HQ 759.48 .R46 2000 IGSL UCB M BARGOED E UNTIL Tl IE MORNING OF O 9 F3 *5 > o: < D o: :0 LL] LL >1 < E a: LL ,, i, emember the Children MOthEI‘S Balance work and Child Care under Welfare Reform Growing Up In Poverty Project 0 wave 1 Findings—California. Connecticut, Florida University of California, Berkeley and Yale University HOW are children and mothers faring under welfare reform? In the summer of 1998 we began to invite single mothers with preschool-age children— all entering new welfare programs—into a four-year study. Our premiere aim is to learn how the upbringing and development of children may be altered by the unprecedented push on their mothers to leave home, find child care, and hold down a job. This monograph details initial findings that stem from the Wave 1 data. The results are largely descriptive, offering a baseline picture for 948 families spread across three states. Some findings offer early warning signals. For example, we find that many youngsters are being placed in low-quality child care settings. We find high levels of social isolation and clinical depression among many women, sure indicators that their children’s early development will be delayed Welfare reform has brought results. But it's a work in progress. If cash assistance and family support programs are to lift this new generation of children out of poverty, much remains to be done. This report delineates key items for this unfinished agenda. The Growing Up in Poverty Project is conducted in collaboration with our research partners, Mathematica Policy Research Inc. and the Manpower Demonstration Research Corporation. The Project receives support from nine foundations and government research agencies. INSTITUTE OF GOVERNMENTAL STUDIES LIBRARY FFB 07 20l7 UNIVERSITY OF CALIFORNIA The Growing Up in Poverty Project Graduate School of Education-PAC E University ofCalifornia. Berkeley 94720 510-642—7223 Bush Center in Child Development and Social Policy Yale University. New Haven 0651 1 203-432—9931 Additional copies may be ordered from the Berkeley office. Wave I Report February 2000 Remember the Children Mothers Balance Work and Child Care under Welfare Reform Growing Up in Poverty Project 2000 Wave 1 Findings—California, Connecticut. Florida Project Co-directors Bruce Fuller University of California, Berkeley Sharon Lynn Kagan Yale University Berkeley Gretchen Caspary, Assistant Project Director Nancy Cohen Desiree French Laura Cascue Africa Hands James Mensing Tampa Jan McCarthy, Site Coordinator Cege Kreischer Yale Jude Carroll, Site Coordinator Kristen Cool Survey Coordinators Susan Sorachman, Mathematica Policy Research Inc, Princeton Greg Hoerz and Jordan Kolovson, MDPC, New York “ TABLE OF CONTENTS Acknowledgments 1 Executive Summary 3 Section 1 PROJECT AIMS 7 How are children and mothers faring under welfare reform? Section 2 MOTHERS' ATTRIBUTES 21 Who are the women entering new welfare programs? Section 3 STATES AND NEIGHBORHOODS 27 How do job markets, welfare rules, and child care contexts differ? Section 4 HOMES AND PARENTING 41 How well do families function? Section 5 FAMILY ECONOMY 51 How do women get by economically? Section 6 MATERNAL AND CHILD HEALTH 59 What gaps exist in access to insurance and clinical services? Section 7 WELFARE TO WORK 65 How do women engage new welfare rules and jobs? Section 8 CHILD CARE 71 Where do young children now spend their days? Section 9 EARLY LEARNING 91 How well are children growing and developing? Endnotes 101 Appendices 107 BERKELEY — YALE GROWING UP IN POVERTY PROJECT A ACKNOWLEDGMENTS ACKNOWLEDGMENTS We are blessed by the wonderful range of individuals who have come together to craft the Growing Up in Poverty Project. This initial report would not be in your hands today if not for this team of highly committed funders, policy makers, and researchers. In light of the contention swirling around welfare reform, we are motivated by the desire to discover whether this policy revolution is advancing the well—being of children. Our policy partners—in Washington, state capitals, and participating counties—have displayed an inspiring willingness to explore this empirical frontier. First and foremost we thank the 948 women who have been willing to share many elements of their lives. They have let us into their homes and into their child care settings. In turn, many have found their way into our hearts. Appreciation also is expressed to the hundreds of child care providers— from kin members to preschool teachers—who allowed us into their settings and put up with all our questions. The Project‘s funders have been enormously sup— portive since 1996 when the Packard Foundation provided the first planning grant. Deanna Gomby and Marie Young at Packard have long supported our work. We are forever grateful. Patricia Graham and the Spencer Foundation board were early and generous supporters, seeing the linkages between family—support policies and children’s early school performance, as was Mike Laracy at the Casey Foundation. The early education institute of the US. Department of Education (OERI) provided a generous grant. Warm appreciation is expressed to Naomi Karp and Donna Hinkle. The Child Care Bureau of the US. Department of Health and Human Services took an early and sustained interest in the Project. Many thanks to Joan Lombardi and Pia Divine. Smaller grants have been crucial, especially in funding dissemination activities. Thanks go to the Peter and Miriam Haas and the Walter and Elise Haas funds, and to the MacArthur, Mailman, and Hancock foundations, all supporting our efforts to engage policy makers and program designers around the country. The California Department of Social Services is supporting new lines of data analysis and focus groups to add a more human understanding of women’s complicated lives, moving beyond all the numbers. The Hewlett Foundation provides bed— rock support to Berkeley’s PACE research center. A countless number of reviewers have constructively pushed us to improve the research program. Our survey directors have been wonderfully skilled and equally fun partners to work with. Susan Sprachman at Mathematica has been with us since the beginning, helping to design interview and assessment instruments, organizing field operations, and most importantly, keeping track of the partici— pating families. The steady assistance and humor of Audrey McDonald and Phyllis Schulman have been a gift. Greg Hoerz and Jordan Kolovson at MDRC, and Lee Robeson at Response Analysis Corporation, played similar roles for the Connecticut survey. Dan Bloom and Bob Granger continue to be supportive colleagues. In each participating state, we have been buoyed by the openness and curiosity of welfare and child care officials. In California, Raul Aldana, Jeanne Flory, Julie Goldsmith, Delores Heaven, Will Lightborne, Alette Lundeberg, Trent Rhorer, Michele Ruther— ford, Jolene Smith, and Kate Welty have been simply wonderful. In Sacramento, Dave Dodds, Teri Ellen, Bill Jordan, Nancy Remley, Oshi Ruelas, Werner Schink, and Jo Weber have provided moral and financial support. Larry Orman and Mark Breshears at Greenlnfo Network led the geocoding and spatial analysis that appear in Section 3. In Connecticut, Mark Heuschkel, Kevin Loveland, Peter Palermino, Joyce Thomas, Patricia Wilson— Coker, Marion Wojick have been very supportive, BERKELEY - YALE [I ACKNOWLEDGMENTS despite the Projects added headaches and the sensitivity of findings. Their commitment to explor— ing the effects of welfare reform on families and children has been a joy to discover. In Florida, several agencies and individuals have provided invaluable assistance and enthusiasm. We are especially grateful to Rene Benton, state Depart- ment of Labor, and Tim Kelley, Department of Children and Families, for facilitating the selection of families. We arrived on the scene during the early months of welfare reform. Rene and Tim were open and trusting during this initial period. In planning and carrying out the study in 'Iampa, several others were equally cooperative: the managers of the welfare orientation and job club sites; Ann Dawson and Kay Doughty who co-chair our local advisory committee; Janet Allyn and Betsy Drake at Partners in Care, a resource and referral agency; Janet Aversa at the 4Cs Early Learning Office of the Hillsborough County school district; Marina Harkness at the Department ofChildren and Families who patiently taught us about the local welfare system; and Linda Stoller at the child care licensing agency. We are grateful to the Childrens Board oinllsborough County for housing our field office, and to Children and Families for donation of office space. Special thanks go to Susan Muenchow and Lisa Odom for their moral and administrative support in Tallahassee. The Projects Technical Advisory Group has always been available to offer guidance and encouragement. Deep appreciation is expressed to David Blau, Ellen Galinsky, Fred Glantz, Suzanne Randolph, Catherine Snow, Jeanne Brooks—Gunn, Carollee Howes, Judith Singer, and Marty Zaslow. We warmly thank our field staffwho did the hard work of inviting women into the study, gently probing into their lives, setting up child care and home visits. A heart felt thanks to Al Barlow, Therese Baumberger, Shannon Borch, Maria Victoria Chavez, Maria Bonita Chavez, Kristen Cool, Grace Dowdy, Jinel DuPrey, Susannah Eldridge, Jody Embrey, Marcia Garman, Christiane fl GROWING UP IN POVERTY PROJECT 2 Gauthier, Nicole Gardner, Cindy Golden, Tara Huis, Dorene Lisowsky, Chun Lee, Kym Lessard, Sonya Mason, Ouyen Ngo, Oanh Nguyen, Tuonganh Nguyen, Elsa Jones Nance, Sonya Patton, Valerie Pervis Pritchett, Judy Peterson, Myra Quitiquit, Denise Rao Monoghan, Beverly Richardson, Ruby Richardson, Peter Rogers, Barbara Rossi, Jean Rustici, Sheri Sweet, Ina Silverman, Lisa Smith Horn, Melinda Stephens, Denise Villanueva, Lam V0, and Da V0. The site coordinators and senior staff appear on the title page. Jan McCarthy and Gege Kreischer direct Tampa operations. Jude Carroll oversees daily work at Yale. Gretchen Caspary and Christiane Gauthier codirect field work in California. Gretchen and Laura Gascue also conducted the data analysis that informs this report. Africa Hands earlier coordinated Berkeley operations. Jim Mensing coordinated San Jose operations during Wave 1. Substudies within the Project are directed or assisted by Jude Carroll, Desiree French, Susan Holloway, Jim Mensing, and Lynna Tsou. Darren Lubotsky and Judith Singer have offered invaluable statistical and modeling advice. Elaine Chen and Lynna Tsou labored over data entry. Bob Hass contributed heavily to this report, taking his typically sharp pencil to our prose. Joanne Klein led the artistic design and crafted the finished product. Thanks go to Nita Winter for the fine photos. Finally, the Proiects center at Berkeley would simply collapse if not for all the support from Elizabeth Burr, Regina Burley, and Diana Smith. Thank you all. Bruce Fuller Sharon Lynn Kagan fl Jan McCarthy Berkeley New Haven Tam pa EXECUTIVE SUMMARY RQMGMDEI’ th v a. a u:t1‘,:-,'a.s¥f.1 v, m ewe" ;~= ? 993.19%“. “2:: zm=:::: ~71: -:1-;w;xmizz:mm—Lmrm Delivering on Promises In 1996 President Clinton added his voice to a growing chorus of policy makers and voters, all pushing to revolutionize America’s welfare system. Whether these reformers, almost four years later, have strengthened the lives of poor children remains an open question. Hopes were high that summer of 1996. Congres— sional leaders vowed to end the “cycle of welfare dependence“ that seemed to entrap poor parents and their children from one generation to the next. The policy revolutionaries expressed a number of goals. They would shrink the welfare rolls and thereby build stronger families. Working mothers, moving into jobs, would offer stronger role models for their children and greater economic stability for their households. Child care programs would grow in number and quality, supporting youngsters early development. In short. the reformers aimed to reduce single mothers welfare dependency and boost childrens futures over time. But is welfare reform delivering on its promises? Can we discern observable effects from these reforms on young children? Does the welfare—to—work impera- tive alter maternal practices, homes, or child care settings in ways that advance children’s well—being? These are the core questions that energize this study. Young Children Enter a New Frontier No one doubts that we have embarked on a grand national experiment. For the first time government is seriously requiring that single mothers with preschool—age children work to qualify for time— limited cash assistance. \X’omen must now juggle the task of raising an infant or toddler with holding down a job. As a result, about one million additional children now spend their days away from their mothers in child care. These youngsters are entering their own 40—h0ur—a-week frontier, being raised by new adults in new settings. What is the quality of their child care? Do these settings advance or impede youngsters’ early learning? How do neighborhoods vary in providing organized child care? These are crucial questions if children’s well—being is truly our first concern. The Growing Up in Poverty Project During the second half of 1998 our research team, working from Berkeley and Yale, randomly selected 948 single mothers with young children. They live in or near one of five cities: San Francisco or San Jose, California; Manchester or New Haven, Connecticut; and Tampa, Florida. The samples proved to be representative of each area’s caseload. Participating women in California and Florida had been enrolled for six months in new welfare pro— grams. In Connecticut, we compared experimental and control groups 18 months after they had entered the new or old program. This report details major findings from the first wave of data collection. Our results stem from interviews of the mothers, visits to their child care BERKELEY — YALE a EXECUTIVE SUMMARY providers. both centers and individual caregivers, and assessments of children‘s early language and social development. The median child was 30 months old when we first interviewed the