Children of the Great Recession

Irwin Garfinkel, Sara McLanahan, and Christopher Wimer EDITORS 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 Lawrence F. Katz Claude M. Steele Karen S. Cook David Laibson Shelley E. Taylor W. Bowman Cutter III Nicholas Lemann Richard H. Thaler Sheldon Danziger Martha Minow Hirokazu Yoshikawa Kathryn Edin Peter R. Orszag 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/FZsS738Acy1J5xYsUOYUYMrKrjKqenPhOCYYCqejqtPhO- qekTzhOyMrjKqenPhOCYYOyrhhKUqen6m7AjqKNJVZgl6CAvU02rzjifY01dSh 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, photocopy- ing, 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 1 Irwin Garfinkel, Sara McLanahan, and Christopher Wimer

Chapter 2 Economic Well-Being 31 Irwin Garfinkel and Natasha Pilkauskas

Chapter 3 Public and Private Transfers 58 Natasha Pilkauskas and Irwin Garfinkel

Chapter 4 Mothers’ and Fathers’ Health 88 Janet Currie and Valentina Duque

Chapter 5 Parents’ Relationships 118 Daniel Schneider, Sara McLanahan, and Kristen Harknett

Chapter 6 Nonresident Father Involvement 149 Ronald B. Mincy and Elia De la Cruz Toledo

Chapter 7 Mothers’ and Fathers’ Parenting 173 William Schneider, Jane Waldfogel, and Jeanne Brooks-Gunn

Chapter 8 Child Well-Being 206 William Schneider, Jane Waldfogel, and Jeanne Brooks-Gunn

Index 228

List of Illustrations

Figure 1.1 Median Household Income Index and Unemployment Rate 6 Figure 1.2 Local Unemployment Rates During Interviewing Periods 9 Figure 2.1 Maternal Employment 35 Figure 2.2 Paternal Employment 36 Figure 2.3 Household Income ($2010) 37 Figure 2.4 Big Gains and Losses 38 Figure 2.5 Poverty Rates 39 Figure 2.6 Hardship (Insecurity) Rates 39 Figure 2.7 Household Income ($2010) by Race-Ethnicity 40 Figure 2.8 Employment by Education 42 Figure 2.9 Income by Education 43 Figure 2.10 Income by Race-Ethnicity and Relationship Status 43 Figure 2.11 Poverty Rate by Education 44 Figure 2.12 Hardship by Education 44 Figure 3.1 Public Assistance Receipt by Child’s Age-Year 62 Figure 3.2 Public Assistance Receipt by Education 63 Figure 3.3 Average Dollar Value of Public Assistance Benefits 64 Figure 3.4 Private Financial Transfers ($2010) 65 Figure 3.5 Doubling Up 65 Figure 3.6 Average Dollar Value of Private Assistance 66 Figure 3.7 Private Financial Transfers and Doubling Up 68 Figure 3.8 Public Transfer Receipt Rates 69 Figure 3.9 Effects of Transfers on Household Income 71 Figure 3.10 Mitigating Effects of Transfers on Poverty 71 Figure 4.1 Mothers’ Health Status Is Fair or Poor 91 Figure 4.2 Fathers’ Health Status Is Fair or Poor 92 Figure 4.3 Mothers’ Health Problem that Limits Work 92 Figure 4.4 Fathers’ Health Problem that Limits Work 93 Figure 4.5 Mothers’ Binge Drinking 94 Figure 4.6 Fathers’ Binge Drinking 94 Figure 4.7 Mothers’ Drug Use 95 Figure 4.8 Fathers’ Drug Use 95 Figure 4.9 Effects of a Recession on Mothers’ Health Status 97 Figure 4.10 Effects of a Recession on Fathers’ Health Status 97 Figure 4.11 Effects of a Recession on Mothers’ Health Problem that Limits Work 98 Figure 4.12 Effects of a Recession on Fathers’ Health Problem that Limits Work 98 Figure 4.13 Effects of a Recession on Mothers’ Binge Drinking 99 Figure 4.14 Effects of a Recession on Fathers’ Binge Drinking 100 viii list of illustrations

Figure 4.15 Effects of a Recession on Mothers’ Drug Use 100 Figure 4.16 Effects of a Recession on Fathers’ Drug Use 101 Figure 5.1 Mothers’ Relationship Status 121 Figure 5.2 Marriage to Bio Fathers or New Partners 122 Figure 5.3 Marriage or Cohabitation to Bio Fathers or New Partners 122 Figure 5.4 Mothers’ Reports of Bio Fathers’ Supportiveness 123 Figure 5.5 Fathers’ Reports of Bio Mothers’ Supportiveness 124 Figure 5.6 Mothers’ Reports of New Partners’ Supportiveness 125 Figure 5.7 Mothers’ Reports of Relationship with Bio Father 125 Figure 5.8 Fathers’ Reports of Relationship with Bio Mother 126 Figure 5.9 Mothers’ Marriage and Marriage or Cohabitation 127 Figure 5.10 Mothers’ Marriage (Bio Father or New Partner) 128 Figure 5.11 Mothers’ Marriage or Cohabitation (Bio Father or New Partner) 128 Figure 5.12 Mothers’ Reports of Bio Fathers’ Supportiveness 130 Figure 5.13 Fathers’ Reports of Mothers’ Supportiveness 130 Figure 5.14 Mothers’ Reports of New Partners’ Supportiveness 131 Figure 5.15 Mothers’ Reports of Quality of Relationship with Bio Father 132 Figure 5.16 Fathers’ Reports of Quality of Relationship with Bio Mother 132 Figure 6.1 Nonresidence Status 153 Figure 6.2 Father Engagement 154 Figure 6.3 Child Support and Visitation 155 Figure 6.4 Formal Child Support per Year 156 Figure 6.5 Informal Child Support per Year 157 Figure 6.6 In-Kind Child Support per Year 157 Figure 6.7 Visitation Days per Month 158 Figure 6.8 Share of Nonresident Fathers Visiting Their Children 158 Figure 7.1 High-Frequency Maternal Spanking by Education 176 Figure 7.2 High-Frequency Maternal Physical Aggression by Education 177 Figure 7.3 High-Frequency Maternal Psychological Aggression by Education 177 Figure 7.4 High-Frequency Maternal Warmth by Education 178 Figure 7.5 High-Frequency Maternal Parenting Activities by Education 179 Figure 7.6 High-Frequency Paternal Spanking by Education 179 Figure 7.7 High-Frequency Paternal Physical Aggression by Education 180 Figure 7.8 High-Frequency Paternal Psychological Aggression by Education 181 Figure 7.9 High-Frequency Maternal Spanking by Unemployment Rate 182 list of illustrations ix

Figure 7.10 High-Frequency Maternal Physical Aggression by Unemployment Rate 182 Figure 7.11 High-Frequency Maternal Psychological Aggression by Unemployment Rate 183 Figure 7.12 High-Frequency Maternal Warmth by Unemployment Rate 183 Figure 7.13 High-Frequency Maternal Parenting Activities by Unemployment Rate 184 Figure 7.14 High-Frequency Paternal Spanking by Unemployment Rate 185 Figure 7.15 High-Frequency Paternal Physical Aggression by Unemployment Rate 186 Figure 7.16 High-Frequency Paternal Psychological Aggression by Unemployment Rate 186 Figure 8.1 Child Internalizing Behavior Problems 209 Figure 8.2 Child Externalizing Behavior Problems 209 Figure 8.3 Child PPVT Scores 210 Figure 8.4 Child Overweight-Obese 211 Figure 8.5 Child Internalizing Behaviors 212 Figure 8.6 Child Externalizing Behaviors 212 Figure 8.7 Child PPVT Scores, Unemployment Rates 213 Figure 8.8 Child Overweight-Obese, Unemployment Rates 213

Table 1.1 Fragile Families Sample Composition, Mothers’ Education 8 Table 2.A1 Full Regression Results, Material Hardship 48 Table 2.A2 Coefficients and Standard Errors, Rate of Change, Economic Outcomes 50 Table 2.A3 Sensitivity of Coefficients, Economic Outcomes 53 Table 2.A4 Coefficients and Standard Errors, Economic Outcomes 54 Table 3.A1 Full Regression Results for SNAP 75 Table 3.A2 Coefficients and Standard Errors, Rate of Change for Transfers 77 Table 3.A3 Sensitivity of Coefficients, Transfers 82 Table 3.A4 Coefficients and Standard Errors, Transfers 84 Table 4.A1 Full Regression Results, Parents’ Physical Health 106 Table 4.A2 Coefficients and Standard Errors, All Outcomes by Maternal Education 108 Table 4.A3 Coefficients and Standard Errors, All Outcomes by Paternal Education 110 Table 4.A4 Sensitivity of Coefficients, Parents’ Health 113 Table 4.A5 Coefficients and Standard Errors, Model 1, All Outcomes 114 Table 5.A1 Full Regression Results, Married to or Cohabiting with Father or New Partner 137 x list of illustrations

Table 5.A2 Coefficients and Standard Errors for Unemployment Rate, Relationship Outcomes 139 Table 5.A3 Sensitivity of Unemployment Rate Coefficients, Relationship Outcomes 143 Table 5.A4 Coefficients and Standard Errors for Unemployment Rate, Relationship Outcomes 145 Table 6.A1 Full Regression Results, Child Support and Visitation 164 Table 6.A2 Coefficients and Standard Errors, Rate of Change, Father Involvement 166 Table 6.A3 Sensitivity of Coefficients, Child Support and Visitation Outcomes 168 Table 6.A4 Coefficients and Standard Errors, Model 1, Child Support and Visitation Outcomes 169 Table 7.A1 Full Regression Results, Maternal Parenting 191 Table 7.A2 Coefficients and Standard Errors, Rate of Change in Unemployment for Maternal Parenting Outcomes 193 Table 7.A3 Sensitivity of Coefficients, Maternal Parenting Outcomes 196 Table 7.A4 Coefficients and Standard Errors, Maternal Parenting Outcomes by Subgroup 197 Table 7.A5 Full Regression Results, Paternal Parenting 198 Table 7.A6 Coefficients and Standard Errors, Rate of Change in Unemployment for Paternal Parenting Outcomes 200 Table 7.A7 Sensitivity of Coefficients, Paternal Parenting Outcomes 202 Table 7.A8 Coefficients and Standard Errors, Paternal Parenting Outcomes by Subgroups 203 Table 8.A1 Full Regression Results, Child Well-Being 219 Table 8.A2 Coefficients and Standard Errors, Rate of Change in Unemployment, Child Well-Being Outcomes 221 Table 8.A3 Sensitivity of Coefficients, Child Well-Being Outcomes 223 Table 8.A4 Coefficients and Standard Errors, Child Well-Being Outcomes by Subgroup 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

he first decade of the twenty-first century in the United States was Ta 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’ strug- gles 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 short- term causal effects of recessions on family functioning and well-being. We introduction 3 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 subsi- dies 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: introduction 5 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

Figure 1.1 Median Household Income Index and Unemployment Rate

104 11 102 10 100 9 98 8 96 7 94 6 92 5 90 4 HII (January 2000 = 100.0) 88 3 Jul 2000 Jul 2001 Jul 2002 Jul 2003 Jul 2004 Jul 2005 Jul 2006 Jul 2007 Jul 2008 Jul 2009 Jul 2010 Jul 2011 Jul 2012 Jul 2013 Jan 2000 Jan 2001 Jan 2002 Jan 2003 Jan 2004 Jan 2005 Jan 2006 Jan 2007 Jan 2008 Jan 2009 Jan 2010 Jan 2011 Jan 2012 Jan 2013 Jan 2014 Apr 2000 Apr 2001 Apr 2002 Apr 2003 Apr 2004 Apr 2005 Apr 2006 Apr 2007 Apr 2008 Apr 2009 Apr 2010 Apr 2011 Apr 2012 Apr 2013 Apr 2014 Oct 2000 Oct 2001 Oct 2002 Oct 2003 Oct 2004 Oct 2005 Oct 2006 Oct 2007 Oct 2008 Oct 2009 Oct 2010 Oct 2011 Oct 2012 Oct 2013 Month and Year Seasonally Adjusted Unemployment Rate

Recessionary periods = 104 Seasonally adjusted household income index (January 2000 = 100.0) Monthly unemployment rate (seasonally adjusted)

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- introduction 7 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 college- educated 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 Some High School School College College + (32%) (26%) (21%) (21%) Married 32 40 57 97 Cohabiting 32 26 23 1 Single 34 34 20 2 Non-Hispanic white 13 22 32 71 Non-Hispanic black 41 46 38 7 Hispanic 45 23 24 7 Other race-ethnicity 1 9 6 15 Poverty 33 21 12 2 Household income ($2010) $37,000 $47,000 $67,000 $176,000 N 1,079 791 767 349 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 introduction 9 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 under- scores 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

Figure 1.2 Local Unemployment Rates During Interviewing Periods

17 Austin 16 Baltimore 15 Boston 14 Chicago 13 Corpus Christi 12 Detroit 11 Indianapolis 10 Jacksonville 9 8 Milwaukee 7 Nashville 6 New York 5 Newark 4 Norfolk Unemployment Rate (%) 3 Oakland 2 Philadelphia 1 Pittsburgh 0 Richmond 200920082007200620052004200320022001200019991998 San Antonio San Jose Birth Age 1 Age 3 Age 5 Age 9 Toledo

Source: Authors’ calculations. 10 children of the great recession our cities. During the year nine interviews, unemployment rates were sky- rocketing 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 introduction 11 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 fixed- effects 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 introduction 13 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 two- thirds, 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- introduction 15 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 income- tested 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, employer- provided 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. introduction 17

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 dis- solve 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 marriage- cohabitation 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 college- educated 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 less- educated 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 ter- rible 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 introduction 19 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 well- being. 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 twenty- first 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 immu- nity 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. introduction 21

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 well- being. 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 individual- level 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 well- ordered family stress model, we found that higher unemployment rates were associated with less rather than more harsh parenting, and that rapid introduction 23 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. introduction 25

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 acces- sible, 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 barri- ers 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 Pechac˘ová 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. introduction 27

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. introduction 29

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 Pechac˘ová. 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-on- historical-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

ecessions are primarily an economic phenomenon. If we are to under- R stand 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-being 33 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 twenty- first 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 well- being 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 college- educated 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-being 35 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

Figure 2.1 Maternal Employment

100 90 College + 80 70 60 Some college 50 40 High school 30 20 Less than high school 10 Percent of Mothers Employed 0 1 3 5 9 (1999–2001) (2001–2003) (2003–2006) (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

Figure 2.2 Paternal Employment

100 90 College + 80 70 Some college 60 50 High school 40 30 20 Less than high school 10 Percent of Fathers Employed 0 1 3 5 9 (1999–2001) (2001–2003) (2003–2006) (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-being 37

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

Dollars of Household Income 0 1 3 5 9 (1999–2001)(2001–2003)(2003–2006) (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

Figure 2.4 Big Gains and Losses

100 22 90 27 31 80 35 18 70 + 40 percent 23 60 17 14 + 10–40 percent 50 22 16 13 40 23 No change 30 23 18 − 10–40 percent

Each Income Change 22 20 15 20 − 40 percent 10 13 17 13 Percent of Households Experiencing 0 1 (1999–2001) to 3 (2001–2003) to 5 (2003–2006) to 5 (2003–2006) to 3 (2001–2003) 5 (2003–2006) 9 (2007–2010) 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-being 39

Figure 2.5 Poverty Rates

35 College + 30 25 Some college 20 High school 15

10 Less than 5 high school

Percent of Households in Poverty 0 1 3 5 9 (1999–2001) (2001–2003) (2003–2006) (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

50 College +

40 Some college 30 High school 20 Less than Percent of Households Experiencing Insecurity 10 high school 0 1 3 5 9 (1999–2001)(2001–2003)(2003–2006) (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

Figure 2.7 Household Income ($2010) by Race-Ethnicity

140,000 White 120,000 Married 100,000 Hispanic 80,000 Cohabiting 60,000 Black Single 40,000 20,000

Dollars of Household Income 0 1 3 5 9 (1999–2001) (2001–2003) (2003–2006) (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 race- ethnicity 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-being 41 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

Figure 2.8 Employment by Education

100 UR 5 percent UR 10 percent –12% 90 –9% 80 –15% +3% –11% –10% –11% –14% 70 –9% 60 –5% 50 40 30 20 10 0 All* Less than High Some College + All* Less than High Some College +* high school* college* high school* college* school school

Predicting Percent of Employment Mother’s Employment 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-being 43

Figure 2.9 Income by Education

160,000 –5% 140,000 120,000 100,000 –15% UR 5 percent 80,000 –13% –20% UR 10 percent 60,000 –14% 40,000 20,000 0 All* Less than High Some College + high school* school* college* Predicted Dollars of Household Income

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

Figure 2.10 Income by Race-Ethnicity and Relationship Status

100,000 –9% 90,000 –17% 80,000 70,000 60,000 –19% –20% UR 5 percent 50,000 –20% –20% 40,000 UR 10 percent 30,000 20,000 Household Income

Predicted Dollars of 10,000 0 Black* Hispanic* White Married Cohab* Single* Race or Ethnicity 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

Figure 2.11 Poverty Rate by Education

40 +42% 35 30 +56% +63% 25 20 UR 5 percent +75% 15 UR 10 percent 10 –33% Predicted Percent of 5 Households in Poverty 0 All* Less than High Some College + high school* school* 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

Figure 2.12 Hardship by Education

70 +24% +56% +26% +16% 60 50 40 UR 5 percent 30 +47% UR 10 percent 20 10

Experiencing Insecurity 0 All* Less than High Some College +

Predicted Percent of Households high school school* 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-being 45 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-being 47

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 distin- guished 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 construct- ing 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 Without Individual Fixed Effects Fixed Effects Unemployment rate 1.17** (5.02) 1.13** (4.10) Education Less than high school 2.85** (6.94) High school 2.33** (5.92) Some college 2.72** (6.97) Relationship status Married 0.58** (-6.34) Cohabiting 1.11* (1.98) Mother’s age 1.00 (-0.34) Race-ethnicity Black 1.22* (2.32) Hispanic 0.96 (-0.43) Other 1.20 (1.43) Immigrant 0.77† (-1.87) Number of children in household 1.08** (3.02) Lived with both parents at age fifteen 0.77** (-4.82) Interview year 2000 0.59** (-2.69) 0.70** (-4.49) 2001 0.83 (-1.28) 0.87** (-3.16) 2002 0.66* (-2.38) 0.72** (-3.87) 2003 0.83 (-1.07) 0.84* (-1.99) 2004 0.59** (-2.95) 0.70** (-3.41) 2005 0.73† (-1.90) 0.77** (-3.25) 2006 1.71 (1.00) 0.89 (-0.35) 2007 0.93 (-0.32) 0.92 (-1.34) 2008 0.98 (-0.14) 1.02 (0.16) 2009 0.77 (-1.26) 0.82 (-1.26) 2010 0.63 (-1.37) 0.77 (-0.97) economic well-being 49

