parenting of mothers and
fathers; parents’ relationship status
and quality
Wu, Chi-Fang, 151