mother. Economic Incentives, Personal Resources, and Community Context Welfare reform rests on the assumption that by altering the economic incentives and moral obliga— tions attached to cash assistance, the lives of women and their children will improve. By setting time limits on cash aid. expanding family supports like child care. and raising tax benefits for the working poor. policy makers hope “to make work pay” and move single mothers from welfare to work. In this study. we move beyond the simple economics of women‘s lives to sketch their varied and complex attributes, along with the character oftheir diverse neighborhoods. We inquire about their households. sources of social support or degree of isolation, and their parenting practices. And for the children. we assess the nature and quality of their new child care settings. W’e explore how a neighborhood‘s infrastructure— especially the availability oforganized child care and effective delivery of financial aid—can mediate the direct effect of welfare reform on children. Presented here are the major findings that emerged from this first wave of maternal interviews and child care assessments. The nine chapters that follow provide details. How Are Children Faring under Welfare Reform? I lbzmg (Iii/dry” are moving info [ow—(1111150! (XI/M (Airmen/21g; 115 theirmarlms mowfi'om wdflm’ to work. This results in part from welfare reform. since single mothers must quickly find a child care provider. often without the financial aid to which they are legally entitled. Our observations ofchild care settings revealed quite low quality. on average. I] GROWING UP IN POVERTY PROJECT 1 But low quality compared to what? Earlier national studies have revealed unevenness in the quality of centers and preschools located in middle—class communities. Comparing our results to this earlier work, we find that children in the new welfare system have entered centers of even lower quality. Educational materials often are scarce, little reading or story telling was observed, and many children typically spend their days with an adult who has only a high school diploma. The two study sites in California represent important exceptions: center-based programs in San Francisco and Santa Clara County exhibited fairly high quality. This is a glimmer of good news. It demonstrates that well targeted subsidies to centers can improve the quality of care for children on welfare. This migration of young children into mediocre child care also is driven by the robust economy and demand for semi—skilled workers. The new welfare rules are not solely responsible for this trend. But policy makers must decide whether to address the development of children with the same intensity that they display in moving single mothers swiftly into jobs. Most participating children were not placed in centers but in home-based care. By this we mean licensed family child care homes or individual kin members or friends (kith and kin). who now qualify for child care vouchers worth up to $5.000 per year. These home—based providers fell below the average quality level of center—based programs. We observed fewer educational materials. much greater use of television and videos. and unclean facilities. In short. we find that the welfare-to-work push on single mothers is placing a growing number of children in mediocre and disorganized child care settings. This represents a lost opportunity. for we also have learned in recent years how high-quality child care can effectively boost early learning. I Child] care subsidies redo/9 unequulfiuctions of poor families and encourage the use of unlicensed cure. The share of women drawing their child care sub— sidy ranged from just 13% in the Connecticut sample to 50% in Florida. In all three project states, women are eligible for child care aid while on welfare and for at least two years after finding a job. This low use ofsubsidies represents a serious break— down in one key component ofwelfare reform. It constitutes a strong disincentive to work. Much policy attention has been paid to inadequate take-up rates for Medicaid and food stamps. The situation for child care subsidies, potentially a significant income support. is even worse. The propensity ofwomen to utilize child care centers, as opposed to kith and kin, is highly corre- lated with the supply of centers in their neighbor— hoods. And what’s striking is the magnitude of inequality that characterizes the supply of centers among the communities in our study. Disparities in supply range from 42 center slots per 100 young children in Tampa to just 1 1 enrollment slots per capita in Santa Clara County. With limited supply oflicensed child care and scarce knowledge of subsidies. many mothers have few options. I lining Children} early learning and development is limited by uneven parenting practices and big/J rates of maternal depression. \We found that certain parenting practices, such as reading frequently with one’s child, often are absent in homes. In addition, the incidence of severe levels of maternal depression was up to three times higher among participating mothers, compared to the national average.‘ Maternal depression is troubling for two reasons: it constrains women’s employability and reduces their children’s odds of thriving. In Connecticut, one in every six women suffered from severe depression. Mothers with preschool—age children experienced higher levels of depression, relative to women with older youngsters} Utilizing a second measure of EXECUTIVE SUMMARY what doctors call “depressive symptoms,” about half of all participating women in California and Florida displayed emotional difficulties. Both factors—disengaged parenting and clinical depression—can substantially retard infants’ and toddlers’ early learning.3 We detected delays in the language development of participating toddlers in California and Florida, relative to national norms. These gaps are not necessarily attributable to welfare reform per se. And no significant differences in child development were found between experimental and control groups in Connecticut.4 But it remains unclear how welfare reform’s promise of advancing children’s life chances will be met until these deeper dynamics are recognized. With many mothers debilitated by mental health problems and a wider range of women not engaged in positive developmental practices at home, how will the welfare—to—work push alone advance children’s well-being? How Are Mothers Faring under Welfare Reform? I A sizable share of women are moving into jolas. Among all participating women in the three states, about half were working and had selected a Child care provider for at least 10 hours of care per week within their initial months of welfare involvement. (Another share had selected child care even though they were not employed.) From the Connecticut experimental data we see that involvement in Jobs First did encourage a higher rate of employment—a 15% margin among mothers in the new program, above the control group. Many were pursuing postsecondary training while drawing cash assistance. And we found that mater- nal education is one of the most consistent predic— tors of both employment and positive parenting practices, similar to findings from earlier studies. BERKELEY — YALE E EXECUTIVE SUMMARY I Wages are low and household economies remain impoverished. The median Florida woman, when most recently employed, earned just $5.45 per hour, yielding a monthly income of $630. Average hourly wages were higher in California ($6.36) and Connecticut ($7.24) before adjusting for the cost ofliving. These women reported median monthly earnings of $700 and $799 in the two states, respectively. Fragile levels of economic support directly touch the lives of young children. Asked whether they had difficulty buying enough food, 28% of the Florida mothers and 32% ofCalifornia mothers said often or sometimes. I Levels ofeeouomie and social support gained hy the 1001218]! (I)? Illlé’l’l’ll. Just 16% of participating women in Connecticut reported that they lived with an adult who provided economic support for their child, compared to 36% among women in Florida. About one-quarter of all women appear to be socially isolated, rarely seeing other adults. Among sampled women in California, 4100 reported that they “feel alone as a parent.“ One fifth of the Florida mothers reported that their household includes one member with an alcohol or drug abuse problem. Such daily sources ofstress undercut the familys stability. Nor do all mothers feel efficacious about the chal— lenge of raising a young child in poverty. In Califor— nia, 390/0 of all women agreed with the statement, “At the end of a long day I find it hard to feel warm and loving toward my child.“ Taken together, these findings suggest that mere nranipulation of economic incentives and penal— ties—the carrots and sticks strategy—may be insufficient. This approach fails to recognize the force ofwomens personal resources. levels of social support. and their emotional health. The uneven availability ofchild care programs is another case of how the focus on engineered a GROWING up IN POVERTY PROJECT A incentives misses a key point: Beyond economic calculations, these mothers are empowered or trapped by their community’s resources. They benefit from a human hand when extended by local organizations. A Cautionary Note These Wave 1 results offer a first snapshot of how mothers and their young children are faring. The findings should be considered tentative, with emerg- ing patterns to be examined over time. To that end, we are collecting Wave 2 information from the same families, 18 months after Wave 1. A third round is planned after the children enter school. Caution is warranted in comparing findings across the three Project states. The Connecticut sample was drawn somewhat differently than the California and Florida samples. Participating women differ in their personal characteristics. The sample of women in Connecticut, for instance, is better educated than women in the other two states. We highlight below where between—state differences may be the result of sampling procedures. Companion Research This report is being published alongside two other studies. One is an evaluation of the economic and employment effects of Connecticut‘s welfare reforms, authored by Dan Bloom and others at the Manpower Demonstration Research Corporation (MDRC) in New York. The second focuses on the health of participating women, especially their mental health, and how such factors help to explain employability and wage rates. It is authored by Sarah Horwitz and Bonnie Kerker at Yale‘s School of Public Health.R Together, these analyses begin to inform the pressing questions around how children are faring under welfare reform, by looking into womens lives and into the new child care settings where their young- sters are being raised. SECTION 1 III This SECtiOII a How are young children and mothers faring under welfare reform? What is the Project attempting to find out? at How were families selected? What did we learn from them? Welfare and Child Care: Two National Worries Families and political leaders alike have struggled with two intertwined issues over the past quarter century. first, the harsh effects of family poverty are felt each day by one in five children across America ~~despite the current economic boom. How to widen job opportunities for poor fami— lies. and lower employment hurdles faced by single mothers. are tasks that have befuddled policy makers for decades. Second. many families with young children struggle to find affordable child care, hunting for quality preschools or caring kin members. Parents— rich. poor. and middle—class—search for safe havens and stimulating settings for their children. Today more than two—thirds of all mothers work outside the home. How to ensure safe, quality child care for all has become an equally challeng— ing policy issue. few question the need to craft policies and sustain local organizations that lighten the burden of poverty on children, while strengthening America‘s patchwork child care “system.” The elusive question is, How can we best accomplish these tandem goals? PROJECT AIMS H ou work almost 40 hours and go to school three days a week. So, it's affecting them a lot. l try to call them every day during my lunch time. l either call home or call their aunt My older daughter. she always tells me 'iiliornrnv, i love you and i miss you, too m shante Putting Evidence on the Table The Growing Up in Poverty Praject shines a bright light on how single mothers and their preschool—age children are faring in the context of welfare reform. The nation’s two worries haunt these families at a very personal level. Since 1992, single mothers have faced increasing pressure to move from welfare to work in a growing number of states. Serious conse— quences await those who are able to work but who fail to find a job. BERKELEY » YALE fl SECTION 1 Some political leaders and commentators applaud the decade—long shrinkage ofthe welfare rolls. But are these single mothers who sever their ties with public assistance really better off than before? And one question that many citizens have largely forgot— ten: How are the young children faring as their mothers leave home and go to work? Never before. prior to the reforms punctuated by President Clinton in 1996. have single mothers with preschool-age children felt such a strong push to work. The federal government and states did, from the late 19705 forward, experiment with “workfare” or welfare—to—work programs. But only 9% of all adults on AFDC were reporting earnings in 1994 in the earlier jOBS program. And many states ex— empted women who had a preschool-age child at home. But now single mothers must engage in work activities when their infant turns 3 months-old in some states. We aim to remember the children. Our starting point is to ask basic empirical questions: How are young children’s lives changing as their mothers move from welfare to work? What is the quality of child care in which youngsters now spend their days? How is the young child’s home environment changing, from altered parenting practices to more supportive