Table 2.A1 Continued With Individual Without Individual Fixed Effects Fixed Effects City Austin 1.47** (10.58) Baltimore 1.07 (0.91) Detroit 1.14* (2.37) Newark 0.89** (-3.12) Philadelphia 1.05 (0.73) Richmond 1.27** (2.62) Corpus Christi 0.89† (-1.80) Indianapolis 1.38** (4.48) Milwaukee 1.25** (4.26) New York 0.83** (-4.88) San Jose 0.74** (-4.70) Boston 1.34** (6.42) Nashville 1.06 (0.90) Chicago 0.82** (-3.80) Jacksonville 0.74** (-4.72) Toledo 1.02 (0.32) San Antonio 1.33** (4.32) Pittsburgh 1.16* (2.13) Norfolk 1.00 (0.03) Constant 0.23** (-7.94) Observations 8,392 15,860 Number of individuals 2,280 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 Table 2.A2 Coefficients and Standard Errors, Rate of Change, Economic Outcomes With Individual Fixed Effects Without Individual Fixed Effects Less than Less than High High Some High High Some All School School College College + All School School College College + Mother’s employment odds ratios (z-stat) Unemployment rate 0.89** 0.93 0.87* 0.80** 1.04 0.93** 0.96 0.93* 0.84** 1.04 (model 1) (-3.82) (-1.48) (-2.45) (-3.42) (0.31) (-2.64) (-1.05) (-1.98) (-5.00) (0.43) Unemployment rate 0.89** 0.94 0.87* 0.80** 1.04 0.93** 0.96 0.93† 0.84** 1.04 (model 2) (-3.73) (-1.39) (-2.32) (-3.49) (0.29) (-2.66) (-0.99) (-1.92) (-5.23) (0.43) Increasing 1.00 1.00 1.00 1.00 1.01 1.00 1.00 1.00 1.00 1.00 unemployment rate (0.52) (0.40) (0.86) (-0.92) (1.16) (0.51) (0.00) (0.31) (-0.25) (1.10) Decreasing 1.00 1.00 0.98 1.00 0.98 1.00 1.01 0.99 1.00 0.99 unemployment rate (-0.50) (0.39) (-1.25) (0.16) (-0.94) (0.11) (1.51) (-1.16) (0.70) (-0.69) Observations 8,446 3,587 2,281 1,856 722 15,851 6,126 4,061 3,927 1,733 Number of 2,301 991 617 502 191 individuals Father’s employment odds ratios (z-stat) Unemployment rate 0.85** 0.84** 0.88 0.86 0.62* 0.89** 0.94 0.89† 0.84** 0.68** (model 1) (-3.77) (-2.63) (-1.45) (-1.56) (-2.21) (-3.50) (-1.56) (-1.83) (-3.46) (-3.44) Unemployment rate 0.84** 0.82** 0.90 0.84† 0.65† 0.88** 0.93† 0.90 0.81** 0.68** (model 2) (-3.81) (-2.83) (-1.26) (-1.77) (-1.90) (-3.57) (-1.67) (-1.59) (-3.21) (-3.06) Increasing 1.00 1.00 1.01 0.99† 1.02 1.00* 1.00 1.00 0.99** 1.00 unemployment rate (-0.52) (-1.22) (1.39) (-1.67) (1.43) (-2.39) (-1.48) (1.49) (-3.94) (0.31) Decreasing 1.00 0.99 0.99 1.00 1.00 1.00 1.00 1.00 1.00 1.01 unemployment rate (-0.33) (-0.40) (-0.55) (-0.04) (-0.03) (0.16) (-0.25) (0.25) (0.20) (0.17) Observations 3,924 1,724 1,091 891 218 11,588 4,236 2,864 2,934 1,499 Number of 1,181 522 333 264 62 individuals Log of household income Unemployment rate -0.04** -0.05** -0.05** -0.03* 0.01 -0.04** -0.05** -0.05** -0.03* 0.01 (model 1) (0.01) (0.01) (0.01) (0.01) (0.06) (0.01) (0.02) (0.01) (0.01) (0.06) Unemployment rate -0.03** -0.04** -0.05** -0.02† 0.03 -0.03** -0.05** -0.05** -0.03* 0.02 (model 2) (0.01) (0.01) (0.01) (0.01) (0.07) (0.01) (0.02) (0.01) (0.01) (0.07) Increasing rate of 0.00** 0.00* 0.00† 0.00* 0.00 0.00 0.00 -0.00 0.00* 0.00 unemployment (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) Decreasing rate of 0.00 0.00 -0.00 0.00 0.00 -0.00 0.00 -0.00 0.00 -0.00 unemployment (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.01) (0.00) (0.00) Observations 15,688 6,062 4,027 3,879 1,720 15,688 6,062 4,027 3,879 1,720 Number of 4,600 1,821 1,162 1,122 495 individuals Poverty (odds ratios) Unemployment rate 1.16** 1.16** 1.20* 1.17 0.93 1.11** 1.09* 1.12* 1.20** 0.81 (model 1) (4.01) (2.91) (2.51) (1.60) (-0.18) (3.63) (2.36) (2.20) (2.99) (-0.67) Unemployment rate 1.14** 1.13* 1.20* 1.16 0.92 1.10** 1.08* 1.13* 1.19** 0.82 (model 2) (3.52) (2.39) (2.43) (1.43) (-0.19) (3.44) (2.07) (2.18) (3.03) (-0.69) Increasing 0.99** 0.99** 1.00 0.99 0.97 1.00** 1.00** 1.00 0.99 0.99 unemployment rate (-3.32) (-3.30) (-0.46) (-1.05) (-1.11) (-2.68) (-3.07) (0.17) (-1.48) (-0.39) Decreasing 0.99 0.99 1.01 0.98 0.81 1.00 0.99 1.01 0.99 1.09 unemployment rate (-0.83) (-1.01) (0.89) (-1.07) (-1.15) (-0.34) (-0.85) (1.01) (-0.76) (0.93) Observations 5,833 3,277 1,571 920 65 15,656 6,045 4,004 3,876 1,018 Number of 1,618 916 433 251 18 individuals (Table continues on p. 52.) Table 2.A2 Continued With Individual Fixed Effects Without Individual Fixed Effects Less than Less than High High Some High High Some All School School College College + All School School College College + Material hardship Unemployment rate 1.17** 1.09† 1.18** 1.30** 1.16 1.13** 1.08* 1.11** 1.23** 1.08 (model 1) (5.02) (1.72) (2.68) (4.13) (1.01) (4.10) (2.11) (3.26) (6.09) (0.77) Unemployment rate 1.19** 1.10* 1.19** 1.33** 1.16 1.13** 1.09* 1.11** 1.24** 1.07 (model 2) (5.36) (2.00) (2.77) (4.41) (0.99) (4.20) (2.25) (3.17) (6.76) (0.57) Increasing 1.00** 1.00† 1.00 1.01† 1.01 1.00 1.00 1.00 1.00 1.00 unemployment (2.75) (1.90) (0.84) (1.74) (1.26) (1.23) (1.08) (0.77) (1.05) (0.08) Decreasing 1.00 1.01 1.01 1.01 0.95† 1.00 1.01 0.99 1.01 0.96* unemployment (0.54) (0.54) (0.48) (1.11) (-1.75) (0.05) (1.18) (-1.00) (0.98) (-1.99) Observations 8,392 3,523 2,341 2,025 503 15,860 6,131 4,064 3,925 1,740 Number of 2,280 971 631 545 133 individuals 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 economic well-being 53

Table 2.A3 Sensitivity of Coefficients, Economic Outcomes With Individual Without Individual Fixed Effects Fixed Effects Mother’s employment odds ratios (z-stat) Unemployment rate (model 3) — — — — Individual unemployment — — — — Unemployment rate (model 4) 0.89** (-2.84) 0.92* (-2.19) Unemployment rate ∗ year nine 1.00 (-0.00) 1.02 (0.50) Father’s employment odds ratios (z-stat) Unemployment rate (model 3) — — — — Individual unemployment — — — — Unemployment rate (model 4) 0.80** (-3.63) 0.87** (-3.69) Unemployment rate ∗ year nine 1.09 (1.27) 1.04 (1.19) Log of household income Unemployment rate (model 3) -0.03** (0.01) -0.03** (0.01) Individual unemployment 0.24** (0.02) 0.43** (0.02) Unemployment rate (model 4) -0.04** (0.01) -0.05** (0.01) Unemployment rate ∗ year nine 0.01 (0.01) 0.02 (0.01) Poverty odds ratio (z-stat) Unemployment rate (model 3) 1.14** (3.39) 1.08** (3.02) Individual unemployment 0.40** (-13.16) 0.25** (-19.18) Unemployment rate (model 4) 1.18** (3.28) 1.12** (4.16) Unemployment rate ∗ year nine 0.97 (-0.45) 0.99 (-0.20) Material hardship odds ratio (z-stat) Unemployment rate (model 3) 1.17** (4.95) 1.12** (4.03) Individual unemployment 0.92 (-1.42) 0.81** (-4.71) Unemployment rate (model 4) 1.15** (3.40) 1.09† (1.94) Unemployment rate ∗ year nine 1.03 (0.57) 1.05 (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 Table 2.A4 Coefficients and Standard Errors, Economic Outcomes Black Hispanic White Married Cohabiting Single Mother’s employment odds ratio (z-stat) Unemployment rate 0.86** 0.91† 0.99 0.96 0.96 0.79** (-3.07) (-1.65) (-0.18) (-0.55) (-0.84) (-4.69) Father’s employment odds ratio (z-stat) Unemployment rate 0.88† 0.84* 0.76* 0.73** 0.83** 0.91 (-1.93) (-2.15) (-2.39) (-2.64) (-2.80) (-1.36) Log of household income Unemployment rate -0.05** -0.04** -0.01 0.00 -0.05** -0.05** (0.01) (0.01) (0.04) (0.02) (0.01) (0.01) Poverty odds ratio (z-stat) Unemployment rate 1.24** 1.03 1.26* 1.09 1.10 1.25** (3.71) (0.54) (2.11) (0.68) (1.57) (4.19) Material hardship odds ratio (z-stat) Unemployment rate 1.20** 1.20** 1.08 1.19* 1.20** 1.13* (3.67) (3.31) (1.01) (2.33) (3.66) (2.55) 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 economic well-being 55

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-being 57

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

he previous chapter documents great differences in employment, Tincome, 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 college- educated, 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, low- income, 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 transfers 59 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 gener- ous 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 moderate- income 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 low- income, 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 imple- mented 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 transfers 61 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 imple- mented 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

Medicaid 40 EITC 30 SNAP TANF 20 UI/other Public Assistance Housing 10 SSI Percent of Households Receiving 0 1 3 5 9 (1999–2001) (2001–2003) (2003–2006) (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 transfers 63

Figure 3.2 Public Assistance Receipt by Education

70 Less than 60 high school High school 50

40 Some college

Public Assistance 30 College +

Percent of Households Receiving 10

0 Medicaid EITC SNAP Housing TANF SSI UI/other

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 under- funded, 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 9,000 8,000 7,000 6,000 5,000 4,000 3,000

Dollars of Public Assistance 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 transfers 65

Figure 3.4 Private Financial Transfers ($2010)

45 40 College + 35 30 Some college 25 High school 20 15 Less than 10 high school

Private Financial Transfers 5

Percent of Households Receiving 0 1 3 5 9 (1999–2001) (2001–2003) (2003–2006) (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 3.5 Doubling Up

45 40 35 College + 30 Some college 25 20 High school 15 Less than 10 high school 5

Percent of Households Doubling Up 0 1 3 5 9 (1999–2001) (2001–2003) (2003–2006) (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

Figure 3.6 Average Dollar Value of Private Assistance

10,000 9,000 All 8,000 7,000 Less than high school 6,000 5,000 High school 4,000 Some college 3,000 College + 2,000

Dollars of Private Assistance 1,000 0 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 transfers 67 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

Figure 3.7 Private Financial Transfers and Doubling Up

50 +52% UR 5 percent UR 10 percent 45 +29% +15% +8% 40 +74% 35 30 +9% 25 +2% +10% +29% 20 Assistance 15 –20% 10 5

Predicted Percent of Private 0 All* Less than High Some College +* All Less than High Some College + high school college high school college school* school PFTs 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 transfers 69

Figure 3.8 Public Transfer Receipt Rates

25 22* All 20 17* 16* 14* 10* Less than 15 10 10* 10* 9* high school 10 7* 6 5 4* 5 4* High school 5 1 0 Some college –1 –5 –3 –2 –3 College + –5 –5 –10 –6 –15 Predicted Percentage Point

Change in Public Assistance –15* –20 Medicaid EITC SNAP TANF 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 college- educated 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 transfers 71

Figure 3.9 Effects of Transfers on Household Income

160,000 +3% 140,000

120,000 Pretransfer 100,000 income 80,000 +3% +17% Post-transfer 60,000 income +9% +18% 40,000

20,000

Predicted Dollars of Household Income 0 All Less than High Some College + high school school 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.

Figure 3.10 Mitigating Effects of Transfers on Poverty

60 –26% 50

40 –26% Pretransfer –21% poverty 30 Post-transfer –26% poverty 20

10 –33%

0 All Less than High Some College + Predicted Percent of Households in Poverty high school school 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 less- educated 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, recession- induced 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 transfers 73

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 transfers 75 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 Without Individual Fixed Effects Fixed Effects Unemployment rate 1.18** (4.64) 1.10** (2.85) Education Less than high school 14.08** (19.93) High school 8.03** (14.49) Some college 4.85** (10.16) Relationship status Married 0.28** (-13.92) Cohabiting 0.76** (-5.71) Mother’s age 0.99† (-1.66) Race-ethnicity Black 1.93** (6.32) Hispanic 1.38* (2.21) Other 1.72* (2.21) Immigrant 0.50** (-4.79) Number of children in household 1.16** (6.04) Lived with both parents at age fifteen 0.73** (-8.13) Interview year 2000 0.54** (-2.85) 0.65† (-1.73) 2001 0.41** (-5.83) 0.57** (-2.63) 2002 0.38** (-4.86) 0.56* (-2.40) 2003 0.43** (-4.48) 0.60* (-2.06) 2004 0.49** (-3.61) 0.66† (-1.68) 2005 0.48** (-3.96) 0.66† (-1.84) 2006 0.38 (-1.54) 0.39* (-2.34) 2007 1.05 (0.21) 0.93 (-0.45) 2008 0.66* (-2.18) 0.81 (-0.80) 2009 0.50** (-3.04) 0.71 (-1.32) 2010 0.55 (-1.49) 0.88 (-0.42) (Table continues on p. 76.) 76 children of the great recession

Table 3.A1 Continued With Individual Without Individual Fixed Effects Fixed Effects City Austin 1.14** (3.52) Baltimore 0.75** (-3.44) Detroit 0.94 (-0.80) Newark 0.75** (-3.81) Philadelphia 0.94 (-0.77) Richmond 1.22† (1.88) Corpus Christi 1.82** (5.98) Indianapolis 1.29** (3.12) Milwaukee 1.88** (9.07) New York 1.22* (2.13) San Jose 0.45** (-6.57) Boston 1.05 (0.65) Nashville 1.35** (3.77) Chicago 1.23* (2.28) Jacksonville 0.86† (-1.82) Toledo 1.57** (4.72) San Antonio 0.94 (-0.62) Pittsburgh 1.19* (2.07) Norfolk 1.22* (2.22) Constant 0.08** (-9.39) Observations 6,430 15,688 Number of individuals 1,771 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 Table 3.A2 Coefficients and Standard Errors, Rate of Change for Transfers With Individual Fixed Effects Without Individual Fixed Effects Less than Less than High High Some High High Some All School School College College + All School School College College + Medicaid odds ratios (z-stat) Unemployment rate 1.19** 1.27** 1.23** 1.07 0.91 1.11* 1.17* 1.07 1.08* 0.97 (model 1) (5.27) (4.64) (3.29) (1.08) (-0.63) (2.55) (2.52) (1.17) (2.23) (-0.30) Unemployment rate 1.19** 1.28** 1.22** 1.06 0.84 1.11* 1.18** 1.07 1.06† 0.94 (model 2) (5.21) (4.77) (3.19) (0.84) (-1.11) (2.45) (2.72) (1.14) (1.66) (-0.53) Increasing 1.00 1.00 1.00 1.00 0.98† 1.00 1.00 1.00 1.00† 0.99† unemployment rate (-0.45) (1.39) (-0.45) (-1.06) (-1.88) (-0.38) (1.36) (-0.09) (-1.67) (-1.74) Decreasing 1.00 1.03* 0.99 0.99 0.95 1.00 1.02 0.99 0.99 0.97 unemployment rate (0.64) (2.37) (-0.42) (-1.01) (-1.53) (0.11) (1.42) (-1.03) (-0.99) (-1.11) Observations 7,529 3,092 2,108 1,943 386 15,820 6,121 4,048 3,919 1,713 Number of 2,060 855 574 526 105 4,604 1,822 1,162 1,124 496 individuals EITC odds ratios (z-stat) Unemployment rate 0.95 0.94 0.84** 1.02 1.14 0.97 0.98 0.89† 0.99 1.15† (model 1) (-1.63) (-1.30) (-2.93) (0.30) (1.26) (-0.80) (-0.34) (-1.82) (-0.14) (1.81) Unemployment rate 0.96 0.95 0.85** 1.02 1.12 0.97 0.98 0.90† 0.99 1.12 (model 2) (-1.38) (-1.07) (-2.63) (0.39) (1.00) (-0.75) (-0.32) (-1.66) (-0.16) (1.31) Increasing 1.00 1.00 1.00 1.00 0.99* 1.00 1.00 1.00 1.00 0.99** unemployment rate (0.52) (1.26) (1.20) (0.02) (-2.25) (-0.54) (-0.02) (1.06) (-1.13) (-2.60) Decreasing 1.02* 1.00 1.03* 1.02 1.04 1.01** 1.01 1.02* 1.01 1.02 unemployment rate (2.37) (0.42) (2.10) (1.25) (1.63) (2.76) (1.40) (2.26) (0.98) (1.04) Observations 9,053 3,535 2,222 2,365 931 15,884 6,139 4,073 3,931 1,741 Number of 2,475 979 604 642 250 4,605 1,823 1,162 1,124 496 individuals (Table continues on p. 78.) Table 3.A2 Continued With Individual Fixed Effects Without Individual Fixed Effects Less than Less than High High Some High High Some All School School College College + All School School College College + SNAP odds ratios (z-stat) Unemployment rate 1.18** 1.16** 1.25** 1.18* 0.89 1.10** 1.09* 1.09 1.16* 1.05 (model 1) (4.64) (2.67) (3.24) (2.17) (-0.44) (2.85) (2.07) (1.52) (2.20) (0.22) Unemployment rate 1.19** 1.17** 1.23** 1.18* 0.79 1.10** 1.09* 1.09 1.15* 0.99 (model 2) (4.62) (2.86) (3.00) (2.18) (-0.78) (2.73) (2.03) (1.58) (2.04) (-0.03) Increasing 1.00 1.00 0.99† 1.00 0.95* 1.00 1.00 1.00 1.00 0.97* unemployment rate (-0.61) (1.27) (-1.85) (-0.69) (-2.09) (-0.71) (1.02) (-0.61) (-1.34) (-2.28) Decreasing 1.02* 1.01 1.02 1.02 0.99 1.00 1.00 1.01 1.01 1.02 unemployment rate (2.06) (1.03) (1.53) (1.34) (-0.10) (0.52) (-0.06) (0.75) (0.39) (0.44) Observations 6,430 2,920 1,906 1,459 145 15,688 6,022 4,026 3,900 1,700 Number of 1,771 819 519 394 39 4,595 1,817 1,159 1,123 496 individuals TANF odds ratios Unemployment rate 1.16** 1.23** 1.15 1.00 — 1.07 1.11† 1.00 1.07 0.79 (model 1) (3.34) (3.47) (1.63) (-0.01) (1.36) (1.86) (0.01) (1.09) (-0.32) Unemployment rate 1.15** 1.25** 1.12 0.98 — 1.06 1.11† 1.00 1.05 0.79 (model 2) (3.22) (3.61) (1.35) (-0.17) (1.30) (1.86) (-0.04) (0.77) (-0.38) Increasing 1.00† 1.00 0.99† 0.99† — 1.00 1.00 1.00 0.99** 0.98 unemployment rate (-1.65) (0.51) (-1.83) (-1.91) (-1.32) (-0.04) (-0.55) (-2.58) (-1.06) Decreasing 1.02* 1.03† 1.02 1.01 — 1.00 1.00 1.01 1.01 1.20 unemployment rate (2.23) (1.95) (0.99) (0.62) (0.56) (0.01) (0.49) (0.75) (1.60) Observations 5,152 2,775 1,373 960 15,750 6,062 4,006 3,912 740 Number of 1,418 777 370 259 4,603 1,821 1,162 1,124 496 individuals Public housing or Section 8 odds ratio (z-stat) Unemployment rate 1.00 1.01 0.98 0.98 — 0.97 0.97 0.95 0.94 1.32 (model 1) (-0.04) (0.16) (-0.28) (-0.16) (-0.77) (-0.43) (-0.76) (-1.23) (0.89) Unemployment rate 0.99 1.01 0.97 0.95 — 0.96 0.98 0.95 0.91† 1.33 (model 2) (-0.16) (0.25) (-0.42) (-0.53) (-0.86) (-0.41) (-0.92) (-1.78) (1.19) Increasing 1.00 1.00 0.99 0.99* — 1.00 1.00 1.00 0.99** 0.97 unemployment rate (-1.09) (0.41) (-1.22) (-2.10) (-1.41) (0.18) (-0.74) (-4.08) (-0.94) Decreasing 1.01 1.01 1.04† 0.95* — 1.01 1.00 1.03** 0.98 1.12 unemployment rate (0.67) (0.61) (1.90) (-2.19) (1.29) (0.21) (3.44) (-1.35) (1.60) Observations 4,427 2,296 1,311 751 15,739 6,059 4,049 3,896 1,190 Number of 1,218 645 349 204 4,604 1,823 1,162 1,123 496 individuals Supplemental Security Income odds ratio (z-stat) Unemployment rate 1.03 0.96 1.24 — — 0.94† 0.95 0.90 1.07 0.44 (model 1) (0.28) (-0.32) (1.11) (-1.70) (-1.04) (-0.91) (0.69) (-1.26) Unemployment rate 1.03 0.98 1.24 — — 0.95 0.97 0.91 1.07 0.38 (model 2) (0.33) (-0.17) (1.12) (-1.43) (-0.66) (-0.89) (0.64) (-1.52) Increasing 1.00 1.01 1.01 — — 1.00† 1.01* 1.00 1.00 1.04* unemployment rate (0.42) (1.15) (0.64) (1.79) (2.35) (0.35) (-0.29) (2.16) Decreasing 1.01 1.01 1.03 — — 1.01 1.01 1.01 1.02 0.94 unemployment rate (0.45) (0.40) (0.77) (1.19) (0.90) (0.61) (1.44) (-1.38) Observations 1,403 701 389 15,828 6,111 4,042 3,811 1,037 Number of 384 193 107 4,602 1,821 1,162 1,124 495 individuals (Table continues on p. 80.) Table 3.A2 Continued With Individual Fixed Effects Without Individual Fixed Effects Less than Less than High High Some High High Some All School School College College + All School School College College + Unemployment insurance or other odds ratio (z-stat) Unemployment rate 1.13* 0.98 1.20 1.36** 0.75 1.13† 0.96 1.23 1.35** 0.74† (model 1) (2.26) (-0.22) (1.62) (2.99) (-1.26) (1.93) (-0.68) (1.59) (2.96) (-1.96) Unemployment rate 1.15* 1.02 1.19 1.42** 0.68 1.14* 1.00 1.21 1.38** 0.71* (model 2) (2.45) (0.17) (1.52) (3.30) (-1.53) (2.26) (-0.01) (1.35) (3.29) (-2.12) Increasing 1.00 1.01 0.99 1.01† 0.98 1.00* 1.01** 0.99 1.01* 0.98 unemployment rate (1.56) (1.62) (-1.47) (1.90) (-1.18) (1.98) (3.41) (-1.50) (2.14) (-1.27) Decreasing 1.00 0.99 0.98 1.02 0.99 1.00 1.01 0.97 1.00 0.97 unemployment rate (-0.19) (-0.33) (-0.99) (0.99) (-0.18) (-0.69) (0.41) (-1.39) (0.36) (-0.62) Observations 2,767 973 749 831 214 15,802 6,008 3,990 3,905 1,493 Number of 743 265 199 221 58 4,603 1,823 1,161 1,123 496 individuals Private financial transfers odds ratio (z-stat) Unemployment rate 1.11** 1.18** 1.04 1.03 1.35* 1.08** 1.12* 1.01 1.07 1.20* (model 1) (3.37) (3.46) (0.67) (0.48) (2.41) (4.05) (2.41) (0.16) (1.04) (2.09) Unemployment rate 1.11** 1.20** 1.04 1.01 1.32* 1.08** 1.13* 1.00 1.06 1.20* (model 2) (3.20) (3.68) (0.58) (0.23) (2.15) (3.75) (2.38) (-0.01) (0.93) (2.00) Increasing 1.00 1.00 1.00 1.00 0.99 1.00† 1.00 1.00 1.00 1.00 unemployment rate (-0.93) (1.21) (-0.47) (-1.56) (-1.36) (-1.87) (0.42) (-1.46) (-1.28) (-0.07) Decreasing 1.00 1.02† 0.99 0.99 1.00 1.00 1.01† 0.98† 1.00 1.01 unemployment rate (0.19) (1.89) (-1.02) (-0.78) (-0.01) (0.08) (1.69) (-1.90) (-0.47) (0.60) Observations 8,220 3,302 2,157 2,098 663 15,691 6,064 4,027 3,880 1,720 Number of 2,243 909 587 570 177 4,600 1,821 1,162 1,122 495 individuals Doubling up odds ratio (z-stat) Unemployment rate 1.05 1.03 1.04 1.09 0.96 1.02 1.00 1.00 1.09* 1.06 (model 1) (1.31) (0.51) (0.54) (1.18) (-0.30) (0.64) (-0.03) (-0.06) (2.20) (0.60) Unemployment rate 1.04 1.01 1.05 1.09 0.95 1.02 0.99 1.00 1.10* 1.07 (model 2) (1.23) (0.23) (0.70) (1.16) (-0.33) (0.59) (-0.20) (-0.07) (2.34) (0.62) Increasing 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00† 1.00 unemployment rate (-0.24) (-1.44) (1.31) (0.21) (-0.51) (-0.33) (-1.40) (0.14) (1.86) (0.35) Decreasing 0.99 1.00 0.99 0.99 1.01 1.00 1.00 0.99 1.00 1.01 unemployment rate (-0.67) (-0.37) (-0.75) (-0.40) (0.38) (-0.78) (-0.69) (-0.92) (0.53) (0.19) Observations 6,891 3,205 1,796 1,510 380 15,884 6,139 4,073 3,931 1,722 Number of 1,890 894 484 409 103 4,605 1,823 1,162 1,124 495 individuals 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 82 children of the great recession

Table 3.A3 Sensitivity of Coefficients, Transfers With Individual Fixed Without Individual Effects Fixed Effects Medicaid odds ratios (z-stat) Unemployment rate (model 1) 1.19** (5.27) 1.11* (2.55) Unemployment rate (model 3) 1.08† (1.93) 1.05 (1.05) Mother’s unemployment 1.85** (6.77) 2.57** (19.74) Bio-social fathers not employed 1.15 (1.60) 1.71** (11.22) Unemployment rate (model 4) 1.14** (3.08) 1.10† (1.68) Unemployment rate * year nine 1.07 (1.27) 1.03 (0.63) EITC odds ratios (z-stat) Unemployment rate (model 1) 0.95 (-1.63) 0.97 (-0.80) Unemployment rate (model 3) 1.04 (0.85) 1.02 (0.47) Mother’s unemployment 0.26** (-14.80) 0.34** (-12.96) Bio-social fathers not employed 1.01 (0.15) 1.05 (1.25) Unemployment rate (model 4) 0.97 (-0.70) 0.97 (-0.63) Unemployment rate * year nine 0.96 (-0.79) 1.00 (-0.03) SNAP odds ratio (z-stat) Unemployment rate (model 1) 1.18** (4.64) 1.10** (2.85) Unemployment rate (model 3) 1.13** (2.59) 1.06† (1.75) Mother’s unemployment 2.25** (8.82) 3.52** (20.66) Bio-social father’s not employed 1.11 (1.19) 1.85** (9.20) Unemployment rate (model 4) 1.17** (3.30) 1.09† (1.88) Unemployment rate * year nine 1.01 (0.26) 1.02 (0.46) TANF odds ratio (z-stat) Unemployment rate (model 1) 1.16** (3.34) 1.07 (1.36) Unemployment rate (model 3) 1.08 (1.25) 1.03 (0.51) Mother’s unemployment 2.96** (11.09) 4.14** (26.86) Bio-social fathers not employed 1.41** (3.40) 1.97** (8.62) Unemployment rate (model 4) 1.11† (1.83) 1.01 (0.20) Unemployment rate * year nine 1.07 (1.00) 1.10 (1.51) Public housing or Section 8 odds ratio (z-stat) Unemployment rate (model 1) 1.00 (-0.04) 0.97 (-0.77) Unemployment rate (model 3) 1.00 (0.01) 0.95 (-0.81) Mother’s unemployment 1.47** (3.76) 1.89** (11.39) Bio-social fathers not employed 1.22† (1.95) 1.49** (6.70) Unemployment rate (model 4) 0.99 (-0.16) 0.95 (-0.79) Unemployment rate * year nine 1.01 (0.20) 1.03 (0.63) Supplemental Security Income odds ratio (z-stat) Unemployment rate (model 1) 1.03 (0.28) 0.94† (-1.70) Unemployment rate (model 3) 0.99 (-0.07) 0.86** (-3.34) Mother’s unemployment 1.29 (0.90) 2.61** (9.92) Bio-social fathers not employed 2.94** (3.34) 1.74** (4.87) Unemployment rate (model 4) 0.97 (-0.26) 0.93 (-1.04) Unemployment rate * year nine 1.08 (0.65) 1.02 (0.21) public and private transfers 83

Table 3.A3 Continued With Individual Fixed Without Individual Effects Fixed Effects Unemployment insurance or other odds ratio (z-stat) Unemployment rate (model 1) 1.13* (2.26) 1.13† (1.93) Unemployment rate (model 3) 1.18* (2.35) 1.16** (2.78) Mother’s unemployment 3.49** (9.60) 2.68** (6.58) Bio-social fathers not employed 1.25† (1.66) 1.27** (3.39) Unemployment rate (model 4) 1.15† (1.86) 1.15† (1.75) Unemployment rate * year nine 0.98 (-0.28) 0.97 (-0.60) Private financial transfers odds ratio (z-stat) Unemployment rate (model 1) 1.11** (3.37) 1.08** (4.05) Unemployment rate (model 3) 1.14** (3.48) 1.11** (3.90) Mother’s unemployment 1.29** (3.14) 1.38** (7.89) Bio-social fathers not employed 1.11 (1.35) 1.34** (4.60) Unemployment rate (model 4) 1.07 (1.59) 1.05 (1.54) Unemployment rate * year nine 1.07 (1.34) 1.06† (1.66) Doubling up odds ratio (z-stat) Unemployment rate (model 1) 1.05 (1.31) 1.02 (0.64) Unemployment rate (model 3) 1.03 (0.72) 1.00 (0.10) Mother’s unemployment 1.08 (0.83) 1.28** (4.42) Bio-social fathers not employed 1.13 (1.34) 1.20** (3.91) Unemployment rate (model 4) 0.98 (-0.35) 0.99 (-0.29) Unemployment rate * year nine 1.12* (2.07) 1.07 (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 Table 3.A4 Coefficients and Standard Errors, Transfers Black Hispanic White Married Cohabiting Single Medicaid odds ratio (z-stat) Unemployment rate 1.02 1.35** 1.11 1.34** 1.22** 1.13* (0.38) (5.43) (1.37) (3.56) (3.86) (2.32) EITC odds ratio (z-stat) Unemployment rate 0.85** 0.95 1.08 1.04 0.95 0.92† (-3.16) (-1.03) (1.22) (0.59) (-1.03) (-1.68) SNAP odds ratio (z-stat) Unemployment rate 1.22** 1.25** 1.07 1.02 1.32** 1.12* (3.56) (3.31) (0.80) (0.15) (4.79) (2.13) TANF odds ratio (z-stat) Unemployment rate 1.12† 1.31** 1.01 1.11 1.24** 1.12† (1.83) (3.24) (0.07) (0.62) (3.07) (1.80) Public housing or Section 8 odds ratio (z-stat) Unemployment rate 0.95 1.14† 0.85 1.03 1.05 0.96 (-0.90) (1.76) (-1.06) (0.18) (0.73) (-0.62) Supplemental Security Income odds ratio (z-stat) Unemployment rate 1.12 0.91 1.23 1.49† 1.23 0.77† (0.90) (-0.54) (0.62) (1.75) (1.30) (-1.94) Unemployment insurance or other odds ratio (z-stat) Unemployment rate 1.06 1.17 1.31* 1.07 1.09 1.16† (0.70) (1.44) (2.12) (0.52) (1.00) (1.78) Private financial transfers odds ratio (z-stat) Unemployment rate 1.19** 1.01 1.08 1.26** 1.04 1.13** (3.52) (0.22) (1.13) (3.15) (0.76) (2.59) Doubling up odds ratio (z-stat) Unemployment rate 1.07 0.95 1.04 1.08 0.99 1.09 (1.19) (-0.87) (0.52) (0.85) (-0.14) (1.64) 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 public and private transfers 85

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-Creating- a-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 transfers 87

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

conomic recessions can represent huge economic and psychological Eshocks 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’ health 89 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’ health 91

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

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 9 (1999–2001) (2001–2003) (2003–2006) (2007–2010) Mothers’ Health Status is Fair or Poor (%) 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

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 9 (1999–2001) (2001–2003) (2003–2006) (2007–2010) Fathers’ Health Status is Fair or Poor (%) 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

0.18 0.16 College + 0.14 0.12 Some college 0.10 0.08 High school 0.06

Limits Work (%) Less than 0.04 high school 0.02 Mothers’ Health Problem that 0 1 3 5 9 (1999–2001) (2001–2003) (2003–2006) (2007–2010) Child’s Age-Year

Source: Authors’ compilation based on the Fragile Families and Child Wellbeing Study. mothers’ and fathers’ health 93

Figure 4.4 Fathers’ Health Problem that Limits Work

0.16 0.14 College + 0.12 Some college 0.10

0.08 High school 0.06

Limits Work (%) 0.04 Less than high school 0.02 Fathers’ Health Problem that 0 1 3 5 9 (1999–2001) (2001–2003) (2003–2006) (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

0.20 0.18 0.16 College + 0.14 Some college 0.12 0.10 High school 0.08 0.06 Less than 0.04 high school

Mothers’ Binge Drinking (%) 0.02 0 1 3 5 9 (1999–2001) (2001–2003) (2003–2006) (2007–2010) Child’s Age-Year

Source: Authors’ compilation based on the Fragile Families and Child Wellbeing Study.

Figure 4.6 Fathers’ Binge Drinking

0.50 0.45 College + 0.40 0.35 Some college 0.30 0.25 High school 0.20 Less than 0.15 high school 0.10 Fathers’ Binge Drinking (%) 0.05 0 1 3 5 9 (1999–2001) (2001–2003) (2003–2006) (2007–2010) Child’s Age-Year

Source: Authors’ compilation based on the Fragile Families and Child Wellbeing Study. mothers’ and fathers’ health 95

Figure 4.7 Mothers’ Drug Use

0.09 0.08 College + 0.07 0.06 Some college 0.05 High school 0.04

0.03 Less than 0.02 high school Mothers’ Drug Use (%) 0.01 0 1 3 5 9 (1999–2001) (2001–2003) (2003–2006) (2007–2010) Child’s Age-Year

Source: Authors’ compilation based on the Fragile Families and Child Wellbeing Study.

Figure 4.8 Fathers’ Drug Use

0.25

0.20 College +

Some college 0.15

High school 0.10

Less than Fathers’ Drug Use (%) 0.05 high school

0 1 3 5 9 (1999–2001) (2001–2003) (2003–2006) (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 explana- tions 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’ health 97

Figure 4.9 Effects of a Recession on Mothers’ Health Status

0.7 +31.5%*** +26.3%*** 0.6 +20.0%*** 0.5 +4.0%

0.4 UR 5 percent 0.3 –10.7% Poor (%) UR 10 percent 0.2

0.1

Mothers’ Health Status is Fair or 0 All Less than High Some College + high school school 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-

Figure 4.10 Effects of a Recession on Fathers’ Health Status

0.6 –6.6% 0.5 +2.1% –2.4% +69.2%*** 0.4 UR 5 percent 0.3 –6.2% UR 10 percent

or Poor (%) 0.2 0.1

Fathers’ Health Status is Fair 0 All Less than High Some College + high school school 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

0.25 +71.0%***

0.20 +45.9%*** UR 5 percent +35.2% +55.4%* 0.15 UR 10 percent 0.10 –85.7%*** Limits Work (%) 0.05 Mothers’ Health Problem that 0 All Less than High Some College + high school school 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

0.25 +31.1% 0.20 +31.9%** +43.1% 0.15 UR 5 percent +91.2% UR 10 percent 0.10 +24.0% Limits Work (%) 0.05 Fathers’ Health Problem that 0 All Less than High Some College + high school school college

Source: Authors’ compilation based on the Fragile Families and Child Wellbeing Study. ***p < .01; **p < .05; *p < .1 mothers’ and fathers’ health 99

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

0.25 +122.9%** +51.2%* 0.20 +29.2%* +23.9% –11.1%

0.15 UR 5 percent

0.10 UR 10 percent

0.05 Mothers’ Binge Drinking (%) 0 All Less than High Some College + high school school 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

Figure 4.14 Effects of a Recession on Fathers’ Binge Drinking

0.50 0.45 +8.6% –14.5% –3.8% –16.1% 0.40 –11.9% 0.35 0.30 UR 5 percent 0.25 0.20 UR 10 percent 0.15 0.10

Fathers’ Binge Drinking (%) 0.05 0 All Less than High Some College + high school school 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

0.14 +62.4% +72.5% 0.12 +46.5%* 0.10 +39.1% UR 5 percent 0.08 –78.9% 0.06 UR 10 percent 0.04 Mothers’ Drug Use (%) 0.02 0 All Less than High Some College + high school school college

Source: Authors’ compilation based on the Fragile Families and Child Wellbeing Study. ***p < .01; **p < .05; *p < .1 mothers’ and fathers’ health 101