households? Pancorous debate rocked the Congress during President Clinton’s first term, after he had promised to "end welfare as we know it." After vetoing two bills and allowing experimentation by the states, Mr. Clinton announced on the last day of July, 1996 that he would sign the Republican leadership’s third bill. The exchange that day between the White House and Capitol Hill encapsulated the decades— long attack on the old welfare system. It also amplified the high hopes expressed over how the well—being of children would be advanced by these historic reforms. President Clinton: "Today the Congress will vote on legislation... to transform a broken system that traps too many people in a cycle of dependence to one that emphasizes work. It gives those on welfare what we want for all families in America, the opportunity to succeed at home and at work. I made my principles for real welfare reform very clear from the beginning. It should impose time limits on welfare. It should give people the Will welfare reform improve children’s lives? High Hopes in 1996 child care and health care they need to move from welfare to work. (This new bill) provides $4 billion more for child care so that mothers can move from welfare to work...You cannot ask somebody on welfare to go to work if they’re going to neglect their children in doing it. This is the best chance we will have for a long, long time to complete the work of ending welfare as we know it and doing better by children." House Speaker Gingrich: "I think the children who have been trapped in poverty are going to have a much better future in this kind of environment where the work ethic is re-estab- lished, where child care is made available, and they have a chance to begin to climb the ladder of opportunity.” Congresswoman Jennifer Dunn: "We’re so glad the President said that he’s going to sign the bill. One of the reasons is he's finally discov— ered that we have taken care of the children. Mothers need to have the peace of mind to be able to move back into the work force."6 E] GROWING up lN POVERTY PROJECT A As welfare experiments were being initiated in the states, a wider civic debate over child care and early education grew louder. In turn, the political will to strengthen child care and preschool options has risen dramatically. President Bush, in 1990, approved the first national child care program, providing state block grants. Mr. Clinton has led efforts to expand and improve the quality of federally funded child care options, moving beyond Head Start. And many governors, have pressed to expand state preschool programs for blue—collar and middle—class families, while boosting child care vouchers for low—income parents, be they welfare poor or working poor. These developments frame the Project’s second set of questions: Are single mothers taking advantage of expanding child care options as they attempt to move from welfare to work? Do these highly variable child care options—from quality preschools to the aunt downstairs—advance or hinder the early learning of toddlers and preschoolers as their mothers go to work.> How Are Mothers and Children Faring? Proponents of welfare reform continue to argue that the pressure for single mothers to move from “wel— fare dependency” to work would yield benefits in terms of higher income, self sufficiency, and a stronger feeling of self confidence. Then children are to benefit, many claimed, since working mothers would become stronger role models, seeking to achieve more and setting higher aspirations for their children. Critics of welfare reform, in contrast, have emphasized that a decline in the welfare rolls does not mean that poor families are better off. They point out that since quality child care is often scarce in poor communities, the children of such families may suffer. Their mothers cannot stay with them at home because they have to work, yet stable child care of reasonable quality will be difficult to find. PROJECT AIMS Our research team—backed by policy leaders at state and local levels—resolved that these competing claims should be informed by hard evidence. De— spite all the polemics swirling around this huge social experiment, little data on how young children are faring has been gathered. This was the starting point for the Growing Up in Poverty Project. This Wave 1 Report describes—for 948 families who entered new welfare systems in California, Connecticut, and Florida—the key foundations of mothers’ and children’s new lives: film The uneven range of personal resources that mothers bring, from variable education levels, uneven social supports, to troubling levels of depression and stress at home. The low—wage labor markets that mothers are entering, as they struggle to meet work require— ments and make ends meet. The varying home environments and child care settings—displaying different levels of quality— in which toddlers and preschoolers are now being raised. The spotty participation in allied family—support programs, from uneven utilization of child care subsidies to mothers’ often tenuous links to Medicaid and the earned income tax credit (EITC). Welfare Reform's Effects on Families: What Are We Learning? The basics of welfare reform—tighter eligibility rules, pressure on clients to work, and services aimed at smoothing the transition to economic independence—have yielded dramatic results. If the main goal ofwelfare reform is to get people off welfare, the post—1992 policy experiments crafted by state governments and the Congress have been remarkably successful. Yet if the broader goal of welfare reform is to reduce the nation’s rate of family poverty and raise poor children’s quality oflife, the jury is still out. We are BERKELEY — YALE E] SECTION 1 Figure 1.1 Falling Welfare Caseioads: Annual percentage change in families on cash assistance I Connecticut I California Florida 1985-86 1990-91 just beginning to learn whether children are better now than they were before this decade of policy change. Child poverty has inched downward over the past few years. At the same time, the child poverty rate remains higher than in the late 1960s and far above levels typically experienced in Europe. What have we learned from recent studies about the effects of welfare reform on families? The best known fact: The number affirm/fies (Inna/Hg my!) (mist/Ina) lizlxflzl/elz. This shrinkage of the welfare rolls has been dramatic in many states. his due not only to new restrictions and work require- ments, but also to the nations booming economy which has created millions of service—sector and semi—skilled jobs. Over the last eight years, the number of families receiving cash assistance has declined from 5 million to under 3 million in 1999, currently the lowest level in over three decades. About half of this decline is attributable to the fast pace of job growth, accord— ing to one recent study by the presidents Council of Economic Advisors.“ Yet tighter eligibility, new work rules, and time limits on receiving cash assis- tance have prompted many families to leave the welfare rolls, and deterred others from signing up. [E GROWING UP IN POVERTY PROJECT A 1995-96 1997-98 Figure 1.1 displays the precipitous drop in welfare caseloads observed in our three Project states: California, Connecticut, and Florida. Over the past five years, the number of families on welfare has declined in California by 22%, 27% in Connecticut, and 62% in Florida. About 60% ofall remaining clients nationally are female—headed households with at least one preschool—age child. Many women zlz‘eflndingjO/Js after leaving vie/film but/em mm (Hang/,2 to live [Ivor/e the poverty line. Depending on the state and the study, between 51% and 71% of women leaving the welfare rolls are employed several months later. They earn $5.50 to $8.09 an hour, failing to raise most out of poverty.“ Over half of former welfare parents, interviewed in earlier studies, reported that they worry about running out of food." More than one-third reported that they are unable to pay rent or utility bills over the past year. The economic situation has improved for some single women with children. Between 1993 and 1995, income from earnings and benefits for the poorest women who head their household rose almost 14%, or about $1,000 yearly. But between 1995 and 1997—as welfare reform accelerated in most states—this same group suffered a 7% drop, losing $580 annually on average. This group equals about 2 million families with incomes below 75% of the federal poverty line, totaling 6 million individu— als when children are included.10 Famiiy participation in income support programs remains low. Families who leave the welfare rolls are using health insurance, food stamps, and child care subsidies at unexpectedly low rates. Almost half of all these families are no longer enrolled in the Medicaid health program, even though most remain eligible. Less than a third still use food stamps, even though two—thirds of this group qualify.11 Sixty percent of all remaining welfare cases involve a single mother with at least one preschool-age child. "u‘yxgfiuq" ~~ . c a“ ,. The use ofchild care subsidies—presumably a key support for enabling women to find and keep a job—remains well below 50%, even in counties that are aggressively trying to expand child care options. The use of child care vouchers has grown markedly in some states. This form of aid more than doubled in Illinois and climbed by 70% in Maryland during 1997, reducing the cost of child care for both welfare clients and working poor parents who had left the rolls.” But the rate of subsidy utilization remains low overall. Among new welfare clients in Los Angeles County, just one in five are using their child care voucher (see Section 8).” Under the old AFDC program, other family supports were tied directly to eligibility for cash assistance, such as Medicaid and food stamps. Under new welfare programs, however, they are not. This may account for why many low—income families who no longer receive cash assistance fail to use other income supports. The effect oflow participation on, for example, the quality of child care selected or children’s levels of hunger, is a question to which we return. PROJECT AIMS The local child care and preschool infiastructure is expanding at a modest rate in many states. Federal and state governments, in the wake of welfare reform, have invested billions of new dollars in child care and preschool programs. The federal government now invests about $1 1 billion annually in child care and early education, including Head Start and block grants to states.14 State spending also has been rising rapidly. California’s total budget for early care and education has risen from $800 million to $2.4 billion in just the past three years.15 Many of these new dollars are in the form of child care vouchers, aimed at widening options for welfare poor and working poor families. Yet expansion of the organized child care infrastruc- ture inside neighborhoods is unfolding at a slower pace. Many single mothers on welfare rely on kith and kin for child care services, often assuming that vouchers cannot be used outside centers or preschools. Despite investments in state preschool programs—most notably in Georgia and North Carolinawthe growth in licensed centers and family child care homes has been sluggish in other states. At times the enrollment capacity of child care centers lags behind population growth. In California, for example, slots in centers grew at an annual rate ofjust 2.2% between 1996 and 1998, 16 despite child population growth of 3.4% per year. What We Don't Know: The Project: 5. Major Questions Gauging the economic future of women leaving the welfare rolls is a crucial research task. Yet as data begin to sketch a more textured picture of such maternal effects, researchers have yet to provide parallel evidence on how young children are faring. By remembering the children, we illuminate their lives as their mothers enter new welfare regimes, focusing on these empirical questions: BERKELEY — YALE III SECTION 1 I How do childreifs home environments change as their mothers spend more time in welfare—to— work activities, then try to hold down a job? I What kinds ofchild care arrangements do women select, and does the quality of these settings help or hinder youngsters early development? D What support do mothers obtain from welfare agencies, household and kin members, and friends as they juggle jobs and child rearing? How do the features of neighborhoods—— including job demand, ethnic composition, and infrastructure—expand or constrain the choices available to parents? These questions require that we look inside house— holds, ranging from how adults share income and support each other emotionally to the mothers parenting practices, all of which may change as women spend less time at home. And we must observe the character and quality ofchild care settings, be they organized centers or homes, to fully understand how young childrens social environ— ments are evolving. Our research design takes seriously the idea that neighborhoods matter. One important case is how the availability ofchild care may condition womens ability to hold down a job. In addition, prior re- search has demonstrated how the character and quality ofchild care settings affect the early learning of poor children.r Blueprint for Wave 1 Data Collection All 948 single mothers and preschool-age children participating in the Project had recently entered new welfare programs in one of our three states. In California and Florida, Wave 1 data were collected during their first 2 to 6 months in the new programs. In Connecticut, all sampled women are participating in a random assignment experiment. We interviewed the mothers and assessed child care settings 18 months m GROWING up IN POVERTY PROJECT A after they had entered the experiment, living under new or old welfare rules. Each half of the sample falls into the experimental or control group. Many women in all three states had been drawing cash assistance from the old AFDC program, although this proportion varies across each state sample, as detailed below. All \Vave 1 data—including maternal interviews, child care observations, and direct assessments of children’s early learning—were collected between June 1998 and January 1999. We are presently collecting Wave 2 data, spaced 18 months after the Wave 1 collection. This will provide a look at longer-term effects on mothers and children. How Did We Sample the 948 Families? Step 1: Invite single mothers into the study. In California and Florida we met women eligible for the study when they attended orientation sessions in welfare offices. All had been deemed eligible to receive cash assistance under new state welfare programs. F.ach mother was unmarried and raising at least one child of 12—42 months when we invited her participation. In Connecticut we sampled single mothers with a young child in the identical age range. Unlike the women in California or Florida, however, about half the sample had been randomly assigned to the state’s new welfare program, Jobs First, 18 months earlier. The other half formed a control group which continues to live under the old AFDC welfare rules. Thus we are following women and children who are participating in a true experiment. Random assignment to the experimental (Jobs First) or the control group took place between late 1996 and early 1997, as each individual applied for cash assistance. A small share (8%) were deemed ineli— gible for cash assistance or never drew cash assistance after applying, but they remain in the sample. In California and Florida, the samples were recruited from welfare orientation sessions after eligibility had been determined. The Connecticut sample allows us to look at the discrete effects of this state’s welfare program. But since these families could have been exposed to the new program for a period up to one year longer than the families in California and Florida, we distinguish below any significant program effects stemming from Connecticut’s experiment; In all three states any sampled family may have dropped out of the cash assistance program within days or weeks of being recruited. Their duration of exposure to new state welfare policies might have been quite short, or ranged up to 6 months in California and Florida and up to 18 months in Connecticut (within our Wave 1 data window). We will be analyzing how the extent of exposure to welfare—to—work programs may lead to differing outcomes for mother and child alike. The Connecticut experiment is directed by the Manpower Demonstration Research Corporation (M DRC), under contract with the state welfare agency. Appendix 1 outlines the full Connecticut study and specifies how the Berkeley-Yale portions fit into the overall design. In all three states, we conducted an interview with each mother, visited the child care settings of most women using care. and directly assessed their children’s early development. A small subset of questions posed to participating women in Connecticut were not included on the California and Florida interviews, and vice versa. This stemmed from differing local priorities among our state and county—level policy colleagues. By the end ofour 1998 recruitment process, 948 eligible mothers with preschool-age children had agreed to participate. They reside in or near one of five cities: San Francisco or San Jose, California; Manchester or New Haven, Connecticut; or Tampa, Florida. These cities represent quite diverse settings in terms of state welfare requirements, local labor demand, and ethnic composition. They also differ widely in their community infrastructure, particu— larly in the availability and of neighborhood child care programs. PROJECT AIMS What Information Did We Collect? Step 2: Interview mothers ahout their households, children, wor/e, welfare involvement, economic and social supports. Just over 80% of all eligible women agreed to participate in the Project. We conducted a detailed interview with each of them that lasted up to 2 hours. During this structured discussion we collected information on a variety of topics, includ— ing the following: A basic demographic profile, including the mother’s school attainment, household composition, age and ethnicity, marital and fertility history. m Prior work experience and the flip—side, engage— ment with welfare and family support programs. m The rewards and stress associated with parenting, as well as the quality oflife with other household members. as Parenting practices related to young children’s early development, including reading practices, approaches to discipline, and learning activities between mother and child. iii? Sources of social support from kin members, friends, fathers and boyfriends. saw Maternal and child health, including assessment of depression and access to health services. «a Engagement in welfare—to—work activities, such as, attending job clubs, searching for work, and knowledge of new welfare rules. 1 Prior use of child care and plans for who will care for their youngster as the mother begins searching for a job. The mothers report on her child’s early language and communication skills, as well as social development and behavioral problems. Many of the findings reported in this report are derived from this maternal interview. BERKELEY — YALE [E] SECTlON 1 How Did We Assess Child Care Quality and Child Development? Step 5: Com/m1 zzfi)//0w—11p interview regard/Hg (Iii/(1' (are iIrmngenmztx. Two weeks following the initial interview we began calling each woman to see if‘she had found child care. By then many women had entered a welfare-to-work activity, often a job club where they prepare resumes and receive job counsel— ing. Many simply look for work on their own. A fair share were doing neither, ignoring the new requirements or unaware that sanctions would soon follow their inaction. During the Wave 1 period in California and Florida we stayed in touch with each mother for up to 6 months to see it‘she had selected a child care pro— vider. In Connecticut the follow—up child care questions were embedded in the maternal interview, since 18 months had elapsed, allowing them to sort out a child care arrangement(s). Also note that child care utilization refers to the woman’s current or most recent provider (if any) during the 18-month period. In California, child care use also pertains to the most recent provider selected since entry to welfare, but the time frame, as before, is up to 6 months. For the Connecticut sample, steps 1, 2, and 3 were managed by MDRC. Interviews were conducted by phone or during a home visit. A total or 56‘) women (or 60% of the total Family sample of“)48) reported using a child care pro— vider—either a formal center or a kin member or friend—for at least 10 hours per week within the survey period. At first glance this proportion seems high, given that the overall share ot‘TANF clients Table 1.1 Participating Women, Child care users, and Clll‘ECt assessments California Step 1 Number of eligible women 473 at sampling sites COIIIIECtiCllt Florida 342‘ 2’I8 Step 2 Number of completed maternal interviews 414 (88%) 3’l1 (90%) 202 (93%) Step 3 Number of women using a child care provider 259 (55%) 167 (54%) ”142 (65%) Step 4 Number of direct child care and/or child assessments 216(460o)? 143 (46%) 165 (82%) Step 2 percentages are for participation rates, relative to eligible women (Step 1). Step 5 and Step 4 percentages are for women using a child care provider and child care or child assessments completed, respectively, relative to the number of completed maternal interviews (Step 2) Number of women at random assignment who we knew would have a child, 12—42 months old, at 18 months [E] GROWING up IN POVERTY PROJECT A following random assignment. There is a six-month window where a small but unknown number of women may have given birth. This report excludes child-only cases. So, does not add to 948‘ ‘ The child assessment completion rate equals 56% in San Francisco and 65% in San Jose. engaged in work activities equaled 21%, 58%, and 28%, respectively, for California, Connecticut, and Florida in 1997. Yet, a significant share of women were already utilizing subsidized child care through programs such as Head Start or state preschools prior to entering the new welfare program. This was true for 23% of the women living in San Jose (Santa Clara county) and 47% of those in Tampa. Pinpointing the discrete effect of welfare reform on the move— ment of children into child care is complicated by this fact; some mothers utilize a child care provider or preschool whether they work or not. The child care questions posed to the mother covered a range of topics: K The amount of time the child spends away from the mother, where, and in what type of child care setting. L f... & \Who pays for child care and participation in subsidy programs. u L. if Sources of information on which the mother relies when it comes to finding and judging the quality ofchild care. The helpfulness of caseworkers or local agencies in finding a child care provider. The quality and flexibility (in hours open) displayed by the child care provider, as reported by the mother. Difficulties in finding or keeping a job due to child care hassles. Step 4: Viki! and (’Z’tl/ZIZZH’ the quality ofchild care settings, and (155655 the child’s early learning and dez’e/npnzmr. We then asked each woman if we could visit her child care provider for a morning or afternoon. The purpose of the visit was to interview the teacher or caregiver and to assess the quality of the setting. Overall, 71% of the mothers using child care allowed us to visit and complete all assessments, including a direct evaluation of the child’s early language development. In addition, PROJECT AIMS each child’s social development and behavior problems were assessed by mothers and child care providers. Table 1.1 summarizes the four stages of data collec- tion and illustrates how sample sizes decreased from one stage to the next. This is typical in studies of low-income families. The study began with 948 participating mothers, out of 1,079 eligible women across the three states (before excluding child-only cases). By “eligible” we mean women who met our selection criteria, single and with at least one preschool- age child, 12—42 months of age. For Connecticut, this includes the full population randomly assigned who met these criteria (360), minus a small number of child-only cases and families that had been exposed to an earlier welfare—to-work program (18 cases). For California and Florida, this includes all women who met our our selection criteria. We then conducted follow—up interviews for the 568 mothers who had selected a child care provider within 6 months ofour initial meeting.18 At step 3 we successfully gained access to 460 providers or to the mothers’ homes to conduct the direct child assessment. This equals 81% of the women who had selected a child care provider. Relative to the original sample of 948 families, we were able to complete in—person assessments of49% of the children. In addition, we asked all mothers to report on their child’s language proficiencies, social development, and behavioral problems. These measures are predictive oflater learning and school performance, as detailed in Section 9. We obtained useable data on these child development measures for 93% of the 948 original families. Wave 2 Data Collection The second round of maternal interviews and child assessments began in September 1999. It includes a phone survey, followed by a 2—hour Visit to each mother’s home. During the visit interviewers assess the mother’s and child’s cognitive and language proficiency, as well as the youngsters social BERKELEY — YALE [B SECTION 1 development. Additional data on children’s early school achievement and behavior inside classrooms will be collected during Wave 3. set for 2001. A Representative Sample of Families? All large field studies are faced with a potential problem linked to the sampling strategy: Is the final sample of families. drawn to maximize randomness, truly representative of the overall population from which the sample was selected? We did not intend to draw a nationally representa— tive sample. Instead, we chose five diverse cities in three states with contrasting welfare reform pro- grams and child care infrastructures. Our city—based family samples. then. should be representative of that portion of their welfare caseloads that met our selection criteria. This sampling strategy aims to obtain samples that were representative of urban caseloads while allowing us to focus on local differ- ences in welfare regimes. neighborhood features. and child care infrastructure. Why Select Families From Five Cities? Our decision to select families from five cities is based on the fact that family support policies have become quite decentralized. Parents and children now face a disparate array ofeligibility rules. uneven benefit levels. and differing incentives and punish— ments to encourage participation in work activities. These local differences are layered on top of differing local economies and the historically decentralized character ofchild care and preschooling. Neighborhoods may matter a lot in mediating the family—level effects of welfare reform. For the most part. local differences in the number of available jobs for low—skilled workers are not determined by government policies. But state and county agencies do set the rules for cash assistance. They also have [13 GROWING UP IN POVERTY PROJECT A the authority to build a stronger child care and early education infrastructure within neighborhoods, or choose not to do so. Potential Sources of Bias in the Family Samples Two possible sources of bias must be noted. Both are linked to how we obtained the family samples. First, before completing the initial interview we asked each woman ifwe could visit the child care provider or her own home to conduct the direct child assessment. We knew that it would be difficult to obtain permission ifwe had not already gained the mothers trust during this initial meeting. While many surveys conduct interviews by telephone in order to minimize bias, doing so makes it difficult to establish a trusting relationship with participants. The decision to select families from five cities is based on the fact that family support policies have become decentralized across America. ‘16,»:4‘ us. ~ , wag/.4 We therefore chose to meet and personally recruit women at the orientation sessions. typically held in local welfare offices (for California and Florida). By doing so. we met face to face. explained the study and its aims. Our research staff reflected the diver— sity of mothers in each local area. In Connecticut. MDRC colleagues randomly assigned all women to the control or experimental group. so we knew that they had passed the welfare eligibility screen. Sam- pling bias for this latter group is not an issue. The down side of our recruitment method is that a significant rate of no—shows for orientation sessions emerged in California and florid-a. as the rolls continued to shrink in 1998. This makes it neces- sary to compare our resulting family sample with known characteristics of each city‘s caseload during the recruitment period. Fortunately. we observed Table 1.2 How representative are the family samples? Mother's Age Ethnicity (%) 0n aid latino Vietnamese ill prior year (median years) Black San Francisco Sample 29 57 18 O 97 Population 26 56 16 3 92 San Jose Sample 29 7 5’l 26 98 Population 26 11 51 ’13 NA Manchester Sample 26 20 ’i9 0 60 Population 25 25 16 O 48 New Haven Sample 25 44 2’1 0 65 Population 24 42 50 O 57 Tampa Sample 52 47 14 ’l 92 Population 50 47 25 0 NA ‘.:te tomcat isuns are for eligible countywide populations of clients: single mothers with preschool-age children seiected dann :1 one montr in mid—1998 For Connecticut, the percentage of women on aid is for the year prior to random assignment. onh slight dilierences between the sample and The second possible source otisample bias pertains know It features of local caseloads of single mothers to the direct child assessment data. On Table 1.2, with preschool—age children ( lable 1.2). under the step 4 column, we see that direct assess— . c ,. ments of childrens eaer language development were .