Figure 4.16 Effects of a Recession on Fathers’ Drug Use

0.25

+3.2% +18.9% 0.20 +18.5% 0.15 +31.5% UR 5 percent UR 10 percent 0.10 +12.0%

Fathers’ Drug Use (%) 0.05

0 All Less than High Some College + high school school 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 exception- ally 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’ health 103 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 self- reported 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’ health 105 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). Table 4.A1 Full Regression Results, Parents’ Physical Health Mothers Fathers With Individual Without Individual With Individual Without Individual Fixed Effects Fixed Effects Fixed Effects Fixed Effects Unemployment rate 0.018*** (0.005) 0.016*** (0.004) 0.001 (0.006) 0.003 (0.004) Relationship status Married -0.075*** (0.022) -0.026** (0.012) Cohabiting 0.008 (0.014) 0.009 (0.015) Mother’s age 0.007*** (0.001) 0.006*** (0.001) Race-ethnicity Black 0.011 (0.017) -0.022* (0.013) Hispanic 0.046* (0.024) 0.015 (0.021) Other -0.005 (0.031) 0.039 (0.038) Immigrant 0.025 (0.022) 0.033 (0.023) Mother’s education High school 0.053*** (0.013) 0.061*** (0.014) Some college or associates degree -0.024** (0.010) -0.053*** (0.015) College + -0.176*** (0.021) -0.141*** (0.020) Number of children in household 0.009** (0.004) 0.007* (0.004) Lived with both parents at age fifteen -0.040*** (0.008) -0.019 (0.013) Interview year 2000 -0.037 (0.030) -0.041 (0.025) -0.031 (0.034) -0.045 (0.037) 2001 -0.036* (0.021) -0.036*** (0.010) -0.041* (0.025) -0.036* (0.019) 2002 -0.046* (0.027) -0.053*** (0.018) -0.079** (0.031) -0.076*** (0.025) 2003 -0.075*** (0.026) -0.071*** (0.015) -0.066** (0.030) -0.066*** (0.024) 2004 0.008 (0.028) 0.00 (0.018) 0.006 (0.032) 0.007 (0.028) 2005 -0.01 (0.026) -0.008 (0.015) -0.017 (0.029) -0.016 (0.029) 2006 -0.11 (0.081) -0.146** (0.063) 0.013 (0.105) -0.047 (0.061) 2007 0.008 (0.033) 0.022** (0.010) 0.010 (0.039) 0.031* (0.016) 2008 0.063** (0.027) 0.048*** (0.016) 0.023 (0.032) 0.024 (0.035) 2009 -0.006 (0.032) 0.003 (0.020) 0.041 (0.035) 0.025 (0.026) 2010 -0.074 (0.055) -0.077** (0.037) 0.012 (0.065) 0.107** (0.051) City Austin 0.013*** (0.005) -0.072*** (0.009) Baltimore -0.035*** (0.010) -0.061*** (0.017) Detroit 0.008 (0.012) -0.063*** (0.017) Newark -0.008 (0.009) -0.081*** (0.014) Philadelphia -0.006 (0.011) -0.060*** (0.017) Richmond 0.042*** (0.011) -0.072*** (0.017) Corpus Christi -0.051*** (0.010) -0.089*** (0.013) Indianapolis 0.020** (0.008) -0.011 (0.012) Milwaukee 0.029*** (0.008) -0.020* (0.010) New York -0.029*** (0.006) -0.093*** (0.010) San Jose -0.071*** (0.008) -0.049*** (0.012) Boston 0.012 (0.008) -0.089*** (0.013) Nashville -0.01 (0.008) -0.108*** (0.012) Chicago -0.023*** (0.008) -0.069*** (0.012) Jacksonville -0.001 (0.009) -0.128*** (0.012) Toledo -0.047*** (0.014) -0.020 (0.015) San Antonio -0.062*** (0.011) -0.076*** (0.013) Pittsburgh 0.005 (0.011) -0.005 (0.016) Norfolk 0.003 (0.009) -0.055*** (0.013) Constant -0.651*** (0.025) -0.802*** (0.029) -0.663*** (0.029) -0.764*** (0.041) N (person * wave) 15,362 15,362 11,890 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 Table 4.A2 Coefficients and Standard Errors, All Outcomes by Maternal Education With Individual Fixed Effects Without Individual Fixed Effects Less than Less than High High Some High High Some All School School College College + All School School College College + Health status is fair or poor Unemployment 0.018*** 0.030*** 0.024*** 0.004 -0.005 0.016*** 0.032*** 0.013 0.007 0.000 rate (model 1) (0.005) (0.009) (0.009) (0.010) (0.013) (0.004) (0.009) (0.008) (0.008) (0.014) Unemployment 0.017*** 0.030*** 0.022** 0.005 -0.004 0.016*** 0.033*** 0.010 0.008 0.002 rate (model 2) (0.005) (0.009) (0.009) (0.010) (0.014) (0.004) (0.009) (0.007) (0.009) (0.013) Increasing -0.000 0.000 -0.001 -0.000 0.001 0 0.000 -0.001*** 0.000 0.001 unemployment (0.000) (0.000) (0.001) (0.001) (0.001) (0.000) (0.001) (0.000) (0.000) (0.001) rate Decreasing 0.001 0.001 -0.002 0.003 -0.001 0.001 0.003 -0.001 0.003 -0.001 unemployment (0.001) (0.002) (0.002) (0.002) (0.003) (0.001) (0.002) (0.002) (0.002) (0.002) rate N (individuals 8,848 2,978 2,721 2,108 627 15,362 4,168 4,218 2,370 1,969 * wave) Number of 2,301 823 735 567 172 4,549 1,206 1,175 673 559 individuals Health problem limits work Unemployment 0.010*** 0.019*** 0.008 0.010* -0.014*** 0.006*** 0.014*** -0.003 0.012** -0.017** rate (model 1) (0.003) (0.006) (0.005) (0.005) (0.006) (0.002) (0.005) (0.004) (0.005) (0.007) Unemployment -0.000 0.001 0.006 0.008 -0.018*** 0.005** 0.014*** -0.004 0.010** -0.019** rate (model 2) (0.001) (0.001) (0.005) (0.005) (0.006) (0.002) (0.005) (0.004) (0.005) (0.008) Increasing 0.009*** 0.015*** -0.000* -0.000 -0.001 -0.000* -0.000 -0.000 -0.000* -0.000 unemployment (0.003) (0.005) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) rate Decreasing -0.000* 0.000 0.001 -0.002 -0.003*** -0.001 -0.000 -0.000 -0.002 -0.003** unemployment (0.000) (0.000) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) rate N (individuals 2,904 1,242 885 654 115 15,345 4,164 4,209 2,367 1,969 * wave) Number of 789 344 239 174 30 4,549 1,206 1,175 673 559 individuals Four or more drinks on one occasion Unemployment 0.008* 0.007 0.014* -0.004 0.026** 0.008** 0.007 0.015** 0.003 0.015 rate (model 1) (0.004) (0.008) (0.008) (0.008) (0.013) (0.004) (0.009) (0.006) (0.006) (0.012) Unemployment 0.000 0.007 0.014* -0.004 0.030** 0.008** 0.006 0.014** 0.002 0.020 rate (model 2) (0.001) (0.008) (0.008) (0.009) (0.013) (0.004) (0.009) (0.006) (0.006) (0.013) Increasing 0.008* -0.000 0.000 -0.001 0.002* -0.000 0.000 -0.000 -0.001 0.002*** unemployment (0.004) (0.000) (0.000) (0.001) (0.001) (0.000) (0.000) (0.000) (0.001) (0.001) rate Decreasing -0.000 -0.000 -0.000 0.000 0.001 -0.000 -0.000 -0.001 -0.000 0.002 unemployment (0.000) (0.002) (0.002) (0.002) (0.003) (0.001) (0.001) (0.001) (0.002) (0.003) rate N (individuals 2,381 785 779 595 216 11,475 3,821 3,574 2,826 1,241 * wave) Number of 844 284 275 208 75 4,434 1,518 1,369 1,069 473 individuals Drug use Unemployment 0.007* 0.010 0.010 0.005 -0.010 0.005 0.008 0.006 0.008 -0.015* rate (model 1) (0.004) (0.007) (0.006) (0.007) (0.009) (0.004) (0.007) (0.008) (0.007) (0.009) Unemployment -0.000 0.009 0.010* 0.005 -0.010 0.005 0.007 0.006 0.007 -0.016* rate (model 2) (0.001) (0.007) (0.006) (0.007) (0.010) (0.005) (0.007) (0.008) (0.007) (0.009) Increasing 0.006* -0.000 0.000 -0.000 0.001** -0.000 -0.000 0.000 -0.000 0.000 unemployment (0.004) (0.000) (0.000) (0.000) (0.001) (0.000) (0.000) (0.000) (0.000) (0.001) rate Decreasing 0.000 -0.001 0.001 -0.000 -0.001 -0.001* -0.001 -0.001 -0.001 -0.001 unemployment (0.000) (0.001) (0.001) (0.001) (0.002) (0.000) (0.001) (0.001) (0.001) (0.002) rate N (individuals 1,565 562 487 401 112 11,229 3,832 3,585 2,828 1,243 * wave) Number of 551 203 170 139 38 4,437 1,520 1,370 1,069 473 individuals 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 Table 4.A3 Coefficients and Standard Errors, All Outcomes by Paternal Education Fathers With Individual Fixed Effects Without Individual Fixed Effects Less than Less than High High Some High High Some All School School College College + All School School College College + Health status is fair or poor Unemployment 0.001 -0.007 -0.002 0.033*** 0.003 0.003 -0.005 0.003 0.028** -0.008 rate (model 1) (0.006) (0.012) (0.010) (0.012) (0.011) (0.004) (0.009) (0.006) (0.011) (0.008) Unemployment 0.002 -0.010 -0.000 0.009 0.009 0.002 -0.005 -0.001 0.031*** -0.009 rate (model 2) (0.006) (0.011) (0.011) (0.010) (0.010) (0.004) (0.009) (0.006) (0.011) (0.008) Increasing -0.000 0.000 0.000 -0.000 -0.000 -0.000* -0.000 -0.000 -0.000 -0.001 unemployment (0.000) (0.001) (0.001) (0.001) (0.001) (0.000) (0.000) (0.001) (0.001) (0.000) rate Decreasing -0.001 0.002 -0.003 -0.003 -0.003 -0.001 0.001 -0.003 0.001 -0.002 unemployment (0.001) (0.002) (0.002) (0.002) (0.002) (0.001) (0.002) (0.002) (0.002) (0.003) rate N (individuals 5,444 1,748 1,717 921 622 11,890 3,307 3,450 2,100 1,851 * wave) Number of 1,598 518 496 261 173 4,026 1,141 1,123 671 556 individuals Health problem limits work Unemployment 0.007** 0.010 0.010 0.010 0.002 0.010*** 0.016** 0.015 0.008 0.001 rate (model 1) (0.003) (0.007) (0.006) (0.007) (0.006) (0.004) (0.007) (0.009) (0.007) (0.006) Unemployment 0.008** 0.012 0.009 0.009 0.009 0.009** 0.016** 0.011 0.009 0.001 rate (model 2) (0.003) (0.007) (0.007) (0.006) (0.006) (0.004) (0.007) (0.009) (0.007) (0.007) Increasing 0.000 0.001* 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 unemployment (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) rate Decreasing 0.001 0.001 -0.001 -0.000 -0.000 -0.000 0.000 -0.002* 0.000 0.001 unemployment (0.001) (0.002) (0.001) (0.001) (0.001) (0.001) (0.002) (0.001) (0.001) (0.002) rate N (individuals 2,008 634 667 336 163 11,884 3,308 3,446 2,098 1,851 * wave) Number of 595 189 196 96 45 4,024 1,140 1,123 670 556 individuals Four or more drinks on one occasion Unemployment -0.003 -0.009 -0.013 0.007 -0.013 -0.004 0.003 -0.010 -0.003 -0.009 rate (model 1) (0.007) (0.014) (0.013) (0.015) (0.015) (0.005) (0.013) (0.008) (0.008) (0.011) Unemployment -0.003 -0.011 -0.013 -0.012 -0.012 -0.004 0.001 -0.010 -0.003 -0.005 rate (model 2) (0.007) (0.015) (0.013) (0.012) (0.012) (0.005) (0.012) (0.010) (0.009) (0.012) Increasing 0.000 0.000 0.001 0.001 0.001 0.000 0.001* 0.000 0.000 -0.000 unemployment (0.000) (0.001) (0.001) (0.001) (0.001) (0.000) (0.001) (0.001) (0.001) (0.001) rate Decreasing -0.001 -0.002 -0.001 -0.002 -0.002 -0.001 -0.002 -0.002 -0.000 0.003 unemployment (0.001) (0.003) (0.003) (0.002) (0.002) (0.002) (0.002) (0.004) (0.003) (0.003) rate N (individuals 2,595 777 800 443 334 8,559 2,339 2,455 1,506 1,340 * wave) Number of 976 296 300 161 120 3,785 1,062 1,054 635 530 individuals (Table continues on p. 112.) Table 4.A3 Continued Fathers With Individual Fixed Effects Without Individual Fixed Effects Less than Less than High High Some High High Some All School School College College + All School School College College + Drug use Unemployment 0.005 0.001 0.006 0.007 0.002 0.006 0.007 -0.001 0.015 0.007 rate (model 1) (0.005) (0.011) (0.010) (0.011) (0.010) (0.005) (0.010) (0.009) (0.011) (0.012) Unemployment 0.005 0.004 0.005 0.013 0.013 0.007 0.010 -0.003 0.018 0.008 rate (model 2) (0.005) (0.012) (0.010) (0.009) (0.009) (0.005) (0.011) (0.008) (0.012) (0.013) Increasing 0.000 0.001** 0 0 0 0.000 0.001 0.000 0.001 -0.001* unemployment (0.000) (0.001) (0.001) (0.001) (0.001) (0.000) (0.001) (0.001) (0.001) (0.000) Decreasing 0.000 0.004 -0.002 0 0 0.001 0.003 -0.000 0.001 -0.001 unemployment (0.001) (0.002) (0.002) (0.002) (0.002) (0.001) (0.002) (0.002) (0.002) (0.002) N (individuals 1,498 482 490 — 149 8,615 2,350 2,459 1,512 1,341 * wave) Number of 562 184 181 54 3,794 1,066 1,055 636 530 individuals 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 Table 4.A4 Sensitivity of Coefficients, Parents’ Health Mothers Fathers With Individual Without Individual With Individual Fixed Without Individual Fixed Effects Fixed Effects Effects Fixed Effects Health status is fair or poor Unemployment rate (model 3) 0.022*** (0.006) 0.020*** (0.005) 0.001 (0.006) 0.002 (0.004) Individual unemployment 0.013 (0.013) 0.046*** (0.009) 0.028* (0.015) 0.055*** (0.016) Unemployment rate (model 4) 0.019*** (0.007) 0.016** (0.006) -0.002 (0.007) -0.001 (0.005) Unemployment rate * year nine 0.002 (0.007) 0.001 (0.007) 0.005 (0.006) 0.007 (0.006) Health problem limits work Unemployment rate (model 3) 0.005* (0.003) 0.004* (0.002) 0.008** (0.003) 0.010*** (0.004) Individual unemployment 0.033*** (0.007) 0.060*** (0.007) -0.005 (0.009) 0.056*** (0.011) Unemployment rate (model 4) 0.012*** (0.004) 0.009*** (0.003) 0.001 (0.004) 0.004 (0.004) Unemployment rate * year nine -0.004 (0.004) -0.004 (0.004) 0.009** (0.004) 0.009* (0.005) Four or more drinks on one occasion Unemployment rate (model 3) 0.009 (0.005) 0.011*** (0.004) -0.003 (0.007) -0.004 (0.005) Individual unemployment 0.006 (0.012) 0.007 (0.008) -0.008 (0.018) -0.004 (0.013) Unemployment rate (model 4) 0.010 (0.006) 0.007 (0.007) -0.006 (0.009) -0.012** (0.006) Unemployment rate * year nine -0.002 (0.006) 0.002 (0.006) 0.004 (0.007) 0.011** (0.005) Drug use Unemployment rate (model 3) 0.006 (0.004) 0.005 (0.004) 0.005 (0.005) 0.005 (0.005) Individual unemployment 0.014 (0.009) 0.038*** (0.005) 0.025* (0.013) 0.054*** (0.015) Unemployment rate (model 4) -0.000 (0.005) -0.000 (0.005) -0.003 (0.007) -0.001 (0.007) Unemployment rate * year nine 0.009** (0.005) 0.007* (0.004) 0.010** (0.005) 0.009** (0.004) 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 Table 4.A5 Coefficients and Standard Errors, Model 1, All Outcomes Mothers Fathers Black Hispanic White Married Cohabiting Single Black Hispanic White Married Cohabiting Single Health status is fair or poor Unemployment 0.023*** 0.019** 0.011 -0.003 0.030*** 0.018** 0.006 -0.001 0.018* -0.003 -0.005 0.013 rate (0.008) (0.009) (0.010) (0.009) (0.008) (0.008) (0.009) (0.011) (0.011) (0.009) (0.009) (0.011) Health problem limits work Unemployment 0.013*** 0.015*** 0.004 0.003 0.014*** 0.010** 0.009 -0.002 0.009 0.010* 0.007 0.002 rate (0.005) (0.005) (0.006) (0.005) (0.005) (0.005) (0.006) (0.006) (0.006) (0.005) (0.005) (0.007) Four or more drinks on one occasion Unemployment 0.010* -0.003 0.021* 0.009 0.007 0.007 -0.003 0.025* -0.022 -0.023** 0.011 0.000 rate (0.006) (0.009) (0.011) (0.008) (0.008) (0.007) (0.010) (0.015) (0.014) (0.011) (0.011) (0.013) Drug use Unemployment 0.019*** 0.002 -0.008 0.001 0.006 0.009 0.005 -0.006 0.007 0.002 -0.003 0.016 rate (0.006) (0.006) (0.008) (0.007) (0.006) (0.006) (0.009) (0.009) (0.010) (0.007) (0.009) (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 mothers’ and fathers’ health 115

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 Well- Being 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 Cause- Specific 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’ health 117

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

n previous chapters, we saw that the Great Recession generated a good Ideal 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’ relationships 119 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 per- ceived 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’ relationships 121

Figure 5.1 Mothers’ Relationship Status

100 90 80 70 60 50 40 30

Percent of Mothers 20 10 0 1 3 5 9 (1999–2001) (2001–2003) (2003–2006) (2007–2010) Child’s Age-Year

Married or cohabiting bio father No coresidential romantic Married or cohabiting new partner

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 college- educated 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 80 70 60 50 40

Percent of Mothers 30 20 10 0 1 3 5 9 (1999–2001) (2001–2003) (2003–2006) (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 80 70 60 50 40

Percent of Mothers 30 20 10 0 1 3 5 9 (1999–2001) (2001–2003) (2003–2006) (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’ relationships 123

Figure 5.4 Mothers’ Reports of Bio Fathers’ Supportiveness

2.0 1.8 1.6 1.4 1.2 1.0 0.8 0.6 0.4

Relationship Supportiveness Score 0.2 0 1 3 5 9 (1999–2001) (2001–2003) (2003–2006) (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

2.0 1.8 1.6 1.4 1.2 1.0 0.8 0.6 0.4

Relationship Supportiveness Score 0.2 0 1 3 5 9 (1999–2001) (2001–2003) (2003–2006) (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 less- educated 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

2.0 1.8 1.6 1.4 1.2 1.0 0.8 0.6 0.4

Relationship Supportiveness Score 0.2 0 1 3 5 9 (1999–2001) (2001–2003) (2003–2006) (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

4.0

3.5

3.0

2.5

2.0

1.5

1.0

Overall Relationship Quality Score 0.5

0 1 3 5 9 (1999–2001) (2001–2003) (2003–2006) (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

4.0

3.5

3.0

2.5

2.0

1.5

1.0

Overall Relationship Quality Score 0.5

0 1 3 5 9 (1999–2001) (2001–2003) (2003–2006) (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’ relationships 127

Figure 5.9 Mothers’ Marriage and Marriage or Cohabitation

70 −7% 60 50 −5% 40 UR 5 percent 30 UR 10 percent 20 Percent of Mothers 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% 80 70 60 50 −17% UR 5 percent 40 +1% −7% UR 10 percent 30

Percent of Mothers 20 10 0 Less than High Some College + high school school 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

Figure 5.11 Mothers’ Marriage or Cohabitation (Bio Father or New Partner)

100 +4% 90 80 −7% 70 −7% −13% 60 UR 5 percent 50 40 UR 10 percent 30

Percent of Mothers 20 10 0 Less than High Some College + high school school college

Source: Authors’ calculations based on Fragile Families and Child Wellbeing Study data. Note: No significant differences in effect of unemployment between subgroups. parents’ relationships 129 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

Figure 5.12 Mothers’ Reports of Bio Fathers’ Supportiveness

1.75 +1% 1.70 −5% 1.65 +3% 0% 1.60 UR 5 percent

Score 1.55 0% UR 10 percent 1.50 1.45

Relationship Supportiveness 1.40 All Less than High Some College + high school school* 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

Figure 5.13 Fathers’ Reports of Mothers’ Supportiveness

1.75 +5% 1.70 −8% −2% −4% −3% 1.65 UR 5 percent 1.60 Score UR 10 percent 1.55 1.50

Relationship Supportiveness 1.45 All* Less than High Some College +† high school** school 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’ relationships 131

Figure 5.14 Mothers’ Reports of New Partners’ Supportiveness

2.5 −40% 2.0 +2% +2% +6% +7%

1.5 UR 5 percent

Score 1.0 UR 10 percent

0.5

Relationship Supportiveness 0 All Less than High Some College +† high school school 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

Figure 5.15 Mothers’ Reports of Quality of Relationship with Bio Father

4.0 3.5 +4% +2% +3% 3.0 +1% +1% 2.5 UR 5 percent 2.0 1.5 UR 10 percent

Quality Score 1.0

Overall Relationship 0.5 0 All Less than High Some College + high school school 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

Figure 5.16 Fathers’ Reports of Quality of Relationship with Bio Mother

4.0 0% 3.5 −2% −4% −6% 3.0 −1% 2.5 2.0 UR 5 percent Score 1.5 UR 10 percent 1.0 0.5

Overall Relationship Quality 0 All Less than High Some College + high school school 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’ relationships 133 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 pre- requisite 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 fore- stall 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’ relationships 135 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’ relationships 137

Table 5.A1 Full Regression Results, Married to or Cohabiting with Father or New Partner With Individual Without Individual Fixed Effects Fixed Effects Unemployment rate -0.056† (0.034) -0.043* (0.019) Education Less than high school — (.) -1.188*** (0.133) High school — (.) -1.079*** (0.108) Some college — (.) -0.998*** (0.116) Mother’s age — (.) 0.006 (0.004) Race-ethnicity Black — (.) -1.127*** (0.070) Hispanic — (.) -0.502*** (0.084) Other — (.) -0.675*** (0.144) Immigrant — (.) 0.732*** (0.147) Children in household — (.) 0.016 (0.017) Lived with both parents at age fifteen — (.) 0.241*** (0.059) Interview year 2000 0.327 (0.210) 0.379* (0.154) 2001 0.025 (0.156) 0.020 (0.143) 2002 0.009 (0.194) 0.101 (0.170) 2003 -0.095 (0.188) -0.067 (0.130) 2004 -0.041 (0.195) 0.053 (0.166) 2005 -0.303 (0.186) -0.190 (0.139) 2006 -0.490 (0.539) -0.478 (0.296) 2007 -0.802** (0.245) -0.447*** (0.103) 2008 -0.206 (0.192) -0.034 (0.157) 2009 -0.144 (0.225) -0.099 (0.157) 2010 -0.304 (0.366) -0.583* (0.286) (Table continues on p. 138.) 138 children of the great recession