\ote that we purposetulh' over sampled \ letnamese— - ’ _ , - _ . . . ' . . . . conducted from 49% of the or1g1nal family cases. American women in San lose. lhis group comprises . _ . , _ , fl , . , . ’ ‘ ‘ (,alifoi‘nias rate was Just 46%, although thls IS a sizable portion or the countx's caseload. Because . , _ a , . t , j primarllv due to mothers in San Francrsco who were us an under studied population. we want to analvze -' , , , , , t- ,. . ' uncomfortable w1th home ViSits. The San Jose chlld ditlerentes between \ tetnamese and other ethnic . . . 1- assessment completion rate equaled 62%. groups. so we drew a greater proportion of \ ietnam-r C‘ ese women into our sample than their proportion Possible bias here stems from the fact that most ot‘the overall caseload. In calculating means For the two (ialitornia counties (including San Francisco), we always down-weight the Vietnamese cases to make them proportionally equal to the overall caseload in 1998. direct assessments were completed in child care settings, since access was easier and less costly than trying to arrange home visits when mother and child would both be at home. In addition mothers who are reticent about having fairly unknown Visitors '7 at l‘ ' ii i'i l. 5: i ‘t’ “i"ri-‘i. i. if; SECTION 1 may differ from those who are comfortable with this proposition. Appendix 3 reports on how the two sets of families differ: those for whom a direct child assessment was completed and those without. IS Welfare Reform Advancing Children's Well-Being? Many of the findings that follow are descriptive. We want to acquaint you with these women, their homes, and their young children. In addition we detail how their contexts differ, ranging from between—state differences in welfare policies to between-neighborhood differences in the availability oforganized child care. Knowing how these women‘s contexts vary, across states and sometimes among cities, helps in understanding how they are touched by welfare reform. \Ve also put forward claims about how children‘s lives are being touched by the economic carrots and sticks wrapped up in recent reforms. For instance. you will see how mothers are placing their young children, averaging 2 and one-half years old, in child care settings that display low quality on average, especially in Connecticut and Florida. Even when compared to oft mediocre quality faced by middle— class parents, the settings in which participating mothers are placing their young children exhibit an even lower degree ofquality. You must then ask. ls the welfare—to—work push a direct cause ofthis migration of young children into poor child care.> Certainly fewer children would be attending these settings if their mothers were not being forced into work activities. And these new settings represent social environments that may impede childrens early cognitive and social develop— ment. But welfare reform is not the only contribut— ing factor: robust job growth and sluggish state action to improve child care quality also contribute to the stultifying settings that are experienced by many TANF children. [[3 GROWING up IN POVERTY PROJECT “‘1 The Connecticut findings are more convincing in terms of attributing effects to welfare reform. Where families in the experimental group are doing better, compared to the control group, the differences are likely real at the 18-month point. You will discover that a higher percentage ofwomen in the new welfare program are working and, in turn, they earn more, compared to the control group who can still stay at home and face little pressure to work. The differences are modest but discernible. We do have more work to do in exploring how the force of welfare—to—work pressures may interact with women’s own personal resources, such as education level, mental health, or social supports, to yield positive or negative outcomes. The findings for California and Florida families must be interpreted carefully. These families had experienced not more than six months of their states new welfare program. No true experiment is being run. although we can contrast women who spend more time on the job, compared to those who are less engaged in the workforce. Descriptive findings can be quite telling. For ex— ample, women‘s reported levels of depression are quite high. The extent to which women read to their children and engage in educationally rich activities is very uneven. Most mothers fail to benefit from child care subsidies. Many locate child care that displays mediocre to downright poor quality. A portion of these events reflect the realities oflife in poor neighborhoods. They are not the causal result of welfare reform. But a portion reflect the inability of welfare reform—and its explicit policy aims—to make a real difference so far. \V’e present comparative data from earlier research that contrasts our participating mothers and chil— dren to wider national samples on levels ofhunger, mental health, child care quality. and indicators of child development. Some readers may argue that we are setting the bar to high. th‘OlH'St), poor families on welfare suffer from hunger more frequently, mothers face more depressed child care markets, and mothers are more often depressed emotionally. So the argument goes. But proponents of welfare reform have promised that children’s lives will get better. If true, then the antecedent determinants of children’s well—being— that come alive inside homes and in child care settings—should be showing some upward move— ment. It is this claim that our baseline data and causal evidence are beginning to inform. In sum, we report on differing patterns among the three state samples of families. This illuminates how state and local conditions—including how welfare reform is being implemented—yields differing experiences for mothers and children. Second, we compare our participating families to the well-being and quality oflife experienced by broader national populations when such comparative data are avail- able. Third. we press forward on the question of whether the aims of policy makers—for instance, extending child care assistance to all welfare families— are truly being met in these initial years of reform. The Importance of Policy and Institutional Histories As we discuss these initial findings with our host welfare and child care agencies, we already see a good deal of hand—wringing and consternation. Indeed several findings are cause for concern. Yet our hope is that this report stimulates positive dialogue. not finger pointing. These families will only be helped through ongoing policy action and attention to the devil in the details of implementation. \We do not intend to direct blame at any current state administration or local agency. Our aim is to illuminate how welfare reform is, or is not, changing the daily lives of women and their young children. Our own hope is that empirical evidence will clarify which issues on the unfinished agenda of reform should receive greater attention. Much work obvi- ously remains in delivering brighter futures for the nation’s children who continue to grow up in poverty. PROJECT AIMS The potential of future policy adjustments to aid poor families depends upon each state’s own past— its policy history and prior generations of institution building inside neighborhoods. One reason that the Section 3 analysis of state and local contexts is so detailed stems from the fact that family support policies have evolved in such variable ways across the states. The rate at which state governments have expanded licensed child care organizations and pushed to improve quality, for example, varies dramatically across the three Project states. Next we describe the basic attributes of the 948 participating women. You will see that these single mothers and their households are diverse in many ways, from women’s prior work experience to how they patch together social and economic support. We detail their similarities and differences in Section 2. Then in Section 3 we explore how state policy contexts differ, as well as wide varability among the local neighborhoods in which these women and young children reside. BERKELEY — YALE [E s-"‘)\_'2"\1MHMN POVUQ'I'Y PRO JLiCT A SECTION 2 In This SECtiOII % Who are the women entering new welfare programs? a What are the basic features of their households? Have they been working recently, or receiving cash assistance? Diverse Women, Diverse Households Let us introduce you to the 948 mothers participat— ing in the Project. Many are quite remarkable as they struggle to make ends meet, raise their preschool— age children, and now face the push to find a job. Here we simply sketch their basic attributes and their household situations. Remaining sections will detail the womens variable characteristics across a spectrum of domains: their work experience, how they gain economic and social support from others, their views on parenting, their uneven emotional health, and the decisions they are making about child care. Indeed, diverse personal resources and family contexts represent key features of the overall story, bearing on the variable well—being of mothers and children. Education Levels Earlier research consistently shows that maternal education levels—which vary substantially among low—income and working—class women—contribute to a variety of positive life events. These range from higher maternal employment rates to stronger child development for their own youngsters.” MOTHERS' ATTRIBUTES ATTRIBUTES u think my best check was $380 for two weeks which is not a whole heck of a lot to pay off the bills, buy groceries. Now, no more food stamps. It's still almost the same as if I was on welfare. I haven't accomplished very much." — Betty We reported in an earlier paper that participating mothers with more formal schooling exhibited pro— development parenting practices, such as reading to their young child and going to the library or museum more frequently. The woman’s amount of work experience did not affect the quality of parenting.20 These findings hold implications for current welfare reforms that tend to restrict women’s ability to pursue further schooling or job training. BERKELEY - YALE E] SECTION 2 Figure 2.1 Education IGVEIS 0f mothers 100% . i1 [(71 , , ., 1. j 14% 50% 0% California Connecticut n = 41 5 n = 308 Figure 2.1 illustrates maternal education levels for participating women by state. School attainment was lowest for sampled women in California, with only 14% having received any type of formal train- ing beyond a high school diploma or GED. Among participating Tampa women the figure rose to 22%. and for Connecticut women it was 35%. In Connecticut the experimental and control groups did not differ in terms of school attainment. Ethnic Diversity The ethnic composition of participating women varies substantially across the three state samples, reflecting the demographics oftheir local communities. For example, 42% of sampled women in California are Latina, compared to just 14% in Tampa and 20% in Connecticut (Figure 2.2). On the other hand, African—American women comprise 47% of the Florida sample and 38% in Connecticut but just 24% ofthe women participating in California. One’s membership in a cohesive ethnic community may hold important implications. For instance, we a GROWlNC UP IN POVERTY PROJECT ’2 Some postsecondarv High school/ GED I Less than high school Florida n = 200 Education levels are significantly different by state. Numbers may not add up to 100% due to rounding. know from national samples of families that ethnic groups vary. on average, in the extent to which they read to their preschool-age children. We know that social supports at home differ. And we know that in some communities, women from particular ethnic groups are more likely to remain on welfare, relative to others. after taking into account alternative factors.31 Future work will examine the extent to which ethnic norms or community contexts affect important outcomes. from maternal employment to child care selection. We oversampled Vietnamese-American women in San Jose (Santa Clara County) given our interest in analyzing this subgroup. This population represents a major part of welfare caseloads in several California cities. Before weighting these women statistically, they comprised 22% of the California sample. Almost no Asian—American women were randomly selected in either Connecticut or Florida. Ethnic differences are apparent between cities within the same state. For example. the San Jose sample was 51% Latina, considerably higher than for San Figure 2.2 Ethnic composition 0f mothers MOTHERS’ ATTRIBUTES 1 1 % 42% 22% 42% 39% 24% California Connecticut Florida n = 414 n = 202 . Asian . Latina I Bl k . - ac (Vietnamese) Wh'te/ Anglo Francisco where they made up just 19% of the sample. The San Francisco city sample was 57% African—American. compared to just 7% in San Jose. Similarly, a majority of participants in Connecticut resided in predominantly black sections of New Haven. The remaining Connecticut women lived in sections of Manchester that were mostly white. Additional Features of Mothers and Their Households Table 2.] reports on additional characteristics of participating women, their households, and their children. These mothers vary in their average ages across the three states. Participating women in Florida are the oldest, averaging 32 years in 1998. Connecticut mothers are the youngest, with a mean age of26, while California participants fall in between, avera rino 29 ,ears. in a I The mean “focal child”—identified as the oldest youngster between 12—42 months of age when we first interviewed the mother—also differed in age among our three state samples. The average focal child in Connecticut was 25 months, compared to 29 months for children in California and Florida. The state samples differ in terms of marital experi— ence. For instance, 31% of the women in California had been married, compared to 42% in Florida.22 This is partially an artifact of the older age distribu— tion among women in the Florida sample. We also find that women with stronger work experience (in the 12 months before joining the study) were more likely to have been married in California and Florida. In Connecticut, women entering the new Jobs First program were less likely to have been married than the control group.23 Home Characteristics and Wage Earners Next we examine the simple social arrangements in which the women live. In the Connecticut control group, for example, the average woman lived with 0.5 other adults. That is, every other woman lived with one other adult, roughly speaking. In California, by contrast, the average woman lived with 1.5 other adults. This average is driven up by the larger households that often characterize Latino and Vietnamese—American households. Yet these household members are sometimes not related, nor reportedly very supportive of the mother, as we detail below. BERKELEY — YALE EB SECTION 2 The average number of children living in the mother‘s household also varies among the state samples. Again this is probably linked to the mothers relative ages across the samples. In California and Florida, where the mean mothers were 2‘) and 32 years, respectively, 2.7 and 2.8 children under 18 resided in the household when we first interviewed the mother. In Connecticut, where the mean mother was just 26 years, households contained an average of2.1 children. To learn more about the household economy, we inquired about wage earners present in the house— hold. In California, for example, there were 0.7 adults in the household who held a job, on average. The propensity of living with a wage earner equaled 0.8 adults in Florida on average. Roughly speaking, about 6 of every 10 women in these two states reside with at least one wage-earning adult (excluding the mother ifshe is working). In Connecticut the interview question differed a bit. We asked whether any number ofwage earners lived in each household. In response, 27% of the experi— mental group said yes. compared with 24% for the control group. In general. the younger Connecticut women are less likely to be residing in a household with a wage—earning adult, compared to those in California and Florida. Women in Connecticut were more likely to be living in public housing or receiving Section 8 housing vouchers. This was true for 41% of the experimental group and 40% ofthe control group women. By comparison, just 20% of the participating women in Florida reported any housing benefits; in California the proportion equaled 30%. Recent Work and Welfare Experience Participating womens prior levels of work experi- ence differed significantly among the state samples. Among women in Florida, fully 84% reported having wage—earning employment sometime during the 12 months prior to their new current enrollment in welfare. This was only true of45% in California. m GROWING up IN POVERTY PROJECT A In Connecticut, 46% of all women had worked in the year prior to random assignment (from MDRC’S background information file, not our survey data). Women in California spent more months on welfare during the prior year, 10.3 months on average. Florida women, on the other hand, received welfare assistance for only 4.4 months during the same time period, on average. Part of this difference may be explained by the older age profile of the Florida mothers. But the main factor may be that the California sample was drawn as San Francisco and Santa Clara counties were re-enrolling existing clients into the new state welfare program. I—Iillsborough County (Tampa) had been imple— menting its welfare reform program for more than a year when the maternal sample was pulled in 1998 (Section 3).“ The question on welfare experience was asked a bit differently in the Connecticut survey. Among the experimental group, 43% of the women had been on welfare for four months or less during the previous 12 months. Among the control group this percentage equaled 38%. Like Florida, the Connecticut sample includes a more representative mix of short— and long-term welfare clients, compared to California where the re—enrollment process brought in a higher share of long-term clients. Summary: Implications for § § Interpreting our Findings. It is important to keep in mind these attributes of the participating women, especially the variations seen across the three state samples. We urge you to remember three things: Individual women vary widely in their education levels, ethnic membership, and prior work experience. These features are associated with positive and negative outcomes which bear on the quality oftheir lives and that of their Voung children. I k The three state maternal samples vary, stemming from the demographic profiles of each county’s caseload and (for California) when the samples were drawn. For instance, the Connecticut maternal sample is more African—American, better educated, and comprised of younger women with younger toddlers. The stage at which welfare reform was being implemented affected the mix between short— and long—term clients drawn into the state sample. fl Having younger toddlers in Connecticut, compared to California and Florida, may hold implications for the type of child care selected. Parents with infants and young toddlers—— whether poor or middle class—tend to select more informal child care arrangements. On the other hand, we will show that the much lower propensity of Connecticut mothers to select center—based child care is more a function of constrained supply than any difference in child age. Next we turn to the cities and neighborhoods in which the participating families reside. Economic and social contexts make a big difference for the quality of poor women’s lives and their children’s well-being, beyond the force of individual—level Y€SOLIFCCS and constraints. MOTHERS' ATTRIBUTES BERKELEY - YALE m SECTION 2 E13 GROWING up IN POVERTY PROJECT A SECTION 3 STATES AND NEIGHBORHOODS III This SECtiOII How do states' welfare rules and child care policies differ? How diverse are families' neighborhoods? To what degree does the availability of child care vary across communities? State Policies and Neighborhoods Matter The centuries—old debate over how to best aid poor families often returns to the difficult balance be- tween personal responsibility and the collective obligation of public organizations. Even the politi— cally attractiye motto, Mug/J /01i€ reveals our ambiva- lence as .1 society over how to assist and empower the poor. Contemporary family policies stem from these competing philosophies about how best to aid low— income parents. For example. cash assistance is now contingent upon looking for and finding a job, even if a mother has a young infant at home. Personal responsibility means that government now requires women to draw on their personal resources and wherewithal to meet the moral obligation ofworking. Yet our society —and its public agencies—continue to exercise a good deal of collective responsibility in aiding poor parents and their young children: refundable tax credits flow to parents who stay off welfare and in the workforce; the expansion of child care vouchers has been dramatic and the federal governments new child health insurance initiative is taking hold. y morn works graveyard. So in . f- the morning I catch the bus . f to her house at six o'clock. Then. me and my son wait until she grits there, then i catch. the bus to inifir‘i‘i. So, “ rarer hard," Shirl Even as cash assistance for the shrinking numbers of welfare-poor families is linked to personal responsi- bility, local community institutions are being expanded to provide crucial family supports, from child care to housing. Much of this new aid is for the more politically popular group, the working poor. These initiatives also benefit a growing number of former welfare clients. And the strength of these small—scale “insti- tutions” may condition the long-term effects of welfare reform on children, especially when it comes to the availability of child care organizations. To fully capture how mothers and children are faring under welfare reform we must understand how their local neighborhoods and human—scale community BERKELEY — YALE EH SECTION 3 organizations vary. And the strength of these very local organizations flows from variable state policies and funding streams, programs which channel dollars to neighborhood organizations and to families in the form of vouchers, from food stamps to chits for child care. This section focuses on tandem levels oforganizational context: we start with state policy environments and work our way down to neighborhood infrastructure. State Welfare and Child Care Policies Under the devolution of family support programs, governors and state legislatures now craft unique or shared elements of welfare programs. In some states, notably California and Florida, counties can further fill—in the picture around eligibility for cash assis— tance and shape the child care infrastructure. In many ways, state and local authorities have become quite influential in adjusting the fulcrum which balances personal obligations against collective, public responsibility for helping poor families. Neighborhood Economies, Norms, and Organizations Recent studies also have shown how the quality of local neighborhoods can mediate the direct influ- ence of family—level poverty on mothers‘ lives and their childrens development. The social norms. micro economies, and available family—support organizations vary dramatically across low—income communities and in ways that affect children. Prevalent adult role models also can be healthy and enabling in one poor community but not in another. Stable neighborhoods display higher levels ofsocial support focused on the care and rearing ofchil— dren.2R And where parents and adolescents are known in the neighborhood. adult supervision and social norms remain influential.“ Closely linked is the argument that local schools and quality preschool programs can nurture stronger networks ofsupportive parents. Much more work remains to be done in specifying [7010 social norms E3 GROWING UP IN POVERTY PROJECT A and concrete organizations affect poor children, but the diversity oflow—income neighborhoods and their effects on children are now being illuminated more brightly. Finally, the strength oflocal institutions can enable women to enter the workforce with fewer barriers in their path. For example, Rachel Gordon and Lindsay Chase-Lansdale recently demonstrated how the per capita supply of child care centers in low- income communities is associated with higher maternal employment rates, after taking into account mothers’ demographic characteristics? Another study found that low-income communities with more churches per capita display a higher supply ofchild care centers and preschools, again taking into account a variety of other factors.28 Here again we see that neighborhood infrastructure matters. Key Contexts Surrounding the Family For all these reasons, it‘s important to learn about how state policy contexts and neighborhood organi- zations may mediate welfare reform’s effects on families. This section focuses on four specific contexts: 5mm ltd/[Ire I‘ll/(’5 and incentives. Time limits. work requirements, and the provision of child care supports vary among the three states, both official policies and uneven implementation on the ground. [ova/job markers. Employers demand for semi— skilled workers varies among the participating cities. influencing womens job opportunities. Neighbor/10011]p01wry and community 21027115. Participating families reside in remarkably diverse communities in terms of average house- hold income, child poverty rates, and family— support services. (Vii/[1’ our I'lifi21.s‘f2'zri‘r1/re. The availability of licensed child care organizations varies remark- ably across participating cities. leading to very different social settings for young children. State Conditions, State Policies Demographic and Policy Contours Our three states display sharp differences in their levels of affluence and historical commitment to supporting low—income families. For example, Connecticut is the richest state in the nation with a median family income of over $62,000 in 1997. California ranked twentieth at $48,000; Florida occupied thirty-fifth position at $43,400.” Just over one in four children in both California and Florida live below the poverty line, compared to one in five in Connecticut. In California, 35% of children reside in over—crowded housing as defined by the census bureau, compared to 20% in Florida and just 9% in Connecticut.” Total child welfare spending equaled $1 17 per child under 18 in Florida for 1996, compared to $197 in California and $227 in Connecticut.