Table 5.A1 Continued With Individual Without Individual Fixed Effects Fixed Effects City Austin — (.) -0.143*** (0.030) Baltimore — (.) -0.008 (0.100) Detroit — (.) -0.120 (0.094) Newark — (.) -0.000 (0.081) Philadelphia — (.) -0.055 (0.096) Richmond — (.) -0.263* (0.106) Corpus Christi — (.) 0.183* (0.081) Indianapolis — (.) -0.123 (0.082) Milwaukee — (.) 0.049 (0.078) New York — (.) 0.085 (0.062) San Jose — (.) 0.085 (0.066) Boston — (.) -0.356*** (0.072) Nashville — (.) -0.079 (0.081) Chicago — (.) 0.356*** (0.079) Jacksonville — (.) 0.127 (0.082) Toledo — (.) 0.008 (0.088) San Antonio — (.) 0.111 (0.071) Pittsburgh — (.) -0.373*** (0.086) Norfolk — (.) 0.118 (0.085) Constant 2.200*** (0.201) Observations 7,187 15,855 Number of individuals 1,951 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 Table 5.A2 Coefficients and Standard Errors for Unemployment Rate, Relationship Outcomes With Individual Fixed Effects Without Individual Fixed Effects Less than Less than High High Some High High Some All School School College College + All School School College College + Mother married to father or new partner Unemployment rate -0.017 0.053 -0.026 -0.148† -0.010 -0.038* -0.034 -0.018 -0.069* 0.006 (model 1) (0.044) (0.072) (0.085) (0.085) (0.188) (0.018) (0.025) (0.063) (0.031) (0.092) Unemployment rate -0.013 0.062 -0.043 -0.140 -0.021 0.002 -0.034 -0.026 -0.066* -0.023 (model 2) (0.044) (0.072) (0.086) (0.086) (0.193) (0.006) (0.026) (0.062) (0.032) (0.099) Increasing 0.001 0.009* -0.009† 0.002 -0.006 -0.001 0.001 -0.005† 0.001 -0.010† unemployment rate (0.002) (0.004) (0.005) (0.005) (0.007) (0.001) (0.002) (0.003) (0.003) (0.005) Decreasing 0.010 -0.004 0.022 0.014 0.021 -0.040* -0.004 0.010 0.005 -0.001 unemployment rate (0.010) (0.016) (0.020) (0.018) (0.038) (0.018) (0.009) (0.007) (0.010) (0.014) Observations 4,120 1,543 1,076 1,209 288 15,867 6,136 4,067 3,924 1,740 Number of individuals 1,111 420 287 325 78 4,604 1,823 1,162 1,123 496 Mother married to or cohabiting with father or new partner Unemployment rate -0.056† -0.058 -0.079 -0.061 0.090 -0.043* -0.052† -0.047 -0.051 0.092 (model 1) (0.034) (0.051) (0.065) (0.069) (0.172) (0.019) (0.028) (0.033) (0.045) (0.117) Unemployment rate -0.052 -0.054 -0.076 -0.061 0.114 -0.045* -0.055† -0.053† -0.053 0.088 (model 2) (0.034) (0.052) (0.065) (0.070) (0.174) (0.018) (0.029) (0.032) (0.044) (0.124) Increasing 0.001 0.002 -0.000 -0.002 0.009 0.000 0.000 -0.004 -0.001 -0.002 unemployment rate (0.002) (0.003) (0.004) (0.004) (0.011) (0.003) (0.003) (0.002) (0.002) (0.005) Decreasing 0.006 -0.004 0.018 0.005 0.022 -0.001 -0.010† 0.017† -0.001 0.005 unemployment rate (0.007) (0.011) (0.015) (0.015) (0.040) (0.002) (0.006) (0.010) (0.007) (0.017) Observations 7,187 3,146 2,033 1,698 302 15,855 6,128 4,066 3,922 1,735 Number of individuals 1,953 863 547 460 81 4,603 1,823 1,162 1,122 495 (Table continues on p. 140.) Table 5.A2 Continued With Individual Fixed Effects Without Individual Fixed Effects Less than Less than High High Some High High Some All School School College College + All School School College College + Mother’s report of father’s supportiveness Unemployment rate -0.000 0.000 -0.017* 0.008 0.005 0.001 -0.000 -0.000 0.007 -0.008 (model 1) (0.005) (0.007) (0.006) (0.009) (0.012) (0.005) (0.008) (0.010) (0.010) (0.014) Unemployment rate -0.000 -0.001 -0.014† 0.008 0.004 0.001 0.000 0.001 0.007 -0.012 (model 2) (0.005) (0.007) (0.007) (0.009) (0.013) (0.005) (0.008) (0.011) (0.011) (0.015) Increasing 0.000 -0.001 0.001 -0.001 0.002 0.000 0.000 0.000 0.000 -0.001† unemployment rate (0.001) (0.002) (0.003) (0.002) (0.002) (0.001) (0.000) (0.001) (0.001) (0.001) Decreasing -0.000 -0.000 0.001† -0.000 -0.000 0.000 -0.001 0.004 -0.001 -0.002 unemployment rate (0.000) (0.000) (0.000) (0.000) (0.001) (0.000) (0.002) (0.003) (0.002) (0.002) Observations 7,632 2,438 1,791 1,998 1,401 7,628 2,438 1,791 1,998 1,401 Number of individuals 3,024 1,080 727 771 444 3,022 1,080 727 771 444 Mother’s report of new partners’ supportiveness Unemployment rate 0.009 0.006 0.022 0.026 -0.111† -0.001 -0.001 -0.001 0.011 0.007 (model 1) (0.011) (0.021) (0.013) (0.027) (0.063) (0.008) (0.013) (0.014) (0.026) (0.048) Unemployment rate 0.008 0.003 0.021 0.031 -0.130 -0.002 -0.003 0.001 0.009 0.013 (model 2) (0.012) (0.022) (0.014) (0.024) (0.080) (0.008) (0.012) (0.014) (0.024) (0.065) Increasing -0.000 -0.005 0.002 0.008 0.014 0.000 -0.001 0.001 -0.002* -0.001 unemployment rate (0.003) (0.005) (0.005) (0.005) (0.019) (0.002) (0.001) (0.001) (0.001) (0.005) Decreasing 0.001† 0.001* 0.002 -0.001† -0.004† -0.000 -0.002 0.003 0.001 0.009 unemployment rate (0.000) (0.001) (0.001) (0.001) (0.002) (0.000) (0.004) (0.005) (0.003) (0.030) Observations 1,942 927 541 408 63 1,939 927 541 408 63 Number of individuals 1,259 599 356 264 39 1,258 599 356 264 39 Mother’s report of overall quality of relationship with bio father Unemployment rate 0.011 0.006 0.017 0.004 0.028 0.018 0.008 0.053 0.010 -0.006 (model 1) (0.013) (0.023) (0.027) (0.018) (0.022) (0.018) (0.017) (0.036) (0.023) (0.026) Unemployment rate 0.011 -0.000 0.025 0.009 0.019 0.017 0.005 0.055† 0.014 -0.017 (model 2) (0.013) (0.025) (0.023) (0.017) (0.020) (0.017) (0.017) (0.033) (0.023) (0.024) Increasing 0.000 -0.002 0.005 0.002 -0.006 -0.001 -0.001 0.001 0.001 -0.003 unemployment rate (0.003) (0.007) (0.004) (0.005) (0.005) (0.003) (0.001) (0.002) (0.001) (0.002) Decreasing 0.000 -0.002 0.002 0.001* -0.002 -0.000 -0.006 0.004 0.004 -0.007† unemployment rate (0.001) (0.001) (0.002) (0.001) (0.002) (0.001) (0.004) (0.005) (0.005) (0.004) Observations 7,653 2,453 1,791 2,000 1,405 7,649 2,453 1,791 2,000 1,405 Number of individuals 3,020 1,079 725 770 444 3,018 1,079 725 770 444 Father’s report of mother’s supportiveness Unemployment rate -0.008* -0.024** -0.013 -0.010 0.017† -0.010† -0.024** -0.009 -0.011 0.004 (model 1) (0.004) (0.008) (0.012) (0.007) (0.009) (0.006) (0.008) (0.011) (0.009) (0.010) Unemployment rate -0.009* -0.026** -0.015 -0.009 0.017† -0.010† -0.024** -0.009 -0.011 0.002 (model 2) (0.004) (0.007) (0.011) (0.008) (0.009) (0.006) (0.008) (0.012) (0.008) (0.011) Increasing -0.000 -0.001 -0.001 0.000 -0.000 0.000 0.000 0.000 0.000 -0.000 unemployment rate (0.000) (0.000) (0.000) (0.000) (0.001) (0.000) (0.000) (0.000) (0.000) (0.001) Decreasing -0.001 -0.004 -0.001 0.000 0.001 -0.001 -0.004 -0.001 0.001 -0.002 unemployment rate (0.001) (0.002) (0.002) (0.002) (0.002) (0.001) (0.003) (0.002) (0.002) (0.002) Observations 6,545 2,023 1,525 1,715 1,279 6,542 2,023 1,525 1,715 1,279 Number of individuals 2,672 922 638 689 421 2,670 922 638 689 421 (Table continues on p. 142.) Table 5.A2 Continued With Individual Fixed Effects Without Individual Fixed Effects Less than Less than High High Some High High Some All School School College College + All School School College College + Father’s report of overall quality of relationship with bio mother Unemployment rate -0.014 -0.006 -0.026 -0.036* -0.003 -0.022† -0.001 -0.025 -0.060** -0.014 (model 1) (0.012) (0.010) (0.026) (0.016) (0.029) (0.012) (0.019) (0.025) (0.023) (0.031) Unemployment rate -0.017 -0.012 -0.030 -0.032† -0.003 -0.024† -0.005 -0.027 -0.060* -0.016 (model 2) (0.012) (0.010) (0.026) (0.018) (0.028) (0.013) (0.020) (0.025) (0.024) (0.032) Increasing -0.004 -0.010* -0.010† 0.007 -0.001 -0.006 -0.001 0.001 -0.000 0.001 unemployment rate (0.003) (0.005) (0.005) (0.006) (0.003) (0.004) (0.001) (0.001) (0.001) (0.001) Decreasing -0.000 -0.002 0.000 0.001 0.000 -0.000 -0.005 -0.027 -0.060* -0.016 unemployment rate (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.020) (0.025) (0.024) (0.032) Observations 6,807 2,146 1,585 1,774 1,298 6,803 2,146 1,585 1,774 1,298 Number of individuals 2,773 974 666 708 423 2,771 974 666 708 423 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 Table 5.A3 Sensitivity of Unemployment Rate Coefficients, Relationship Outcomes With Individual Fixed Without Individual Effects Fixed Effects Mother married to father or new partner Unemployment rate (model 1) -0.017 (0.044) -0.038* (0.018) Unemployment rate (model 3) 0.040 (0.053) -0.022 (0.025) Mother’s unemployment -0.081 (0.126) -0.371*** (0.077) Bio-social fathers not employed — — — — Unemployment rate (model 4) -0.067 (0.060) -0.040 (0.027) Unemployment rate * year nine 0.084 (0.069) 0.002 (0.029) Mother married to or cohabiting with father or new partner Unemployment rate (model 1) -0.056† (0.034) -0.043* (0.019) Unemployment rate (model 3) -0.034 (0.079) -0.048* (0.024) Mother’s unemployment -0.056 (0.040) -0.230*** (0.051) Bio-social father’s not employed — — — — Unemployment rate (model 4) -0.111* (0.047) -0.063* (0.026) Unemployment rate * year nine 0.086† (0.051) 0.032 (0.024) Mother’s report of father’s supportiveness Unemployment rate (model 1) -0.000 (0.005) 0.001 (0.005) Unemployment rate (model 3) -0.000 (0.004) -0.001 (0.007) Mother’s unemployment 0.029 (0.018) -0.018 (0.014) Bio-social father’s not employed 0.010 (0.018) -0.048** (0.017) Unemployment rate (model 4) 0.008 (0.007) 0.013 (0.009) Unemployment rate * year nine -0.017* (0.007) -0.024** (0.008) Mother’s report of new partners’ supportiveness Unemployment rate (model 1) 0.009 (0.011) -0.001 (0.008) Unemployment rate (model 3) 0.016 (0.014) 0.007 (0.010) Mother’s unemployment -0.029 (0.028) -0.046** (0.015) Bio-social fathers not employed -0.012 (0.029) -0.032 (0.022) Unemployment rate (model 4) -0.001 (0.014) -0.010 (0.013) Unemployment rate * year nine 0.014 (0.014) 0.011 (0.012) (Table continues on p. 144.) Table 5.A3 Continued With Individual Fixed Without Individual Effects Fixed Effects Mother’s report of overall quality of relationship with bio father Unemployment rate (model 1) 0.011 (0.013) 0.018 (0.018) Unemployment rate (model 3) 0.005 (0.014) -0.002 (0.025) Mother’s unemployment 0.009 (0.048) 0.011 (0.031) Bio-social father’s not employed -0.072 (0.054) -0.189*** (0.038) Unemployment rate (model 4) 0.012 (0.016) 0.036 (0.022) Unemployment rate * year nine -0.002 (0.015) -0.038* (0.018) Father’s report of mother’s supportiveness Unemployment rate (model 1) -0.008* (0.004) -0.010† (0.006) Unemployment rate (model 3) -0.008 (0.005) -0.011 (0.007) Mother’s unemployment 0.008 (0.020) 0.001 (0.014) Bio-social fathers not employed -0.006 (0.015) 0.011 (0.014) Unemployment rate (model 4) -0.005 (0.004) -0.003 (0.006) Unemployment rate * year nine -0.006 (0.006) -0.013† (0.007) Father’s report of overall quality of relationship with bio mother Unemployment rate (model 1) -0.014 (0.012) -0.022† (0.012) Unemployment rate (model 3) -0.038* (0.016) -0.037* (0.016) Mother’s unemployment 0.084† (0.046) -0.066† (0.039) Bio-social fathers not employed -0.091† (0.050) -0.033 (0.035) Unemployment rate (model 4) 0.010 (0.007) 0.025 (0.017) Unemployment rate * year nine -0.044** (0.013) -0.085*** (0.016) 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 Table 5.A4 Coefficients and Standard Errors for Unemployment Rate, Relationship Outcomes Married at Cohabiting Single at White Black Hispanic Baseline at Baseline Baseline Mother married to father or new partner Unemployment rate -0.074 -0.011 0.032 -0.051 -0.040 -0.040 (0.087) (0.081) (0.075) (0.120) (0.066) (0.088) Mother married to or cohabiting with father or new partner Unemployment rate -0.132 -0.023 0.014 -0.041 -0.081 -0.087† (0.082) (0.051) (0.063) (0.119) (0.056) (0.048) Mother’s report of father’s supportiveness Unemployment rate 0.003 -0.002 0.000 0.006 -0.006 -0.002 (0.007) (0.011) (0.005) (0.008) (0.007) (0.010) Mother’s report of new partners’ supportiveness Unemployment rate 0.048† -0.011 -0.006 -0.060 0.009 0.016 (0.025) (0.022) (0.015) (0.037) (0.026) (0.011) Mother’s report of overall quality of relationship with bio father Unemployment rate 0.015 0.013 0.012 0.021† -0.007 0.021 (0.021) (0.022) (0.020) (0.012) (0.017) (0.037) Father’s report of mother’s supportiveness Unemployment rate -0.006 0.005 -0.018*** 0.008 -0.026*** -0.026 (0.005) (0.009) (0.003) (0.005) (0.006) (0.021) Father’s report of overall quality of relationship with bio mother Unemployment rate 0.000 -0.025 -0.031* -0.015 -0.035** 0.041 (0.023) (0.035) (0.013) (0.025) (0.010) (0.031) 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 < .1 146 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’ relationships 147

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

revious chapters in this volume show that big increases in unemploy- Pment 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 fer- tile 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 expec- tations. 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, dis- courages 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 expecta- tions of mothers, children’s relatives, and other community stakeholders, and thereby garner permission to see their children.11 nonresident father involvement 151

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 involvement 153

Figure 6.1 Nonresidence Status

60 College + 50 Some college 40

30 High school 20

Percentage of Fathers Less than 10 high school 0 1359 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

70 60 Child support orders

50 Formal support

40 Informal support 30 In-kind support 20 Visitation

Percentage of Fathers 10 0 1 3 5 9 (1999–2001) (2001–2003) (2003–2006) (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 involvement 155

Figure 6.3 Child Support and Visitation

Days 100 8.0 Formal

7.5 Informal 80 7.0 In-kind 60 6.5 Visited at least 6.0 once a month 40

Percentage of Fathers 5.5 Visitation days per month 20 5.0 (right axis) Less than High Some College + high school school college

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

−20%† $4,000 −13%* 3,500 −26% ** 3,000 2,500 2,000 UR 5 percent 1,500 UR 10 percent Amount Paid −11% 1,000 500 0 All Less than High Some high school school 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 involvement 157

Figure 6.5 Informal Child Support per Year

$1,000 −16% −13% −16% 800 −21%

600 UR 5 percent

400 UR 10 percent Amount Paid 200

0 All Less than High Some high school school 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

100

80 −10% 60 0% −4% +1% UR 5 percent 40 UR 10 percent

20 Percentage of Fathers

0 All Less than High Some high school school 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

6 +7% +10% +11% 5 +11% 4 UR 5 percent 3 UR 10 percent 2

Number of Days 1 0 All Less than High Some high school school 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

100

60 0% −2% +3% UR 5 percent +1% 40 UR 10 percent

Percent of Fathers 20

0 All Less than High Some high school school college +

Source: Authors’ calculations using Fragile Families and Child Wellbeing Study data circa 1998 to 2010. nonresident father involvement 159 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 com- pletely 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 involvement 161

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 endogene- ity 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 involvement 163

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’ transi- tions 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 random- effects 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. Table 6.A1 Full Regression Results, Child Support and Visitation Visited Child at Least Once in a Visitation Days Formal Support ($) Informal Support ($) In-Kind Support (%) Month (%) in a Month Unemployment rate -105.786* (42.779) -28.374 (27.841) -0.000 (0.008) 0.000 (0.006) 0.088 (0.102) Education High school 334.324† (177.230) 52.036 (89.787) 0.023 (0.016) 0.034* (0.016) 0.400 (0.322) diploma or GED Some college 671.730** (205.984) 189.925† (106.701) 0.061*** (0.018) 0.041* (0.020) 0.715† (0.401) College or more 3,100.268*** (482.163) 422.031* (214.283) 0.035 (0.040) 0.037 (0.041) 0.714 (0.847) Father’s age 42.093*** (11.394) -1.371 (5.427) -0.001 (0.001) 0.000 (0.001) 0.025 (0.021) Race-ethnicity Black -1,034.714*** (248.830) 52.626 (123.905) 0.009 (0.023) 0.001 (0.025) 0.958* (0.483) Hispanic -318.475 (320.514) 49.924 (162.010) -0.061* (0.027) -0.068* (0.027) -0.206 (0.560) Other -894.627 (551.623) 85.379 (267.756) 0.028 (0.047) -0.013 (0.042) -0.340 (0.842) Immigrant -22.040 (301.317) 27.267 (134.539) 0.034 (0.025) 0.045† (0.025) 0.440 (0.505) Father working at 603.068*** (169.155) 205.791* (82.863) 0.083*** (0.015) 0.067*** (0.016) 0.712* (0.315) baseline Father incarcerated -102.146 (610.280) -125.858 (279.072) 0.080 (0.050) 0.054 (0.052) 0.028 (1.046) at baseline Father visited mother 413.793* (190.662) 287.043** (89.341) 0.221*** (0.016) 0.228*** (0.016) 3.499*** (0.352) at hospital at birth Mother wants father 416.382 (374.574) 222.048 (147.820) 0.168*** (0.029) 0.132*** (0.031) 1.863** (0.588) involved in child’s life Child is a boy 116.076 (146.169) 78.808 (72.509) 0.004 (0.013) 0.016 (0.013) 0.123 (0.276) Child’s age 26.792 (19.527) -1.776 (8.754) 0.001 (0.002) 0.001 (0.002) 0.019 (0.033) Interview year 2000 1,102.462† (611.478) 39.588 (214.017) -0.007 (0.061) -0.027 (0.072) 0.159 (1.428) 2001 834.838† (503.882) 1.519 (155.537) -0.086† (0.044) -0.100* (0.051) -1.269 (0.979) 2002 1,069.775* (517.190) -239.732 (165.671) -0.197*** (0.050) -0.179*** (0.054) -3.205** (1.023) 2003 1,227.786* (522.906) -143.417 (159.538) -0.212*** (0.049) -0.206*** (0.052) -4.039*** (0.998) 2004 1,365.337** (523.584) -82.673 (167.743) -0.181*** (0.049) -0.189*** (0.051) -3.947*** (1.010) 2005 1,411.739** (520.231) -55.321 (160.673) -0.202*** (0.047) -0.215*** (0.051) -4.402*** (1.018) 2006 2,310.491** (889.802) 366.420 (342.080) -0.142 (0.094) -0.139 (0.099) -4.299* (1.890) 2007 1,520.125* (630.981) 120.470 (303.254) -0.144* (0.069) -0.113 (0.071) -4.266** (1.397) 2008 1,692.102** (530.494) -84.422 (178.650) -0.244*** (0.052) -0.246*** (0.055) -5.550*** (1.046) 2009 1,798.178** (568.166) -151.810 (191.796) -0.298*** (0.060) -0.297*** (0.062) -6.318*** (1.112) 2010 2,097.808** (645.335) -148.872 (245.967) -0.305*** (0.073) -0.319*** (0.073) -6.835*** (1.369) City Austin Baltimore 77.934 (402.413) 59.886 (146.896) 0.016 (0.036) 0.013 (0.036) 0.257 (0.792) Detroit -193.327 (350.910) 169.426 (134.627) 0.067* (0.032) 0.090* (0.036) 1.411† (0.755) Newark -10.339 (348.506) 32.650 (130.840) 0.058† (0.033) 0.081* (0.038) 1.926* (0.786) Philadelphia 254.297 (361.561) 296.350† (152.260) 0.069† (0.039) 0.111** (0.036) 1.975** (0.760) Richmond -145.142 (372.341) 68.065 (136.579) 0.038 (0.034) 0.068† (0.036) 1.807* (0.789) Corpus Christi -85.087 (487.812) -148.804 (210.997) 0.024 (0.054) 0.046 (0.037) 0.652 (0.786) Indianapolis -36.411 (424.041) 294.915† (155.320) 0.075† (0.039) 0.081* (0.040) 1.695* (0.833) Milwaukee -133.686 (347.903) 117.281 (138.939) 0.095** (0.033) 0.126*** (0.038) 2.401** (0.825) New York -181.168 (329.456) -126.829 (146.668) 0.072† (0.039) 0.084* (0.036) 1.514* (0.762) San Jose -116.062 (456.762) 307.224* (155.411) 0.059 (0.037) 0.076* (0.038) 1.534† (0.822) Boston 273.718 (509.043) 219.914 (152.454) 0.057 (0.038) 0.084† (0.043) 1.222 (0.865) Nashville 305.899 (505.603) 157.174 (190.258) 0.029 (0.047) 0.043 (0.053) 0.975 (1.108) Chicago -323.019 (444.805) 264.978 (192.258) 0.052 (0.047) 0.088† (0.051) 1.374 (1.105) Jacksonville 87.072 (535.984) 104.238 (183.857) 0.046 (0.047) 0.043 (0.050) 0.734 (1.049) Toledo 50.755 (456.268) 0.978 (187.441) 0.014 (0.049) 0.026 (0.052) 0.614 (1.102) San Antonio 34.456 (557.841) -246.391 (254.021) 0.001 (0.068) 0.066 (0.053) 1.182 (1.132) Pittsburgh -156.824 (524.408) -120.042 (217.213) 0.069 (0.055) 0.137* (0.058) 0.890 (1.218) Norfolk -339.945 (439.780) -93.464 (208.587) 0.056 (0.054) 0.101† (0.052) 1.563 (1.126) Constant 138.471 (1,074.857) -35.449 (380.128) 0.327** (0.102) 0.009 (0.052) 0.388 (1.112) Observations 4,068 9,206 9,206 8,965 8,965 Number of 1,678 3,353 3,353 3,157 3,157 individuals 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 < .1 166 children of the great recession

Table 6.A2 Coefficients and Standard Errors, Rate of Change, Father Involvement Less than Some High High College and All School School College + Formal child support Unemployment rate -105.786* -16.853 -168.682** -152.260† (model 1) (42.779) (29.482) (60.606) (87.879) Unemployment rate -101.144* -22.849 -175.896** -143.482 (model 2) (44.039) (30.946) (64.206) (93.080) Increasing unemployment -2.356 -0.204 -1.471 -0.780 (2.378) (1.451) (3.893) (5.271) Decreasing unemployment -7.309 -6.349 -15.518 6.423 (11.825) (7.606) (17.621) (26.787) Observations 4,068 3,395 1,612 869 Number of individuals 1,678 1,285 683 370 Informal child support Unemployment rate -28.374 -29.800 -21.246 -24.683 (model 1) (27.841) (43.363) (42.686) (59.046) Unemployment rate -10.017 -6.440 -8.119 -10.400 (model 2) (30.459) (44.910) (46.500) (60.059) Increasing unemployment -1.049 -1.710 -1.025 -0.395 rate (1.504) (2.028) (2.407) (3.409) Decreasing unemployment 12.637 13.392 8.494 11.895 rate (7.895) (10.552) (11.803) (17.238) Observations 4,068 3,395 1,612 869 Number of individuals 1,678 1,285 683 370 In-kind child support Unemployment rate -0.000 0.001 -0.004 0.002 (model 1) (0.008) (0.010) (0.010) (0.013) Unemployment rate 0.001 0.002 -0.002 0.005 (model 2) (0.008) (0.012) (0.011) (0.013) Increasing unemployment -0.000 -0.000 -0.000 -0.000 rate (0.000) (0.000) (0.000) (0.001) Decreasing unemployment 0.000 -0.000 -0.000 0.001 rate (0.001) (0.002) (0.002) (0.003) nonresident father involvement 167

Table 6.A2 Continued Less than Some High High College and All School School College + Observations 4,068 3,395 1,612 869 Number of individuals 1,678 1,285 683 370 Visited child at least once in the last month Unemployment rate 0.000 0.001 -0.002 0.002 (model 1) (0.006) (0.009) (0.009) (0.011) Unemployment rate 0.001 0.003 -0.003 0.004 (model 2) (0.006) (0.009) (0.009) (0.011) Increasing unemployment -0.001 -0.001 -0.000 -0.001 rate (0.000) (0.001) (0.001) (0.001) Decreasing unemployment -0.001 -0.000 -0.003 0.000 rate (0.002) (0.003) (0.002) (0.003) Observations 8,965 3,248 3,425 1,829 Number of individuals 3,157 1,192 1,227 708 Visitation days in a month Unemployment rate 0.088 0.085 0.097 0.067 (model 1) (0.102) (0.169) (0.156) (0.209) Unemployment rate 0.112 0.128 0.091 0.117 (model 2) (0.104) (0.173) (0.156) (0.220) Increasing unemployment -0.012* -0.015† -0.009 -0.011 rate (0.005) (0.008) (0.008) (0.011) Decreasing unemployment -0.010 -0.001 -0.036 0.023 rate (0.025) (0.039) (0.042) (0.059) Observations 8,965 3,248 3,425 1,829 Number of individuals 3,157 1,192 1,227 708 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 Standard Models Coefficients Errors Formal child support Unemployment rate (model 1) -105.786* (42.779) Unemployment rate (model 3) -87.790* (42.582) Mother’s unemployment -189.256 (115.192) Bio-social fathers not employed -563.707*** (109.329) Unemployment rate (model 4) -106.541† (54.912) Unemployment rate * year nine -2.170 (56.114) Informal child support Unemployment rate (model 1) -28.374 (27.841) Unemployment rate (model 3) -23.971 (27.809) Mother’s unemployment -8.955 (73.953) Bio-social father’s not employed -182.413** (66.977) Unemployment rate (model 4) -34.514 (31.493) Unemployment rate * year nine 22.703 (36.307) In-kind child support Unemployment rate (model 1) -0.000 (0.008) Unemployment rate (model 3) 0.001 (0.008) Mother’s unemployment -0.003 (0.014) Bio-social fathers not employed -0.066*** (0.013) Unemployment rate (model 4) -0.005 (0.009) Unemployment rate * year nine 0.012† (0.007) Visited child at least once in the last month Unemployment rate (model 1) 0.000 (0.006) Unemployment rate (model 3) 0.002 (0.006) Mother’s unemployment -0.005 (0.014) Bio-social fathers not employed -0.079*** (0.013) Unemployment rate (model 4) -0.004 (0.007) Unemployment rate * year nine 0.011† (0.007) Visitation days in a month Unemployment rate (model 1) 0.088 (0.102) Unemployment rate (model 3) 0.120 (0.102) Mother’s unemployment -0.119 (0.254) Bio-social fathers not employed -1.334*** (0.262) Unemployment rate (model 4) 0.042 (0.122) Unemployment rate * year nine 0.132 (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 involvement 169

Table 6.A4 Coefficients and Standard Errors, Model 1, Child Support and Visitation Outcomes Black Hispanic White Formal child support Unemployment rate -81.399† -114.297** -108.447 (47.229) (42.379) (153.350) Informal child support Unemployment rate -36.303 -31.333 -3.935 (32.780) (54.459) (78.532) In-kind child support Unemployment rate -0.002 -0.022 0.001 (0.008) (0.012) (0.016) Visited child at least once in the last month Unemployment rate 0.002 -0.005 0.006 (0.007) (0.009) (0.015) Visitation days in a month Unemployment rate 0.077 0.009 0.247 (0.131) (0.183) (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 Chase- Landsdale 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 involvement 171

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

n previous chapters, we see that recessions take an economic toll on fam- Iilies. 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-year- old, 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 puni- tive 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’ parenting 175 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 ado- lescence they gain greater autonomy and parents perform less monitoring 176 children of the great recession

Figure 7.1 High-Frequency Maternal Spanking by Education

1.0 0.9 College + 0.8 0.7 Some college 0.6 0.5 High school Mean 0.4 0.3 Less than 0.2 high school 0.1 0 1 3 5 9 (1999–2001) (2001–2003) (2003–2006) (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’ parenting 177

Figure 7.2 High-Frequency Maternal Physical Aggression by Education

1.0 0.9 College + 0.8 0.7 Some college 0.6 0.5

Mean High school 0.4 0.3 Less than 0.2 high school 0.1 0 1 3 5 9 (1999–2001) (2001–2003) (2003–2006) (2007–2010) Child’s Age-Year

Source: Authors’ calculations based on data from the Fragile Families and Child Wellbeing Study.