“ All three states have state—funded preschool programs, but they are highly variable in scope and quality. None ofthe three Project states has created an earned income tax credit program to provide stronger incentives to stay off the welfare rolls. Nationwide, 10 states have enacted such programs. Welfare and Family Support The states also differ in their eligibility rules and incentives aimed at supporting clients’ welfare—to- work transition. Devolution means that states and local counties can set divergent work requirements, levels of cash assistance, disregards for earned income, and links to allied family support programs (like Medicaid, food stamps, and child care vouch— ers). 'lable 3.1 details these variations. Basic cash payments to welfare clients were set at $565 monthly in California for a family of three, compared to $303 in Florida in 1999. Connecticut falls in between at $464 monthly. Yet Connecticut offers a stronger incentive to work by allowing women to retain all of their earnings, provided STATES AND NEIGHBORHOODS earned income does not add-up to more than the federal poverty line. This is known as a 100% income disregard. The other two states reduce welfare checks in proportion to income earned as women move into jobs. In Connecticut, once a woman earns income that exceeds the federal poverty line she becomes ineligible for cash assistance.32 On other hand, women in Connecticut face the most stringent time limit for receiving cash assis- tance, just 21 months of eligibility over a client’s lifetime. Connecticut is granting six-month exten— sions to clients who make a “good faith effort” to find a job, or whose income falls below monthly cash aid levels. In contrast, Florida enacted a 48- month lifetime limit, while California set the limit at 60 months. Work requirements tied to cash assistance are roughly equal across the participating states and rise over time under federal mandates. It’s important to note that the pressure on women to gain employ— ment after qualifying for cash assistance is almost immediate. Once approved, they must attend an orientation session and a “job club” to begin search— ing for work. Mothers in Florida are only exempt from work requirements prior to their infant turning three months-old. W», .. . .. ma». .» 0‘ , vm m»; and» ma- L . wen/ramwsee- vamr: an; Single mothers with infants and preschoolers, prior to the 1996 welfare overhaul, were typically exempt from work requirements. But under the new rules, set by states, California and Connecticut, mothers must begin work activities when their infant turns 12 months of age. Women in Florida are only exempt from the work requirement prior to their infants three month birthday. And in all three states the parent’s clock is ticking, against their lifetime time limit, when they choose not to work with a young baby. BERKELEY - YALE m SECTION 3 Table 3.1 Contrasting state welfare rules: Provisions affecting mothers and children California‘ Cash aid levels (TANF) Monthly cash aid for family of 5 with $565 parent working 0 hours in 1998 Percentage of cash aid retained when moving from O to 20 hours of work weekly at minimum wage 81% retained COIIIIECtiCllt Florida $464 $503 100% retained2 60% retained Time limits on cash aid Limit on consecutive months Lifetime limit 24 months3 60 months NA 24 of any 36 months5 21 months4 48 months WOl‘k requirements Weekly work hours required (for single parents) Exemption from work requirement for mothers with infants Reduction in cash aid after first failure to meet work requirement 26 hours per weekf‘ Variable up to Yes, if child is under 12 months7 Partial reduction 20 hours per week 35 hours per week Yes, if child is under Yes, if child is 12 mos, child capS under 3 months Partial reduction Full elimination SOURCES: San Francisco Department of Human Services, 1998. San Francisco County (fa/WORKS Plan (Submitted January 7). San Francisco 1997. Ca/ifornia State We/fare Reform, as provided va6’ 7542 and AB'iOOS. San Jose. Santa Clara County Social Services Agency. 1997. Employment Support initiative Action Plan (May 16, 1997). San Jose. Florida Legislature. 1996. WAGES: A Plan to Reform ll/e/fare in Florida (Conference Committee on Welfare Reform, CSSB 1662i, Tallahassee. Weil, A, 1999. ”Program Redesign by States in the Wake of Welfare Reform: Making Sense of the Effects of Devolution.” Washington DC: Urban institute. ' Rules differ slightly between San Francisco and Santa Clara counties, 7 Connecticut is experimenting with a full income— disregard. That is, welfare checks are not reduced by a percentage of wage earnings until the client’s income reaches the federal poverty level. State legislation specifies an 18-month limit, with 24 months permitted. Both San Francisco and Santa Clara counties moved to the 24-month limit for almost all client cases. an GROWING up IN POVERTY PROJECT A ; Families may receive a 6—month extension beyond the time limit if they have income below the payment standard and have made a good faith effort to obtain or retain employment, or if they encounter circumstances beyond their control that preclude employment above the payment standard. ‘ A client must go off cash assistance for 12 months after receiving cash assistance for 24 consecutive months. . This is the level effective July 1998. it increased to 52 hours per week in July 1999, uniform under state legislation. ' State law specifies that parents with an infant 6—12 months of age may be exempted from work require— ments. Both counties have chosen the 12-month mark. in most cases, however, the months during which the beneficiary is drawing cash benefits will be counted against the 5—year lifetime limit. “‘ if a mother gives birth while on welfare, the infant exemption for this child does not apply. support to families State guarantee of child care subsidy after leaving cash aid Maximum family income before losing child care subsidy Child care copayment for single mother with two children earning $12,000 per year California‘ For 2 years2 300% of poverty line $0 monthly for full—time care STATES AND NElGHBGRHOGDS Table 3.2 Contrasting child care subsidy rules, supports, and work incentives COIIIIECtiCIIt No time limit for subsidy3 300% of poverty line Between $5-$315 monthly for full-time cared Florida For 2 years 200% of poverty line $48 monthly for full-time care support to organizations Maximum reimbursement to centers for preschool—age care Child care information and strength of resource and referral agencies Between 0.9 and 1.5 standard devia- tions of market rate ($450—$600) Mandate of co-location by R&R staff in welfare offices Up to $325 monthly5 Weak locally, central 800 phone number $310 monthly (statewide estimate)6 Strong locally, co—location of R&R staff in welfare offices SOUPCES San Frantisco Department of Human Services. 1998. San Fra/7C/SCO County Ca/WORKS Plan (Submitted January 7). San Frantisto Santa Clara County Social Services Agency. 1997 Employment Support lnit/‘at/ve Act/on P/an { May 16, 1997i San Jose Florida Legislature. 1996, WAGES-A Plan to Refer/77 lll'e/fare in F/or/da l Conference Committee on Welfare Peform, CSSB 1662i Tallahassee. Government Acmuntmg Office 1998. We/fare reform: States efforts 8 expand r/ii/dcal’e programs Washington DC: 0A0 lHEHS-98—27l Piiles rliffc-r slightly between San Francisco and Santa Clara counties We specify the least restrictive provisions. Crowded the partitipant's income does not exceed 752/7 of the state median income. ‘7 Experimental group must apply within 6 months after leaving cash assistance. L For experimental group, co—payments ranged between 2% and 10% of gross income. 2 Equals $2.70 per hour for 28 hours of care weekly. $435 maximum for special needs children. 1 Long, S, Kirby, 0., Kurka, R, Waters, S. 1998. Chi/d care assistance under we/fare reform: Ear/y responses by the states, Washington 00: Urban Institute. BERKELEY _ YALE SECTION 3 Child Care Support Differing levels of child care support also have emerged among the states, as governors and legisla— tors implement their particular renditions of welfare reform. All three states now allow former parents to retain child care subsidies for up to two years after their cash assistance ends (Table 3.2). Connecticut has no time limit on receipt of child care subsidies. Welfare clients may earn an income that is about twice the federal poverty level in Florida, and almost three times the poverty line in California and Connecticut (75% of the state median income), before losing their child care subsidy. Required levels of co—payments for child care, paid by mothers out of pocket, differ across the states. The level is relatively high in Florida—$48 monthly for full—time care, to no or nominal co—payments in California. In Connecticut, women on welfare pay between $5 and $315 per month in co—payments, depending upon their income. jobs First clients typically are required to pay between 2% and 10% of their gross income in co—payments. k. Most telling, however, is not what benefits look like on paper but what share of women actually take-up their child care subsidy. Our data from Connecticut. for example, reveal that less than one in seven of all mothers actually draw their child care voucher, worth up to $13,000 annually. Certain cities, such as, Tampa and San Francisco stand out, with subsidy utilization at one in every two welfare parents. Section 8 details the disparate ways in which ambi— tious child care subsidy programs are implemented by states and counties. The dollar reimbursements that flow to child care organizations and individual caregivers, set by state and county governments, also vary widely. This issue is crucial since it determines the level of purchasing power exercised by parents and incentives for providers to remain in the market. The monthly reimbursement rate to providers in urban California counties is almost twice that allowed in comparable areas of Florida. a GROWING UP IN POVERTY PROJECT A The strength of local child care resource and referral (R&R) agencies varies across the five Project cities. Connecticut does not support a network of commu- nity—level RSCR agencies, as do legislatures in Cali— fornia and Florida. In California, every county has at least one RESCR agency with community—level offices. In Florida, a similar network is funded by the state and local foundations. Welfare offices in Tampa house R&R staff members who provide child care counseling on site and effectively urge clients to sign-up for their child care voucher on the spot— whether they choose to use a center program or a kin member for child care. This has resulted in a relatively high take—up rate. he‘s»; .1. am". 1 . 7» {LT-a5» . Sizzling rates of economic growth in San Jose and Tampa have trickle-down effects for semi-skilled workers, including women moving From welfare to work. In Connecticut, clients who ask their welfare case— worker about child care assistance are told that they should call an 800 number to reach Infoline, an information agency serving the entire state. Our data show that very few mothers entering the welfare system have contact with any local R&R organiza— tion. For the small share ofclients in Connecticut who figure out the child care subsidy system, pay— ments are disbursed centrally by a private firm on behalf ofthe state welfare agency. Local Job Markets Overall unemployment rates remained low in 1998 across the Project cities. Iet over 7% of all black adults were jobless in the four cities for which data are available. This proportion reached 10% in San jose. And the share of Latino adults without a job in New Haven equaled 9% in the second half of 1998. STATES AND NEIGHBORHOODS Figure 3.1 Local unemployment rates bV ethnicity, 1998 100/0— 8% — 6% — 4% — Percent unemployed 2% — 0% j San Francisco San JOSE I Latino New Haven Tampa Black White/Anglo Monthly averages during second half of 1998. Excludes adults not in the labor force, that is, not looking for work. Source: Current Population Survey (CPS), US. Bureau of the Census. These numbers exclude adults who are not in the labor force, that is, those who are no longer looking for a job. ‘5 Historical levels of unemployment are important to note. In 1990, prior to the current economic boom, jobless rates among blacks were at or above 10% in all four cities, reaching 13% in San Francisco. Latino unemployment equaled 13% in New Haven and 9% in San Jose. The difference between 1990 and 1998 numbers help to explain why welfare rolls have shrunk, given the hot job market’s current ability to absorb semi-skilled workers. Furthermore, ethnic differences in employment rates may hold implications for child-level effects. These urban labor markets also vary in the kinds of jobs that are available to women with limited work experience and skills. In 1999 a higher proportion of women were employed in agriculture and construc— tion-related jobs in Tampa, for example, than in the Manchester area (18% versus 7% respectively).34 The San Francisco labor market is most robust for women entering the service sector: over 44% of all employed women are in services, compared to just 31% in San Jose where the manufacturing sector is generating more jobs. Retail and wholesale trade generates about one—fifth of all jobs for women in all five Project cities. Overall, the sizzling rates of economic growth experienced in Santa Clara County and Tampa in recent years have trickle—down effects in terms of the sustained labor demand for semi—skilled workers. Labor structures in Manchester, New Haven, and San Francisco appear to be less vibrant. Variability in Where Families Live Poor families live in quite diverse neighborhoods. The stereotype of an ethnic minority family surviv— ing in the bleak inner city is true all too often. But BERKELEY — YALE E SECTION 3 Figure A Children in poverty Connecticut. New Haven County 7 \' , -: ‘ \ Percent oi Chlldreu Ill Poverty By US Census Block Group i Fanuiius l [I] Cuuhlv Lines l j I Locallons oiParueipahng i E; Highway Source PACE 1mm Chums. 