Figure 7.3 High-Frequency Maternal Psychological Aggression by Education

1.0 0.9 College + 0.8 0.7 Some college 0.6 0.5 High school Mean 0.4 0.3 Less than 0.2 high school 0.1 0 1 3 5 9 (1999–2001) (2001–2003) (2003–2006) (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

Figure 7.4 High-Frequency Maternal Warmth by Education

1.0 0.9 College + 0.8 0.7 Some college 0.6 0.5 High school Mean 0.4 0.3 Less than 0.2 high school 0.1 0 1 3 5 9 (1999–2001) (2001–2003) (2003–2006) (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-year- olds 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’ parenting 179

Figure 7.5 High-Frequency Maternal Parenting Activities by Education

1.0 0.9 College + 0.8 0.7 Some college 0.6 0.5 High school Mean 0.4 0.3 Less than 0.2 high school 0.1 0 1 3 5 9 (1999–2001) (2001–2003) (2003–2006) (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

Figure 7.6 High-Frequency Paternal Spanking by Education

1.0 0.9 College + 0.8 0.7 Some college 0.6 0.5 High school Mean 0.4 0.3 Less than 0.2 high school 0.1 0 1 3 5 9 (1999–2001) (2001–2003) (2003–2006) (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

Figure 7.7 High-Frequency Paternal Physical Aggression by Education

1.0 0.9 College + 0.8 0.7 Some college 0.6 0.5 High school Mean 0.4 0.3 Less than 0.2 high school 0.1 0 1 3 5 9 (1999–2001) (2001–2003) (2003–2006) (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 less- educated 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’ parenting 181

Figure 7.8 High-Frequency Paternal Psychological Aggression by Education

1.0 0.9 College + 0.8 0.7 Some college 0.6 0.5 High school Mean 0.4 0.3 Less than 0.2 high school 0.1 0 1 3 5 9 (1999–2001) (2001–2003) (2003–2006) (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 college- educated 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

Figure 7.9 High-Frequency Maternal Spanking by Unemployment Rate

80 70 60 50 UR 5 percent 40 UR 10 percent

Percent 30 –44% –53% –22% –24% 20 0% 10 0 All Less than High Some College + high school school* 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.

Figure 7.10 High-Frequency Maternal Physical Aggression by Unemployment Rate

80 70 60 UR 5 percent 50 40 UR 10 percent –52% –55% –40% –4%

Percent 30 –95% 20 10 0 All Less than High Some College + high school school 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’ parenting 183

Figure 7.11 High-Frequency Maternal Psychological Aggression by Unemployment Rate

80 +18% 70 –30% 60 –4% +8% –18% 50 UR 5 percent 40 UR 10 percent

Percent 30 20 10 0 All Less than High Some College + high school school 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

80 –12% 70 –3% –2% +5% 60 +2% 50 UR 5 percent 40

Percent 30 UR 10 percent 20 10 0 All Less than High Some College + high school school college

Figure 7.13 High-Frequency Maternal Parenting Activities by Unemployment Rate

+12% –3% 80 +1% 0% –8% 70 60 50 UR 5 percent 40 UR 10 percent Percent 30 20 10 0 All Less than High Some College + high school school 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’ parenting 185

Figure 7.14 High-Frequency Paternal Spanking by Unemployment Rate

50 –29% –21% –31% 40 –30% –47% 30 UR 5 percent Percent 20 UR 10 percent

0 All Less than High Some College + high school school 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 inter- preting 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 50 40 UR 5 percent

30 UR 10 percent Percent 20 –46% –33% –39% –43% –91% 10 0 All Less than High Some College + high school school 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 +43% +25% 50 40 –3% –28% –35% UR 5 percent 30 UR 10 percent Percent 20 10 0 All Less than High Some College +* high school school 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’ parenting 187

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 ori- gin), 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’ parenting 189 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 high- frequency 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’ parenting 191

Table 7.A1 Full Regression Results, Maternal Parenting High Frequency Maternal Spankinga With Individual Fixed Without Individual Effects Fixed Effects Mothers Unemployment rate -0.01** (-2.64) -0.02*** (-3.45) Education Less than high school 0.01 (0.59) High school 0.04* (2.23) Some college 0.03* (2.18) Relationship status Married 0.03* (2.47) Cohabiting -0.00 (-0.03) Mother’s age -0.00*** (-5.41) Race-ethnicity Black -0.01 (-0.55) Hispanic -0.04* (-2.54) Other 0.04 (1.27) Immigrant -0.05*** (-3.81) Children in household -0.01*** (-3.30) Lived with both parents at -0.01 (-0.96) age fifteen Interview year 2000 0.11* (2.18) 0.03 (1.02) 2001 0.11** (2.74) 0.02 (0.89) 2002 0.21*** (4.71) 0.07† (1.91) 2003 0.20*** (4.50) 0.05† (1.67) 2004 0.11* (2.54) -0.03 (-1.09) 2005 0.60*** (3.54) 0.01 (0.31) 2006 0.28* (2.21) 0.04 (0.41) 2007 -0.08† (-1.72) -0.16*** (-15.6) 2008 -0.01 (-0.31) -0.14*** (-5.70) 2009 0.04 (0.93) -0.11*** (-3.58) 2010 0.09 (1.22) -0.07† (-1.72) (Table continues on p. 192.) 192 children of the great recession

Table 7.A1 Continued High Frequency Maternal Spankinga With Individual Fixed Without Individual Effects Fixed Effects City Austin -0.02*** (-3.47) Baltimore -0.07*** (-4.32) Detroit 0.05** (2.86) Newark -0.03† (-1.67) Philadelphia -0.06** (-3.14) Richmond -0.01 (-0.74) Corpus Christi -0.02 (-1.51) Indianapolis 0.01 (0.60) Milwaukee -0.05*** (-3.34) New York -0.07*** (-4.86) San Jose -0.01 (-0.50) Boston -0.07*** (-4.80) Nashville 0.05*** (3.41) Chicago -0.01 (-0.39) Jacksonville -0.02 (-1.34) Toledo 0.00 (0.01) San Antonio -0.05** (-2.75) Pittsburgh -0.05** (-2.71) Norfolk -0.00 (-0.14) Constant 0.12** (2.66) 0.40*** (11.10) Observations 13,3369 13,369 Number of individuals 4,502 4,604 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. aIncludes years one through nine. ***p < .001; **p < .01; *p < .05; †p < .1 Table 7.A2 Coefficients and Standard Errors, Rate of Change in Unemployment for Maternal Parenting Outcomes With Individual Fixed Effects Without Individual Fixed Effects Less than Less than High High Some High High Some All School School College College + All School School College College + Mothers High-frequency maternal spanking (z-stat)a Unemployment -0.01** -0.02† -0.01 -0.01 -0.05** -0.02*** -0.01 -0.01* -0.01 -0.04* rate (model 1) (-2.64) (-1.77) (-0.70) (-0.75) (-3.08) (-3.45) (-1.57) (-2.08) (-1.03) (-2.12) Unemployment -0.02* -0.02 -0.01 -0.01 -0.03 -0.02*** -0.01 -0.01* -0.01 -0.04* rate (model 2) (-2.15) (-1.63) (-0.62) (-0.65) (-1.45) (-3.91) (-1.60) (-1.95) (-1.12) (-2.10) Increasing -0.00 -0.00 -0.00 0.00 0.00 0.00* 0.00 0.00 0.00 0.00 unemployment (-0.15) (-0.33) (-1.29) (1.04) (0.38) (2.02) (1.02) (0.63) (0.39) (0.83) rate Decreasing 0.00 0.00 0.00 -0.00 0.01 0.00* -0.00* 0.00 -0.00 0.00 unemployment (0.69) (0.55) (0.63) (-0.71) (1.35) (2.13) (-2.44) (0.91) (-0.69) (-0.15) rate Observations 133,369 4,458 4,098 3,368 1,445 133,369 4,458 4,089 3,368 1,445 Number of 4,502 1,554 1,356 1,108 484 4,604 1,593 1,391 1,124 496 individuals High-frequency maternal physical aggression (z-stat)b Unemployment -0.02*** -0.02† -0.02† -0.02 -0.03† -0.03*** -0.05*** -0.05* -0.05** -0.04 rate (model 1) (-3.32) (-1.88) (-1.68) (-1.45) (-1.74) (-5.78) (-3.73) (-2.46) (-2.91) (-1.20) Unemployment -0.02*** -0.02† -0.02† -0.02 -0.03† -0.03*** -0.02† -0.03*** -0.02† -0.03 rate (model 2) (-3.32) (-1.88) (-1.68) (-1.45) (-1.74) (-5.90) (-1.79) (-4.66) (-1.80) (-1.45) Increasing 0.00 0.00 0.00 0.00 0.00 0.00* 0.00† 0.00 0.00 0.00 unemployment (0.78) (0.54) (0.69) (0.02) (0.20) (2.31) (1.81) (0.95) (0.20) (1.56) rate (Table continues on p. 194.) Table 7.A2 Continued With Individual Fixed Effects Without Individual Fixed Effects Less than Less than High High Some High High Some All School School College College + All School School College College + Decreasing 0.00* 0.00 0.00 0.00 0.01 0.00 0.00† 0.00 -0.00 0.00 unemployment (2.10) (1.22) (1.12) (0.55) (1.42) (1.17) (1.77) (0.53) (-0.36) (0.92) rate Observations 9,080 3,001 2,798 2,308 973 9,080 3,001 2,798 2,308 973 Number of 4,068 1,398 1,255 994 421 4,068 1,398 1,255 994 421 individuals High-frequency maternal psychological aggression (z-stat)b Unemployment -0.00 0.01 0.02† -0.04** -0.02 -0.01† -0.01 -0.01 -0.06** 0.02 rate (model 1) (-0.48) (0.60) (1.70) (-2.61) (-0.71) (-1.77) (-0.42) (-0.39) (-3.07) (0.47) Unemployment -0.00 0.01 0.02† -0.03* -0.01 -0.01 0.01 0.00 -0.04** -0.02 rate (model 2) (-0.12) (0.91) (1.81) (-2.53) (-0.55) (-1.61) (0.77) (0.34) (-3.11) (-1.01) Increasing 0.00 0.00 0.00 -0.00 -0.00 0.00 0.00 0.00 -0.00 0.00 unemployment (0.45) (1.03) (0.34) (-0.54) (-0.50) (1.57) (1.27) (1.38) (-0.40) (0.46) rate Decreasing 0.00** 0.01* 0.00 0.00 0.01 0.00† 0.01* 0.00 -0.00 -0.00 unemployment (2.56) (2.03) (1.60) (0.02) (0.98) (1.68) (2.43) (1.31) (-0.89) (-0.11) rate Observations 9,143 3, 030 2,813 2,321 979 9,143 3, 030 2,813 2,321 979 Number of 4,071 1,399 1,256 995 421 4,071 1,399 1,256 995 421 individuals High-frequency maternal warmth (z-stat)b,c Unemployment -0.00 0.00 0.01 -0.00 -0.02 -0.00 -0.01 0.04 -0.03 -0.02 rate (model 1) (-0.10) (0.15) (0.36) (-0.21) (-0.55) (-0.05) (-0.18) (1.05) (-0.57) (-0.43) Unemployment -0.00 0.01 0.01 -0.01 -0.03 0.00 0.01 0.02 -0.02 -0.04 rate (model 2) (-0.15) (0.29) (0.25) (-0.32) (-0.84) (0.01) (0.44) (0.96) (-0.90) (-1.06) Increasing 0.00** 0.00† 0.00 0.00† -0.00 0.00† 0.00 0.00† 0.00* -0.00 rate of (2.74) (1.78) (1.51) (1.84) (-0.43) (1.67) (1.18) (1.67) (2.39) (-0.24) unemployment Decreasing 0.00 0.00 -0.00 0.00 -0.01 0.00 0.00 -0.00 0.00 -0.00 rate of (0.12) (1.09) (-0.61) (0.10) (-1.43) (0.33) (1.23) (-0.52) (0.37) (-0.40) unemployment Observations 6,743 2,256 2,112 1,701 674 6,743 2,256 2,112 1,701 674 Number of 3,550 1,203 1,110 887 350 3,550 1,203 1,110 887 350 individuals High-frequency maternal parenting activities (z-stat)b Unemployment 0.00 -0.00 -0.01 0.02 -0.01 -0.00 0.02 0.02 0.05* 0.02 rate (model 1) (0.24) -0.01 (-0.44) (1.61) (-0.68) (-0.02) (1.26) (1.32) (2.35) (0.46) Unemployment 0.00 -0.00 -0.01 0.02 -0.01 -0.00 0.00 -0.01 0.01 -0.01 rate (model 2) (0.10) (-0.01) (-0.53) (1.33) (-0.43) (-0.08) (0.02) (-0.45) (0.69) (-0.31) Increasing 0.00 0.00 0.00 0.00† -0.00 0.00 0.00 0.00 0.00* -0.00 rate of (1.62) (1.39) (0.23) (1.83) (-0.16) (1.00) (0.78) (0.26) (2.07) (-0.93) unemployment Decreasing -0.00 0.00 -0.00 -0.00 0.00 -0.00 -0.00 -0.00 -0.00 0.00 rate of (-0.50) (0.46) (-0.70) (-0.83) (0.97) (-0.50) (-0.15) (-0.55) (-0.61) (0.90) unemployment Observations 11,237 3,683 3,465 2,821 1,268 11,237 3,683 3,465 2,821 1,268 Number of 4,425 1,506 1,347 1,086 486 4,425 1,506 1,347 1,086 486 individuals 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. aIncludes years one through nine. bIncludes years three through nine. cA subsample of families received in-home visits and were assessed for warmth. ***p < .001; **p < .01; *p < .05; †p < .1 196 children of the great recession

Table 7.A3 Sensitivity of Coefficients, Maternal Parenting Outcomes With Individual Fixed Without Individual Effects Fixed Effects Mothers High-frequency maternal spanking (z-stat)a Unemployment rate (model 1) -0.01** (-2.64) -0.02*** (-3.45) Unemployment rate (model 3) -0.03* (-2.11) -0.02*** (-3.85) Mother’s unemployment 0.02 (0.68) -0.01 (-0.65) Bio-social fathers not employed -0.05† (-1.64) -0.03 (-1.60) Unemployment rate (model 4) -0.03*** (-3.76) -0.03*** (-3.62) Unemployment rate * year nine 0.02** (2.68) 0.02*** (3.36) High-frequency maternal physical aggression (z-stat)b Unemployment rate (model 1) -0.02*** (-3.63) -0.03*** (-5.78) Unemployment rate (model 3) -0.06** (-2.57) -0.05* (-2.26) Mother’s unemployment 0.03 (0.93) -0.01 (-0.80) Bio-social fathers not employed 0.00 (0.05) 0.02 1.49 Unemployment rate (model 4) -0.04*** (-4.92) -0.05*** (-4.95) Unemployment rate * year nine 0.03*** (3.32) 0.03*** (3.66) High-frequency maternal psychological aggression (z-stat)b Unemployment rate (model 1) -0.00 (-0.48) -0.01† (-1.77) Unemployment rate (model 3) -0.04 (-1.55) -0.01 (-0.51) Mother’s unemployment 0.03 (0.90) 0.01 0.37 Bio-social fathers not employed -0.02 (-0.47) 0.01 (0.44) Unemployment rate (model 4) -0.01 (-1.30) -0.02† (-1.67) Unemployment rate * year nine 0.01 (1.35) 0.01 (1.27) High-frequency maternal warmth (z-stat)b,c Unemployment rate (model 1) -0.00 (-0.10) -0.00 (-0.05) Unemployment rate (model 3) 0.04 (0.95) 0.02 (0.40) Mother’s unemployment -0.07 (-1.39) -0.08*** (-3.39) Bio-social fathers not employed 0.03 (0.57) -0.02 (-0.72) Unemployment rate (model 4) 0.00 (0.04) -0.00 (-0.11) Unemployment rate * year nine -0.00 (-0.14) 0.00 (0.15) High-frequency maternal parenting activities (z-stat)b Unemployment rate (model 1) 0.00 (0.24) -0.00 (-0.02) Unemployment rate (model 3) 0.01 (0.38) -0.00 (-0.24) Mother’s unemployment 0.01 (0.52) 0.02 (1.15) Bio-social fathers not employed 0.01 (0.43) -0.02 (-1.09) Unemployment rate (model 4) 0.02** (2.74) 0.03* (1.98) Unemployment rate * year nine -0.03*** (-3.62) -0.03*** (-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. aIncludes years one through nine. bIncludes years three through nine. cA subsample of families received in-home visits and were assessed for warmth. ***p < .001; **p < .01; *p < .05; †p < .1 Table 7.A4 Coefficients and Standard Errors, Maternal Parenting Outcomes by Subgroup With Individual Fixed Effects Black Hispanic White Married Cohabiting Single Mothers High-frequency maternal spanking (z-stat)a Unemployment rate -0.01 -0.02† -0.02 -0.01 -0.02† -0.01 (-0.60) (-1.62) (-1.24) (-1.27) (-0.65) (-0.56) High-frequency maternal physical aggression (z-stat)b Unemployment rate -0.02† -0.02† -0.01 -0.03* -0.02 -0.03* (-1.82) (-1.64) (-0.69) (-2.43) (-1.54) (-2.35) High-frequency maternal psychological aggression (z-stat)b Unemployment rate -0.02† 0.00 0.00 -0.00 -0.00 -0.01 (-1.95) (0.27) (0.09) (-0.21) (-0.05) (-0.60) High-frequency maternal warmth (z-stat)b,c Unemployment rate -0.04* 0.04† -0.04† -0.03 0.02 -0.01 (-2.37) (1.94) (-1.78) (-1.39) (1.32) (-0.33) High-frequency parenting activities (z-stat) Unemployment rate -0.01 -0.01 0.01 -0.01 0.01 -0.01 (-0.71) (-0.75) (0.67) (-0.40) (1.06) (-0.50) 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. aIncludes years one through nine. bIncludes years three through nine. cA subsample of families received in-home visits and were assessed for warmth. *p < 0.05; †p < 0.1 198 children of the great recession

Table 7.A5 Full Regression Results, Paternal Parenting High-Frequency Paternal Spankinga With Individual Fixed Without Individual Effects Fixed Effects Fathers Unemployment rate -0.01 (-1.63) -0.01† (-1.75) Education Less than high school 0.05* (2.25) High school 0.08*** (3.45) Some college 0.08*** (4.07) Relationship status Married 0.09*** (7.07) Cohabiting 0.05*** (3.84) Fathers’ age -0.00** (-2.85) Race-ethnicity Black -0.04* (-2.00) Hispanic -0.06* (-2.08) Other -0.03 (-0.91) Immigrant -0.06* (-2.16) Children in household -0.01† (-1.95) Lived with both parents at age nineteen -0.03* (-2.39) Interview Year 2000 — (—) — (—) 2001 0.27*** (3.88) 0.17** (3.00) 2002 0.30*** (5.04) 0.21*** (4.63) 2003 0.24*** (4.46) 0.15*** (3.47) 2004 0.23*** (3.82) 0.13** (2.62) 2005 0.17** (2.95) 0.07† (1.74) 2006 0.09 (0.89) — (—) 2007 -0.01 (-0.09) -0.12† (-1.94) 2008 -0.04 (-0.69) -0.12* (-2.41) 2009 -0.06 (-1.21) -0.15** (-2.92) 2010 — (—) -0.12* (-2.05) mothers’ and fathers’ parenting 199

Table 7.A5 Continued High-Frequency Paternal Spankinga With Individual Fixed Without Individual Effects Fixed Effects City Austin 0.09*** (12.10) Baltimore 0.02 (0.82) Detroit 0.11*** (6.03) Newark 0.01 (1.01) Philadelphia 0.03* (2.22) Richmond 0.09*** (5.66) Corpus Christi 0.13*** (10.10) Indianapolis 0.11*** (9.50) Milwaukee 0.08*** (9.01) New York 0.04*** (4.08) San Jose 0.12*** (8.96) Boston -0.01 (-0.68) Nashville 0.23*** (20.20) Chicago 0.06*** (5.59) Jacksonville 0.11*** (9.24) Toledo 0.11*** (5.47) San Antonio 0.15*** (10.50) Pittsburgh 0.16*** (10.90) Norfolk 0.13*** (8.86) Constant 0.14† (1.77) 0.16** (2.73) Observations 9,142 9,142 Number of individuals 3,220 3,220 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. aYears one through nine. ***p < .001; **p < .01; *p < .05; †p < .1 Table 7.A6 Coefficients and Standard Errors, Rate of Change in Unemployment for Paternal Parenting Outcomes With Individual Fixed Effects Without Individual Fixed Effects Less than Less than High High Some High High Some All School School College College + All School School College College + Fathers High-frequency paternal spanking (z-stat)a Unemployment -0.01 -0.01 -0.02 0.00 -0.01 -0.01† -0.01 -0.02† -0.01 -0.01 rate (model 1) (-1.63) (-1.10) (-1.48) (0.16) (-0.59) (-1.75) (-1.35) (-1.69) (-0.39) (-1.18) Unemployment -0.01† -0.01 -0.02† -0.00 -0.01 -0.01† -0.01 -0.02* -0.01 0.01 rate (model 2) (-1.71) (-0.98) (-1.67) (-0.01) (-0.41) (-1.84) (-1.23) (-2.00) (-0.71) (-0.64) Increasing -0.00 -0.00 0.00 -0.00 0.00* 0.00 -0.00* 0.00 -0.00 0.00** unemployment (-0.64) (-1.06) (0.19) (-1.26) (2.03) (0.16) (-2.00) (1.39) (-0.93) (2.74) rate Decreasing -0.00 0.00 -0.00 -0.00 0.00 -0.00 0.00 -0.00* -0.01* 0.00 unemployment (-0.85) (0.41) (-1.38) (-0.94) (0.89) (-1.12) (0.29) (-2.37) (-1.96) (0.71) rate Observations 9,142 2,203 1,774 1,455 890 9,142 2,203 1,774 1,455 890 Number 3,220 1,071 829 645 364 3,220 1,071 829 645 364 of individuals High-frequency paternal physical aggression (z-stat)b Unemployment -0.01† -0.01 -0.01 -0.01 -0.02 -0.02*** -0.02** -0.02† -0.01 -0.02 rate (model 1) (-1.66) (-0.67) (-0.67) (-0.82) (-1.17) (-4.04) (-2.66) (-1.67) (-0.69) (-1.15) Unemployment -0.01† -0.01 -0.01 -0.01 -0.02 -0.02*** -0.02** -0.02† -0.01 -0.01 rate (model 2) (-1.68) (-0.70) (-0.60) (-0.79) (-1.10) (-3.97) (-2.61) (-1.63) (-0.77) (-0.90) Increasing -0.00 -0.00 -0.00 0.00 -0.00 0.00 0.00 0.00 0.00 0.00 unemployment (-0.63) (-0.25) (-0.94) (0.41) (-0.15) (0.61) (0.33) (0.57) (0.00) (0.82) rate Decreasing -0.00 -0.00 0.00 -0.00 0.00 -0.00 -0.00 0.00 -0.00 0.00 unemployment (-0.18) (-0.29) (0.05) (-0.13) (0.28) (-0.31) (-0.23) (0.67) (-1.37) (0.35) rate Observations 4,919 1,736 1,333 1,158 692 4,919 1,736 1,333 1,158 692 Number 2,659 993 742 603 321 2,659 993 742 603 321 of individuals High-frequency paternal psychological aggression (z-stat)b Unemployment -0.00 -0.02 -0.03 0.02 0.05 -0.01 -0.03*** -0.03 0.02 0.06 rate (model 1) (-0.14) (-1.09) (-1.18) (1.16) (1.54) (-1.03) (-3.40) (-1.28) (1.52) (1.60) Unemployment 0.00 -0.02 -0.02 0.02 0.06† -0.01 -0.03** -0.03 0.02 0.06† rate (model 2) (0.03) (-1.03) (-0.96) (1.07) (1.77) (-0.86) (-3.04) (-1.20) (1.28) (1.84) Increasing -0.00 0.00 -0.00 -0.00 0.00 0.00 0.00 -0.00 -0.00 0.00 unemployment (-0.90) (0.34) (-1.14) (-1.12) (0.20) (0.66) (1.35) (-0.01) (-0.90) (0.50) rate Decreasing 0.00 0.00 0.01 0.00 0.01 0.00 0.01† 0.01 -0.00 0.01† unemployment (1.43) (0.53) (1.57) (0.12) (1.37) (1.15) (1.86) (1.08) (-0.085) (1.87) rate Observations 4,919 1,736 1,333 1,158 692 4,919 1,736 1,333 1,158 692 Number 2,659 993 742 603 321 2,659 993 742 603 321 of individuals 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. aIncludes years one through nine. bIncludes years three through nine. ***p < .001; **p < .01; *p < .05; †p < .1 Table 7.A7 Sensitivity of Coefficients, Paternal Parenting Outcomes With Individual Without Individual Fixed Effects Fixed Effects Fathers High-frequency paternal spanking (z-stat)a Unemployment rate (model 1) -0.01 (-1.63) -0.01† (-1.75) Unemployment rate (model 3) -0.01 (-0.38) -0.01 (-1.50) Mother’s unemployment -0.02 (-0.52) -0.01 (-0.44) Bio-social fathers not employed 0.05 (1.47) 0.05* (-2.22) Unemployment rate (model 4) -0.03** (-2.88) -0.03** (-2.67) Unemployment rate * year nine 0.02* (2.44) 0.03* (2.28) High-frequency paternal physical aggression (z-stat)b Unemployment rate (model 1) -0.01† (-1.66) -0.02*** (-4.04) Unemployment rate (model 3) -0.03 (-1.12) -0.03* (-2.26) Mother’s unemployment -0.01 (-0.16) -0.01 (-0.52) Bio-social father’s not employed -0.02 (-0.38) 0.05* (2.16) Unemployment rate (model 4) -0.03** (-2.96) -0.04*** (-5.05) Unemployment rate * year nine 0.02* (2.52) 0.03*** (3.73) High-frequency paternal psychological aggression (z-stat)b Unemployment rate (model 1) -0.00 (-0.14) -0.01 (-1.03) Unemployment rate (model 3) -0.03 (-0.76) -0.04 (-1.15) Mother’s unemployment 0.01 (0.07) -0.00 (-0.05) Bio-social father’s not employed 0.18** (2.87) 0.07* (2.37) Unemployment rate (model 4) -0.03 (-0.76) -0.04 (-1.15) Unemployment rate * year nine 0.01 (0.07) -0.00 (-0.05) 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. aIncludes years one through nine. bIncludes years three through nine. ***p < .001; **p < .01; *p < .05; †p < .1 Table 7.A8 Coefficients and Standard Errors, Paternal Parenting Outcomes by Subgroups Black Hispanic White Married Cohabiting Single Fathers High-frequency paternal spanking (z-stat)a Unemployment rate -0.03* 0.01 -0.00 -0.01 -0.02† -0.02 (-2.19) (0.75) (-0.03) (-0.54) (-1.82) (-1.01) High-frequency paternal physical aggression (z-stat)b Unemployment rate -0.01 0.00 -0.01 -0.01 -0.01 — (-0.50) (0.04) (-0.99) (-1.28) (-0.77) (—) High-frequency paternal psychological aggression (z-stat)b Unemployment rate 0.01 0.00 0.01 0.02 -0.01 — (0.25) (0.15) (0.53) (1.12) (-0.55) (—) 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. aIncludes years one through nine. bIncludes years three through nine. *p < .05; †p < .1 204 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’ parenting 205

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

n this final chapter, we ask how a large change in the unemployment Irate, 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 well- being—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-being 207 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-year- old 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-being 209

Figure 8.1 Child Internalizing Behavior Problems

14 College + 12 10 Some college 8 High school Mean 6 Less than 4 high school 2 1 3 5 9 (1999–2001) (2001–2003) (2003–2006) (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

12 College + 10 Some college 8

Mean High school 6 Less than 4 high school 2 1 3 5 9 (1999–2001) (2001–2003) (2003–2006) (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

120

110 College +

100 Some college

90 High school Mean 80 Less than 70 high school

60 1 3 5 9 (1999–2001) (2001–2003) (2003–2006) (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 well- being 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-being 211

Figure 8.4 Child Overweight-Obese

0.50 0.45 College + 0.40 0.35 Some college 0.30 0.25 High school 0.20 Percent 0.15 Less than 0.10 high school 0.05 0 1 3 5 9 (1999–2001) (2001–2003) (2003–2006) (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 10 –16% UR 5 percent 8 –14%

Mean 0% UR 10 percent 6 0% 0% 4 2 All Less than High Some College + high school school 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.

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 0% 12 –9% –9% –10% 10 +11% UR 5 percent 8

Mean UR 10 percent 6 4 2 All Less than High Some College + high school school college

Figure 8.7 Child PPVT Scores, Unemployment Rates

120 110 +1% +1% 100 0% –2% –3% UR 5 percent 90 Mean 80 UR 10 percent 70 60 All Less than High Some College + high school school 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 vari- ability 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.)

Figure 8.8 Child Overweight-Obese, Unemployment Rates

50 45 40 35 UR 5 percent 30 25 +20% –50% –19% –85% +6% UR 10 percent

Percent 20 15 10 5 0 All Less than High Some College + high school school college

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 well- being, 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 well- being, rather than the unemployment rate. These results are summarized in table 8.A3. To estimate the effects of a deep recession, we use the fixed- effects 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-being 215 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 self- report 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 obesity- overweight. 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-being 217

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 inner- focused 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 outward- focused behaviors. The questions were drawn from the aggression and rule- breaking 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 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.

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-being 219

Table 8.A1 Full Regression Results, Child Well-Being Internalizinga With Individual Without Individual Fixed Effects Fixed Effects Unemployment rate -0.11 (-1.39) -0.09 (-0.88) Education Less than high school 1.79*** (7.00) High school -0.36 (-1.85) Some college -1.19 (-7.05) Relationship status Married -1.45 (-6.1) Cohabiting -0.3 (-1.9) Mother’s age -0.02 (-1.54) Race-ethnicity Black -0.17 (-1.15) Hispanic 0.55* (2.4) Other 0.75* (2.08) Immigrant 0.67 (2.41) Children in household 0.19*** (3.69) Lived with both parents -0.16 (-1.79) at age fifteen Interview year 2000 — (—) — (—) 2001 — (—) — (—) 2002 0.49 (0.65) 3.17*** (4.4) 2003 -0.30 (-0.43) 2.74*** (3.79) 2004 -2.78*** (-3.70) 0.06 (0.08) 2005 -2.97*** (-4.11) -0.03 (-0.04) 2006 -1.68 (-0.91) 2007 — (—) 0.49 (0.52) 2008 -3.14*** (-4.05) -0.45 (-0.6) 2009 -3.12*** (-4.55) -0.25 (-0.28) 2010 — (—) 1.22 (0.9) 220 children of the great recession

Table 8.A1 Continued Internalizinga With Individual Without Individual Fixed Effects Fixed Effects City Austin 1.16*** (8.66) Baltimore -0.15 (-0.46) Detroit 0.52 (1.33) Newark 1.08** (3.15) Philadelphia 0.73* (2.24) Richmond 0.59 (1.62) Corpus Christi 0.67 (1.58) Indianapolis 1.24** (2.86) Milwaukee 1.20** (2.73) New York 0.86* (2.02) San Jose 0.61 (1.31) Boston 0.32 (0.74) Nashville 0.54 (1.2) Chicago 0.28 (0.63) Jacksonville -0.08 (-0.17) Toledo 0.79† (1.73) San Antonio 2.15*** (4.9) Pittsburgh 1.48*** (3.39) Norfolk 0.35 (0.73) Constant 6.17*** (6.95) Observations 8,297 8,297 Number of individuals 3,861 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 Table 8.A2 Coefficients and Standard Errors, Rate of Change in Unemployment, Child Well-Being Outcomes With Individual Fixed Effects Without Individual Fixed Effects Less than Less than High High Some High High Some All School School College College + All School School College College + Internalizing-coefficients from OLS (z-stat)a,c Unemployment rate (model 1) -0.11 -0.18 -0.19 -0.02 0.02 -0.09 -0.09 -0.32** 0.06 -0.06 (-1.39) (-1.00) (-1.40) (-0.12) (0.10) (-0.88) (-0.45) (-2.72) (0.66) (-0.36) Unemployment rate (model 2) -0.10 -0.13 -0.20 -0.03 0.04 -0.09 -0.03 -0.32** 0.05 0.03 (-1.25) (-0.69) (1.48) (-0.19) (0.20) (-0.77) (-0.13) (-2.68) (0.58) (0.13) Increasing unemployment rate 0.01 0.01 0.01 0.01 -0.01 0.01 0.01 0.00 0.01 0.02 (1.00) (0.41) (0.65) (1.20) (-0.47) (1.02) (0.81) (0.32) (0.69) (1.01) Decreasing unemployment rate 0.01 0.07† -0.02 0.01 0.03 0.02 0.08* 0.00 -0.02 0.05 (0.84) (1.80) (-0.57) (-0.21) (0.78) (1.50) (2.45) (-0.12) (-1.37) (0.72) Observations 8,297 2,667 2,603 2,135 892 8,297 2,667 2,603 2,135 892 Number of individuals 3,861 1,295 1,205 960 401 4,487 1,542 1,368 1,090 487 Externalizing-coefficients from OLS (z-stat)a,c Unemployment rate (model 1) -0.08 0.05 -0.18 -0.17 0.11 -0.04 0.23† -0.47* 0.07 -0.07 (-0.80) (0.24) (-1.00) (-1.00) (0.42) (-0.45) (1.74) (-2.24) (0.52) (-0.32) Unemployment rate (model 2) -0.06 0.14 -0.21 -0.17 0.08 -0.02 0.36† -0.46** 0.06 -0.06 (-0.58) (0.65) (-1.15) (-1.01) (0.32) (-0.17) (1.90) (-2.58) (0.52) (-0.26) Increasing rate of unemployment 0.02*** 0.02† 0.03** 0.02 -0.01 0.02*** 0.04** 0.02* 0.01 0.01 (3.30) (1.67) (2.92) (1.57) (-0.32) (2.64) (2.68) (2.04) (0.57) (0.39) Decreasing rate of unemployment 0.03 0.08† 0.01 0.01 -0.01 0.02† 0.10** -0.01 -0.02 -0.01 (1.45) (1.77) (0.29) (0.25) (-0.26) (1.62) (2.91) (-0.28) (-0.54) (-0.10) Observations 8,320 2,643 2,634 2,150 893 8,320 2,643 2,634 2,150 893 Number of individuals 3,862 1,293 1,207 960 402 4,487 1,542 1,368 1,090 487 (Table continues on p. 222.) Table 8.A2 Continued With Individual Fixed Effects Without Individual Fixed Effects Less than Less than High High Some High High Some All School School College College + All School School College College + PPVT-coefficients from OLS (z-stat)a,c Unemployment rate (model 1) -0.00 -0.02 -0.03 0.04 0.01 -0.04* -0.05 -0.04 -0.01 -0.70 (-0.09) (-0.70) (-1.39) (1.59) (0.13) (-2.05) (-1.57) (-1.48) (-0.28) (-1.48) Unemployment rate (model 2) -0.01 -0.20 -0.03 0.33 -0.00 -0.04* -0.05† -0.04 -0.01 -0.07 (-0.39) (-0.90) (-1.47) (1.45) (-0.02) (-2.39) (-1.71) (-1.49) (-0.54) (-1.23) Increasing rate of unemployment -0.00 0.00 0.00 -0.00 0.00 0.00 0.00 0.00 -0.00 0.00 (-0.30) (0.16) (1.00) (-0.89) (0.21) (0.50) (1.02) (0.67) (-0.52) (0.93) Decreasing rate of unemployment -0.01** -0.01* -0.00 -0.01 -0.01 -0.01 -0.01 0.00 -0.01 -0.01 (-2.88) (-2.05) (-0.21) (-1.51) (-0.95) (-1.41) (-1.38) (-0.27) (-1.50) (-0.64) Observations 8,057 2,683 2,542 2,023 809 8,057 2,683 2,542 2,023 809 Number of individuals 3,893 1,327 1,216 951 399 4,487 1,542 1,368 1,090 487 Overweight/obese-coefficients from linear probability models (z-stat)b,c Unemployment rate (model 1) -0.00 0.01 -0.02 0.01 -0.03 -0.01 0.01† -0.03** 0.00 -0.03 (-0.37) (0.79) (-1.49) (0.69) (-1.37) (-1.35) (1.71) (-2.71) (0.01) (-1.50) Unemployment rate (model 2) -0.00 0.01 -0.02 0.01 -0.03 -0.01 0.02† -0.03** 0.00 -0.02 (-0.28) (0.98) (-1.51) (0.68) (-1.33) (-1.32) (1.93) (-2.69) (0.03) (-1.34) Increasing rate of unemployment 0.00 0.00 0.00 -0.00 -0.00 -0.00 0.00 -0.00 -0.00 0.00 (0.15) (0.35) (0.20) (-0.32) (-0.36) (-0.50) (0.64) (-1.34) (-0.41) (0.28) Decreasing rate of unemployment 0.00 0.00 -0.00 -0.00 0.00 0.00 0.00† -0.00 0.00 0.00 (0.71) (1.37) (-0.27) (-0.18) (0.20) (1.37) (1.90) (-0.08) (0.18) (0.70) Observations 7,698 2,643 2,396 1,916 743 7,698 2,643 2,396 1,916 743 Number of individuals 3,740 1,291 1,169 916 364 4,487 1,542 1,368 1,090 487 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 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 child well-being 223

Table 8.A3 Sensitivity of Coefficients, Child Well-Being Outcomes With Individual Without Individual Fixed Effects Fixed Effects Internalizing coefficients from OLS (z-stat)a,c Unemployment rate (model 1) -0.11 (-1.39) -0.09 (-0.88) Unemployment rate (model 3) 0.00 (0.00) -0.18 (-0.97) Mother’s unemployment -0.08 (0.25) 0.60** (2.83) Bio-social fathers not employed -0.12 (-0.39) 0.18 (0.91) Unemployment rate (model 4) 0.06 (0.50) -0.03 (-0.19) Unemployment rate * year nine -0.22* (-2.04) -0.09 (-1.10) Externalizing coefficients from OLS (z-stat)a,c Unemployment rate (model 1) -0.08 (-0.80) -0.04 (-0.45) Unemployment rate (model 3) -0.12 (-0.28) -0.31 (-0.75) Mother’s unemployment 0.05* (2.05) 0.91† (1.89) Bio-social father’s not employed 0.08 (1.50) 0.33 (0.86) Unemployment rate (model 4) 0.11 (0.78) 0.14 (0.90) Unemployment rate * year nine -0.24† (-1.84) -0.23* (-1.97) PPVT coefficients from OLS (z-stat)a,c Unemployment rate (model 1) -0.00 (-0.09) -0.04* (-2.05) Unemployment rate (model 3) -0.06 (-1.17) -0.03 (-0.66) Mother’s unemployment -0.01 (-0.14) -0.20*** (-4.80) Bio-social fathers not employed 0.07 (1.10) -0.07* (-2.26) Unemployment rate (model 4) -0.01 (-0.25) -0.05† (-1.75) Unemployment rate * year 9 0.00 (0.26) 0.02 (0.87) Overweight-obese coefficients from linear probability models (z-stat)b,c Unemployment rate (model 1) -0.00 (-0.37) -0.01 (-1.35) Unemployment rate (model 3) -0.00 (-0.13) -0.00 (-0.15) Mother’s unemployment -0.01 (-0.35) 0.10** (2.79) Bio-social fathers not employed 0.02 (1.23) 0.04 (1.29) Unemployment rate (model 4) 0.01 (0.44) 0.00 (0.15) Unemployment rate * year nine -0.01 (-0.95) -0.01 (-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. bCoefficients from linear probability models. c Years three through nine only. ***p < .001; **p < .01; *p < .05; †p < .1 Table 8.A4 Coefficients and Standard Errors, Child Well-Being Outcomes by Subgroup Black Hispanic White Married Cohabiting Single Internalizing coefficients from OLS (z-stat)a,c Unemployment rate 0.06 -0.18 -0.30* -0.05 -0.06 -0.13 (0.41) (-1.02) (-2.13) (-0.31) (-0.44) (-0.90) Externalizing coefficients from OLS (z-stat)a,c Unemployment rate -0.28 0.05 -0.22 -0.16 -0.15 0.09 (-1.61) (0.22) (-1.20) (-0.88) (-0.90) (0.49) PPVT coefficients from OLS (z-stat)a,c Unemployment rate -0.01 -0.02 0.08** -0.01 -0.02 0.01 (-0.70) (-0.80) (2.96) (-0.38) (-1.17) (0.70) Overweight-obese coefficients from linear probability models (z-stat)b,c Unemployment rate 0.01 -0.02 -0.02 -0.01 0.01 -0.01 (0.42) (-1.17) (-1.27) (-0.87) (0.74) (-0.73) 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 child well-being 225

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 School- Age 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-being 227

“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 Mixed- Methods 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 regression results, 219–20; well-being American Recovery and Reinvestment outcomes, sensitivity of coefficients, Act of 2009 (ARRA), 60–61 223; well-being outcomes and rate of Anderson, Elijah, 150 change in unemployment, co-efficients and standard errors, 221–22; well- Bitler, Marianne, 61 being outcomes by subgroup, co- Brooks-Gunn, Jeanne, 18–20 efficients and standard errors, 224 child support, 150–52. See also nonresident Canada, 25 father involvement Charles, Kerwin, 103 Conflict Tactics Scale, 175 children: born to unwed mothers, 3; mal- Conger, Rand D., 2–3, 21, 174 treatment of, 175; parenting of (see consumer confidence index, 47 parenting of mothers and fathers) Current Population Survey (CPS): Child children’s well-being, 19–20, 206–7, Support Supplement, 152; sample 215–17; additional estimates, size, 23 214–15; economic uncertainty and, Currie, Janet, 16, 90, 103 214; externalizing behavior problems by child age-year and educa- data: demographic changes, accounting tion, 209; externalizing behavior for, 3–4; Fragile Families and Child problems by unemployment rate and Wellbeing Study as basic, 3 (see also education, 212; internalizing behav- Fragile Families and Child Wellbeing ior problems by child age-year and Study); limitations of, 23–24; proce- education, 209; internalizing behav- dure for looking at, 9–11 ior problems by unemployment rate DeCicca, Philip, 103 and education, 212; local unemploy- disadvantaged families: data on, 7–8; dis- ment rates and, 207, 210–13; mea- advantage, measure of, 7–8; economic sures of, 217; method for the study well-being, impact of recessions on, of, 218; overweight-obese by child 45; fragile families sample composition age-year and education, 211; over- by mother’s education, 8; policy impli- weight-obese by unemployment rate cations of findings for, 26. See also and education, 213; PPVT scores parents’ relationship status and quality by child age-year and education, divorce rates, economic conditions and, 210; PPVT scores by unemployment 119–20 rate and education, 213; previous dot-com recession, 2, 5–6 research on the effects of economic double-dip recession, 5 hardship and the Great Recession, doubling up, 15, 59–60, 64–68, 73 207–8; trends in internalizing and Duque, Valentina, 16, 90 externalizing behaviors from ages three to nine, 208–10; variables for Earned Income Tax Credit (EITC), 4, the study of, 217–18; well-being, full 14–15, 32, 46, 58–64, 67–69, 73 index 229 economic insecurity: effects of the Great resident father involvement and, 18, Recession on, 44, 44–45; of families 153–58; parenting of mothers and with children from birth through age fathers and, 18, 176–82, 185–86; nine, 38, 40; hardship rates of fami- parents’ relationships and, 17, lies by child age-year and mother’s 121–26, 128–34; policy implications education, 39; material hardship, full of findings for, 25–26; poverty and, regression results, 48–49; as a mea- 38–39, 42–44, 44; public and private sure of economic well-being, 32–33, transfers and, 15, 63–72 46; public and private transfers, EITC. See Earned Income Tax Credit impact of, 72 Elder, Glen H., Jr., 1–3, 21, 174 economic well-being, 13–14; economic employment: by education, 42; as a mea- outcomes, coefficients and standard sure of economic well-being, 31–32, errors, 54; economic outcomes, sen- 46; for mothers and fathers with chil- sitivity of coefficients, 53; economic dren from birth through age nine, outcomes rate of change, coeffi- 35–36. See also local unemployment cients and standard errors, 50–52; rates; unemployment and joblessness employment by education and local ethnicity. See race and ethnicity unemployment rate, 42; of families with children from birth through age families: data on, richness of, 8; dis nine, 35–41; Great Recession, impact advantaged (see disadvantaged of, 33, 41–45; hardship by educa- families); economic being of, tion and child age-year, 39; hardship with children from birth through by education and local unemploy- age nine, 35–41; effects of Great ment rate, 44; household income, Recession on, 20–22; instability in, big gains and losses by education 3; involvement in childrens’ lives of and child age-year in, 38; household nonresident fathers (see nonresident income by education and child age- father involvement); parenting (see year, 37; household income by race- parenting of mothers and fathers); ethnicity, parents’ relationship status, parents’ relationship (see parents’ and child age-year, 40; income by relationship status and quality) education and local unemployment family stress model: future research on, 24; rate, 43; income by race-ethnicity, implications of findings for, 22–23; as parents’ relationship status, and legacy of Elder study, 2; marital qual- local unemployment rate, 43; ity and economic crises, relationship material hardship, full regression of, 120; natural experiments, testing results, 48–49; maternal employ- with, 2–3; parenting and economic ment by education and child age- hardship, relationship of, 207 year, 35; measures of, 31–33, 46; Farm Crisis Study, 2 paternal employment by education fathers: binge drinking, by education and and child age-year, 36; poverty rate child age-year, 94; binge drinking, by education and child age-year, effects of a recession on, 100; drug 39; poverty rate by education and use, by education and child age-year, local unemployment rate, 44; prior 95; drug use, effects of a recession research on recessions and, 34–35; on, 101; employment, simulated supplemental analyses of, 47 effects of the Great Recession on, education: children’s well-being and, 41–42; employment of by educa- 209–13; economic insecurity and, tion level, 36; health outcomes and 38–40, 44, 44–45; economic well- behavior, effects of uncertainty on, being and, 13–14, 33; effects of 101; health outcomes and behavior Great Recession and, 20–21, 45; by child age-year, 90–96; health employment and, 35–36, 42; health outcomes and behavior during reces- and, 16, 90–103; income and, sions, 96–100; health outcomes and 36–37, 42, 43, 58; as a measure behavior during recessions, con- of well-being, 10; of mothers as clusions regarding, 102–4; health measure of disadvantage, 7–8; non- outcomes and behavior of, 16, 230 children of the great recession fathers (continued) (see local unemployment rate); health 88–89; health outcomes by educa- impact during, 102; natural experi- tion, coefficients and standard errors, ments based on, 2; public and private 110–12; health problems that limit transfers, effect on, 67–70; public work, by education and child age- and private transfers during the, year, 93; health problems that limit impact of, 70–72; severity of, 4–6; work, effects of a recession on, 98; unemployment, income, and length health status, effects of a recession of, 5–6 on, 97; health status is fair or poor, by education and child age-year, Harknett, Kristen, 17 92; involvement in childrens’ lives Hatchett, Shirley, 120 of nonresident (see nonresident health: behaviors, patterns of for parents father involvement); mothers’ sup- by child age-year, 93–96; behaviors, portiveness by education and local patterns of for parents with increase unemployment rate, 130; parent- in unemployment rate, 99–100; ing of, 19, 178–80, 185–86 (see also children’s, 208, 210, 211, 213, 216; parenting of mothers and fathers); fathers’ binge drinking by educa- positive sample selection, impact tion and child age-year, 94; fathers’ of, 146n16; relationship with bio binge drinking by education and mother by education and local local unemployment rate, 100; of unemployment rates, 132; reports fathers by education and child age- of bio mothers’ supportiveness year, 92; of fathers by education and by education and child age-year, local unemployment rate, 97; fathers’ 124; reports of relationship with drug use by education and child bio mother by education and child age-year, 95; fathers’ drug use by age-year, 126. See also parenting of education and local unemployment mothers and fathers; parents’ rela- rate, 101; during Great Recession tionship status and quality years, 102; measures of, 104–5; FFS. See Fragile Families and Child method used in study of, 105, Wellbeing Study 115; of mothers, impacts of the fixed-effects models, 11–12 2007 recession on, 90; mothers’ food insecurity, 72 binge drinking by education and food stamps. See Supplemental Nutrition child age-year, 94; mothers’ binge Assistance Program (SNAP) drinking by education and local foreclosure rate, 47 unemployment rate, 99; of mothers fragile families. See disadvantaged families by education and child age-year, 91; Fragile Families and Child Wellbeing of mothers by education and local Study (FFS): as basis for analyses, 3; unemployment rate, 97; mothers’ data from, 6–9; health, as source on, drug use by education and child 90, 104; limitations of, 23–24; age-year, 95; mothers’ drug use by material hardship, questions measur- education and local unemployment ing, 32–33; nonresident fathers, as rate, 100; outcomes, coefficients and source on, 153, 160; parenting of standard errors, 114; outcomes by mothers and fathers, as source on, maternal education, coefficients and 174; public and private financial trans- standard errors, 108–9; outcomes by fers, as source on, 61; relationship sta- paternal education, coefficients and tus and quality, as source on, 120, 133 standard errors, 110–12; of parents, future research, 24 16, 88–89; of parents, sensitivity of coefficients, 113; of parents dur- Garfinkel, Irwin, 13–15, 90, 103, 152 ing recessions, conclusions regard- Great Depression Study, 1–2 ing, 102–4; physical, full regression Great Recession: economic well-being results, 106–7; physical, patterns of and, 34–35, 41–45 (see also eco- for parents by child age-year, 91–93; nomic well-being); effects on fami- physical, patterns of for parents with lies, 20–22; estimating the effect of increase in unemployment rate, index 231

96–99; problems of fathers that 155–60; parenting of mothers and limit work by education and local fathers and, 19, 180–88; parents’ unemployment rate, 98; problems relationship and, 17, 126–33; public of mothers that limit work by educa- and private transfers and, 15, 67–70, tion and local unemployment rate, 74–75; role in fixed-effects models 98; problems that limit fathers’ work of, 12–13. See also unemployment by education and child age-year, 93; and joblessness problems that limit mothers’ work by education and child age-year, 92; marriage, economic conditions and rates race-ethnicity, differential impacts of, 119 due to, 102; recessions and, prior material hardship. See economic insecurity research on, 89–90; relationship sta- McLanahan, Sara, 17 tus, differential impacts due to, 102; Medicaid, 4, 14–15, 58–64, 67–69, 73 uncertainty, effects of, 100–101; method/methodology: children’s well- unemployment of the individual and, being, used in study of, 218; fixed- 101–2; variables in analyses, 105 effects models, 11–13; health of hedonic adaptation theory, 12 mothers and fathers, used in study Hispanics. See race and ethnicity of, 105, 115; natural experiments to housing: assistance, cash and in-kind, 58; address omitted variable bias, 2–3, assistance, federally funded, 59–60, 11, 22; nonresident father involve- 62–64, 73; doubling up, 15, 59–60, ment, used in study of, 162–63; par- 64–68, 73 enting of mothers and fathers, used Hoynes, Hilary, 61 in study of, 190; parents’ relationship status and quality, used in study of, income: big gains and losses by families 135–36 with children by child age-year, 38; Mills, Bradford, 72 education level and, 36–37, 42, Mincy, Ronald, 17–18 43, 58; by education level and local mothers: binge drinking, effects of a reces- unemployment rate, 43; effects sion on, 99; binge drinking by educa- of public and private transfers on tion and child age-year, 94; drug use, household, 71; of families by child effects of a recession on, 100; drug age-year, race-ethnicity,and parents’ use by education and child age-year, relationship status, 40; of families 95; educational attainment as mea- with children from birth through age sure of disadvantage, 7–8; education nine, 36–38; family, effects of the of, policy implications of findings for, Great Recession on, 42; household 25–26; effects of Great Recession on as a measure of economic well-being, relationship status, 126–29; employ- 32, 46; household by education and ment, simulated effects of the Great child age-year, 37; median household Recession on, 41–42; employment income index, 2000–2014, 6; the by education level, 35; fathers’ sup- recessions of the twenty-first century portiveness by education and local and, 5–6, 34–35 unemployment rate, 130; health, impacts of the 2007 recession on, Kuka, Elira, 61 90; health outcomes and behavior, effects of uncertainty on, 101; health local unemployment rates: children’s well- outcomes and behavior during reces- being and, 19–20, 207, 210–13; city sions, 96–100; health outcomes and of birth vs. current residence, 27n18; behavior during time with children economic well-being and, 13–14, from birth through age nine, 90–96; 41–42, 47; FFS data and, combin- health outcomes and behavior of, 16, ing, 8–9; health outcomes/behaviors 88–89; health outcomes and behav- and, 16, 96–100; individual-level ior of during recessions, conclusions measures and aggregate rates, 27n23; regarding, 102–4; health outcomes during interviewing periods, 9; non- by education, coefficients and stan- resident father involvement and, 18, dard errors, 108–9; health problems 232 children of the great recession mothers (continued) 151–52; father engagement by edu- that limit work, effects of a recession cation and child age-year, 154; father on, 98; health problems that limit involvement rate of change, coef- work by education and child age- ficients and standard errors, 166–67; year, 92; health status, effects of a financial support and visitation, recession on, 97; health status is fair impact of unemployment on, 150; or poor by education and child age- formal child support per year by edu- year, 91; labor force participation of, cation, 156; informal child support 3; marriage and marriage or cohabi- per year by education, 157; in-kind tation by local unemployment rate, child support per year by education, 127; marriage bio father or new 157; measures of child support out- partner by education and local comes, 160–61; measures of visitation unemployment rate, 128; marriage outcomes, 161; method for the study or cohabitation by education and of, 162–63; mother’s unemploy- local unemployment rate, 128; mar- ment and, 159; nonresidence status riage or cohabitation to bio fathers by education and child’s age-year, or new partners by education and 153; share of nonresident fathers child age-year, 122; marriage to bio visiting their children by education, fathers or new partners by education 158; stress of economic adversity and child age-year, 122; married and, 157–58; trends in child-support to or cohabiting with father or new orders, payments, and visitation dur- partner, full regression results, 137– ing period from children’s birth to 38; new partners’ supportiveness by age nine, 152–55; variables for the education and local unemployment study of, 161–62; visitation and rate, 131; parenting of, 18–19, child support, reciprocal relationship 176–78, 181–85 (see also parenting between, 169n11; visitation days per of mothers and fathers); relationship month by education, 158 status by child age-year, 121; relationship with bio father by education parenting of mothers and fathers, 18–19, and local unemployment rates, 132; 173–74, 188–89; economic hard- reports of fathers’ supportiveness by ship and, impact on children of, 207; education and child age-year, 123; maternal parenting, full regression reports of new partners’ supportive- results, 191–92; maternal parent- ness by education and child age-year, ing activities by education and child 125; reports of relationship with age-year, 179; maternal parenting bio father by education and child activities by unemployment rate and age-year, 125; sample composition, education, 184; maternal parenting education and, 8. See also parents’ outcomes, sensitivity of coefficients, relationship status and quality 196; maternal parenting outcomes Mykerezi, Elton, 72 and rate of change in unemployment, coefficients and standard natural experiments, 2–3, 11 errors, 193–95; maternal parenting Nepomnyaschy, Lenna, 152 outcomes by subgroup, coefficients nonresident father involvement, 17–18, and standard errors, 197; maternal 149, 159–60; child support and physical aggression by education and visitation, coefficients and standard child age-year, 177; maternal physi- errors, 169; child support and visita- cal aggression by unemployment tion, effects of the Great Recession rate and education, 182; maternal on, 155–59; child support and psychological aggression by educa- visitation, full regression results, tion and child age-year, 177; mater- 164–65; child support and visitation, nal psychological aggression by sensitivity of coefficients, 168; child unemployment rate and education, support and visitation by education, 183; maternal spanking by education 155; empirical evidence regarding, and child age-year, 176; maternal index 233

spanking by unemployment rate and Recession and, 42; income loss of education, 182; maternal warmth by mothers by local unemployment rate, education and child age-year, 178; 43; married to or cohabiting with maternal warmth by unemployment father or new partner, full regression rate and education, 183; measures results, 137–38; measures of rela- of, 189; method used in study of, tionship quality, 134–35; measures of 190; paternal parenting, full regres- relationship status, 134; method for sion results, 198–99; paternal study of, 135–36; mothers’ marriage parenting outcomes, sensitivity of and marriage or cohabitation by local coefficients, 202; paternal parent- unemployment rate, 127; mothers’ ing outcomes and rate of change marriage bio father or new partner in unemployment, coefficients and by education and local unemploy- standard errors, 200–201; paternal ment rate, 128; mothers’ marriage or parenting outcomes by subgroup, cohabitation by education and local coefficients and standard errors, unemployment rate, 128; moth- 203; paternal physical aggression ers marriage or cohabitation to bio by education and child age-year, fathers or new partners by education 180; paternal physical aggression by and child age-year, 122; mothers unemployment rate and education, marriage to bio fathers or new part- 186; paternal psychological aggres- ners by education and child age-year, sion by education and child age-year, 122; mothers relationship status 181; paternal psychological aggres- by child age-year, 121; mothers’ sion by unemployment rate and reports of fathers’ supportiveness by education, 186; paternal spanking by education and child age-year, 123; education and child age-year, 179; mothers’ reports of fathers’ sup- paternal spanking by unemployment portiveness by education and local rate and education, 185; previous unemployment rate, 130; mothers’ research on, 174–75; trends in with reports of new partners’ supportive- children ages three to nine, 175–80; ness by education and child age-year, unemployment rates, impact of with 125; mothers’ reports of new part- children ages three to nine, 180–88; ners’ supportiveness by education unemployment rates and, rapidly and local unemployment rate, 131; changing, 184–85; variables in study mothers’ reports of relationship of, 189–90 with bio father by education parents’ relationship status and quality, and child age-year, 125; mothers’ 17, 118–19, 133–34; differential reports of relationship with bio effects of recessions by, 45; effects father by education and local of Great Recession on relationship unemployment rates, 132; reces- quality, 129–33; effects of Great sions and romantic relationships, Recession on relationship status, 119–20; relationship outcomes, 126–29; fathers’ reports of bio coefficients and standard errors for mothers’ supportiveness by educa- unemployment rate, 139–42, 145; tion and child age-year, 124; fathers’ relationship outcomes, sensitivity reports of mothers’ supportiveness of unemployment rate coefficients, by education and local unemploy- 143–44; supplemental analyses of, ment rate, 130; fathers’ reports of 136; trends during period from chil- relationship with bio mother by dren’s birth to age nine, 120–26; education and child age-year, 126; variables for study of, 135 fathers’ reports of relationship with Patterson, Richard, 120 bio mother by education and local Peabody Picture Vocabulary Test (PPVT), unemployment rates, 132; health 207–8, 210, 212–15 effects of, 102; household income Piketty, Thomas, 5 by child age-year and, 40; income Pilkauskas, Natasha, 13–15, 103 loss by mothers during the Great policy implications, 25–26 234 children of the great recession

poverty: children, effect on, 207; of effects of recessions by, 45; health families with children from birth effects differentiated by, 102; house- through age nine, 38–39; the hold income by child age-year and, Great Recession and, 34–35, 40; income loss during the Great 42–44; as a measure of economic Recession of mothers by, 42; income well-being, 32, 46; mitigating loss of mothers by local unemploy- effects of public and private trans- ment rate, 43; parenting practices fers on, 71; official threshold for, and, 183–84, 187–89; policy impli- 32; rate by education and local cations of findings for, 26 unemployment rate, 44; rates of recession(s): definition of, 4; double-dip, families with children from birth unemployment during, 5; economic through age nine by mother’s well- being and, 34–35 (see also eco- education, 39 nomic well-being); health and (see PPVT. See Peabody Picture Vocabulary health); romantic relationships and, Test 119–20 (see also parents’ relationship private cash/financial transfers, 15, status and quality). See also dot-com 59–60, 64–68, 72–73 recession; Great Recession provider role strain, 150 public and private transfers, 14–15, Saez, Emmanuel, 5 58–60; coefficients and standard Schneider, Daniel, 17 errors, 84; doubling up by educa- Schneider, William, 18–20 tion and child’s age-year, 65; effects Seltzer, Judith, 152 of the Great Recession on, 67–70; SNAP. See Supplemental Nutrition effects of transfers on household Assistance Program income by education, 71; expanded SSI. See Supplemental Security Income by the American Recovery and State Children’s Health Insurance Reinvestment Act of 2009, 60; Program, 4 helping effects of, 70–72; mea- Supplemental Nutrition Assistance sures of, 73–74; mitigating effects Program (SNAP), 4, 14–15, 32, 46, of transfers on poverty by educa- 58–64, 67–69, 73, 75–76 tion, 71; previous research on the Supplemental Security Income (SSI), Great Recession and, 61; private 14–15, 58–60, 62–64, 68, 73 assistance, average dollar value of by education, 66; private financial TANF. See Temporary Assistance for transfers and doubling up by local Needy Families unemployment rate and education, Temporary Assistance for Needy Families 68; private financial transfers by (TANF), 14–15, 58–60, 62–64, education and child age-year, 65; 67–69, 73 public assistance benefits, average Toledo, Elia De la Cruz, 17–18 dollar value of, 64; public assistance receipt by child age-year, 62; pub- unemployment and joblessness: child lic assistance receipt by education, support compliance and, 151–52; 63; public transfer receipt rates by the dot-com recession and, 5; effects local unemployment rate and educa- of during the Great Recession, tion, 69; rate of change for, coef- 21–22; fathers’ financial support for ficients and standard errors, 77–81; and visitation of children, impact received during child’s age one to on, 150; future research on, 24; the nine, 61–67; sensitivity of coeffi- Great Recession and, 4–6, 34–35; cients, 82–83; SNAP, full regression health effects of, 101–2; informal results for, 75–76; supplemental child support compliance and, 152; analyses, 74–75 local unemployment rates (see local unemployment rates); nonresident race and ethnicity: children’s well-being fathers’ visitation of children and, and unemployment rates, impact on 152; unemployment rate, 2000– the relationship of, 215; differential 2014, 6

14371-09-Index-5thPgs.indd 234 8/5/16 12:17 PM index 235

Unemployment Insurance (UI), 14–15, net transfers, 14. See also public 58–64, 73 and private transfers women: joblessness during the Great Veum, Jonathan, 152 Recession, 4–5; labor force participation of, 3; as mothers (see mothers). Waldfogel, Jane, 18–20 See also parenting of mothers and welfare state programs: policy implica- fathers; parents’ relationship status tions of findings for, 25; safety and quality net, better-developed, 4; safety- Wu, Chi-Fang, 151