1998. US Bureau oi the Census 1090 , in many other cases, low—income parents struggle to move into a blue—collar or lower middle—class neigh— borhood, seeking safety, better schools, or greener public spaces for their children. And many women going on welfare come from middle-class back— grounds, suddenly in dire straits due to divorce, desertion, or death of a partner. To better grasp the neighborhoods in which partici— pating families live, we developed maps which show family locations against characteristics oftheir neighborhoods. Figure A, for example, introduces you to this way of illustrating community diversity. Please turn to Appendix 4 to examine detailed maps of four Project cities. The yellow dots on these maps—for San Francisco (Figure 3.2), San Jose and Santa Clara County (Figure 3.3), New Haven (Figure 3.4), Tampa and Hillsborough County (Figure 3.5) in Appendix 4— indicate the locations of participating families. The shades of red show the share ofchildren who live in families below the poverty line for each block—group. A typical census tract contains seven or eight block—groups. The concentration or disbursed character of the yellow dots indicates families housing patterns. For instance, the San Francisco map reveals that a sizeable number of families live in public housing, as indicated by the cluster of yellow dots in the Bay El GROWING up IN POVERTY PROJECT A View and Tenderloin neighborhoods. In San Jose, the bulk of participating mothers and children reside in the low—income East San Jose and Milpitas areas. Yet a significant number also live in more blue-collar and middle—class neighborhoods west of the down- town area. The Tampa map similarly illustrates the suburbanization ofpoverty, with families fanning— out north and east of the downtown area. Neighborhood Economies To better understand the diversity of neighborhoods we examined the median household income of all residents within the block—groups where participat- ing families lived in 1998—99. Participating families resided in one of 542 different block—groups across the five Project cities. These neighborhoods manifest a wide range of economic and institutional condi— tions, including wide variability in average house— hold income. Median income fot the 143 block— groups in which San Jose families resided equaled about $52,000 in 1998. In contrast, household income in the 133 Tampa block—groups was $35,500. Disparities in regional economies—the vibrant heart ofSilicon Valley versus Tampas southern price structure—make a substantial difference in shaping the neighborhoods in which poor families live.5R Participating families, predictably, were heavily concentrated in poorer block-groups and census tracts. For example, 60% ofparticipating Connecti— cut families lived in just one—fourth of the census tracts. The remaining 40% of the sample was distributed across the other three-fourths. Figure 3.6 displays the median household income for the one— quarter of “high concentration" census tracts. For instance, the portion of the California sample concentrated in one—fourth ofthe full sample‘s tracts lived in communities with an estimated median income of$38,317 in 1998. In 'I‘ampa, the high concentration tracts hosted families with a median income of $23,28 1 .3“ If we focus on the tracts at the 25th percentile, we see that large portions of sampled families predictably live in quite poor STATES AND NEIGHBORHOODS Figure 3.6 Families live in diverse neighborhoods: Income levels of census tracts with most participating families $40,000 $38,317 $30,000 " “$15595“ : .] $20,000 $10,000 $0 California - Median tract $22375 Connecticut Florida Tract at 25th percentile Figures are for the one—quarter of census tracts with the highest concentration of participating families. Source: CUP Project, US. Bureau of the Census and Claritas 1998 income estimates. communities. Yet the overall finding remains: participating families reside in a range of neighbor— hoods, contrasting the conventional imagery of deeply depressed ghettos and barrios. Ethnic Neighborhoods Racial segregation marks many poor communities in America. just as it characterizes middle—class sub— urbs. This pattern ofethnic separation appears for a subset of communities in which our families live. Looking at the maps in Appendix 4, we can see concentrations of sampled families is quite distinct ethnic neighborhoods. Many Latino families, for instance. live on the south side of New Haven. In San Francisco. there are Latino families clustered in the Mission District and black households in the Hunters Point and Bayview districts. But many participating families are spread across a wider range Ofmore racially integrated communities. Figure 3.7 illustrates the range of ethnic communi- ties in which participating mothers and children reside. The New Haven bar, for instance reports on the percentage of families’ that reside in predomi— nantly black, Latino, and Anglo block-groups. Note that the majority of block—groups in which partici— pating families live are predominantly Anglo, not black or Latino. That is, a majority of block group residents are Anglo. San Jose is the exception, where families are concentrated in predominantly Latino neighborhoods. Note that this analysis illustrates the range of neighborhoods across which all participat— ing families reside, not the concentration of families who live within the poorer subset of neighborhoods. The ethnic composition of neighborhoods holds at least two implications. First, we know that ethnic groups, in general, differ in the extent to which they engage in reading practices and other forms of parenting that are related to children’s early learning, BERKELEY - YALE EB SECTION 3 Figure 3.7 Families live in diverse neighborhoods: Largest ethnic group in all blOCk groups (305) or participating families 1 00% :’ I ‘ r * f 65% E 60% l ’ l l 50% 0% 5 °/o New Haven Tampa Share of- 805, White/ Anglo Share of 305, Asian - Share of 305, Black I Share of 805, Latino San JOSE Sources: US. Bureau of the Census and Claritas ethnic composition estimates for ”I998. as discussed in Section 2. This is confounded with parents social-class position and their own educa— tion levels. And clearly there is wide variability in parenting practices within any particular ethnic group. Second. evidence from earlier studies reveals that the availability orchild care centers and preschools varies systematically among dirl‘erent ethnic communities. even after accounting for maternal employment rates and other determinants. For instance. the per capita supply orcenter-based programs is markedly lower in Latino communities as compared with those in low-income African—American or Anglo neighborhoods.“V Taken together, these differences suggest the pres— ence of either (a) differing norms about child rearing and maternal roles within these communities. and thus variation in the levels ol‘expressed demand for child care organizations. or (b) uneven capacity or communities to obtain necessary funding for family— support organizations. More work is required to understand how norms and organizational conditions GROWING UP IN POVERTY PROJECT A among ethnic communities may affect mothers“ lives and the development or their children. At the same time, it‘ a significant portion or families reside in economically diverse communities, this opens up avenues for strengthening family—support organiza— tions. including child care. Local Child Care and Preschool Organizations The ability of mothers with preschool—age children to move from welfare to work depends heavily upon the local availability otchild care. Our community analysis discovered that the per capita availability or licensed child care and preschool organizations varies dramatically among the five cities and across neighlmrhoods within cities. This exercise, more broadly, provides a feel for the organizational infra— structure round in these contrasting neighborhoods. liigure 3.8 contrasts per capita availability orchild care slots among three Project cities. The left—hand cluster of bars reports on how many enrollment slots existed within centers or preschools per 100 preschool-age children in 1998. In Tampa, for instance, 42 slots existed (most were filled by children) per 100 young children in the median zip code, averaging across all zips in which participating children lived. In contrast, San Francisco centers operated 35 slots and San Jose, 11 slots per capita. The right-hand cluster ofbars shows that FCCH‘s partially offset this gap in center slots. The highest per capita availability of FCCH slots is observed in San Jose, 12 per capita, compared to 5 slots in Tampa. Another way to examine supply inequalities is to determine how many centers or FCCHs are situated close to participating families. We geocoded the location of every licensed child care organization in the participating cities. then determined how many centers and FCCHs were located within a one—mile radius of each participating family. Figure 3.9 reports these counts for the median family in each of three cities. This method adjusts for the concentration of ‘3 F? i E *3 ill, N L N LE l 9:4 lei; BO R H {33 B 5 families in a subset ofall zip codes, rather than treating each zip code equally in the estimation. We see that the availability of nearby centers remains highest for Tampa mothers and children. For ex— ample, after arraying the number of nearby centers for all Tampa families, from the highest count to the lowest, the median mother lived within one mile of 8 different centers. The median mother in San Jose lived within a mile of6 centers. This suggests that overall supply is lower across the wide range of zip codes in which participating families reside. Centers are effectively located in the zip codes in which welfare—poor families are concentrated. Center availability is very constrained in the main zips in which Connecticut families reside. This analysis was run combining New Haven and Manchester, revealing that three centers are located within one mile of the median mother’s home, about half the level of availability observed in Tampa and San Jose. Figure 3.8 Center and preschool supply is higher in Tampa, while family child care homes partially fill the gap 50— 40* 30- 20* 10- 11 Chlld SlOtS per 100 children 0-5 years Old Center supply I Tampa Child slots per ’IOO children for the median zip code among zips in which Project families reside. San Francisco 12 Family child care home supply San JOSE BERKELEY c YALE 37 SECTION 3 Figure 3.9 Tampa and San JOSE families live close to more centers and preschools 20 _\ U1 Count of centers or homes within one mile of median family _\ U1 0 Tampa I Centers and preschools ji177 l San Jose New Haven Family child care homes Ceocoded data for CUP families and child care organizations. Again. FCCHs partly compensate for disparities in center supply. The median New Haven/l\lanchester mother lives within a mile of 13 FCCHs. This figure is 1—7 for the median family in San Jose. and just 6 for the median Tampa mother. ’l‘ampa is populated by many centers. a fair share of which are for—profit centers ol‘low quality. as we document in Section 8. Summary The opportunities taken up by single mothers are determined by both their personal resources (Ill/Z, the local availability of jobs and family-support organi— zations. You saw in Section 2, for instance. how the Connecticut sample contains the best educated \VOIIIC’IL I‘C‘l'dl'iVC [0 [lIC OIlICI’ [\\'0 SILIIC samples. BUT their propensity to place their children in centers. or in higher quality care of any kind. is lowest. This is surprising in that prior research consistently shows that more highly educated mothers display a E] GROWING UP IN POVERTY PROJECT A greater likelihood ol‘selecting center—based care. But it‘s not a surprising outcome once you see the relatively low availability ol‘licensed child care organizations in poor Connecticut communities. Again. women‘s potential to build from their per— sonal resources—to find jobs or quality child care— are enhanced or constrained by the local context. As we move next to reporting on the actions taken by these women in various aspects of‘lite—employ- ment. finding health services, and making child care choices—keep in mind three ingredients or their local contexts: 53$ (ilmnp/mezr mint )‘t’HliI/Ht’fl] quire 1011' in 1998. Ii'ir/J f/Jt’ important except/0H (if/'0[2/85371853‘ among [)AU‘A’ tin/(MIT. \V’e can not discern why African-American women would be less able to benefit from widespread employment growth. But it does suggest that those participating cities with more heavily black caseloads—San Francisco and New Haven—may see less success in transitioning clients from welfare to work. High joblessness among Latinos in New Haven may further suppress the rate at which New Haven women move into the work force. I State and local policies set the incentives associated with staying on or leaving the welfizre rolls. Tampa’s low level of cash assistance and ample supply of center—based child care may encourage exit from welfare more strongly than in other states that display higher benefit levels and a scarce supply of child care, most notably Con— necticut. Another way to put this: If Connecti— cut had a richer supply of quality child care, more women might be able to hold-down a job. Q The variability in where families live prompts all kinds of questions about the availability of jobs and child care nearby. Are welfare clients who live in blue—collar communities more likely to leave the rolls when richer supplies of organized Child care are present? Do housing projects or concentrations oflow— income families constrain welfare-to—work efforts, relative to economically diverse neighborhoods? One way to look at the mothers and child’s immedi— ate context is to focus on their households, women’s social supports, and ways in which mothers interact with their young child. This is domain on which the HCXI section f0CUS€S. STATES AND NEIGHBORHOODS BERKELEY — YALE m SECTiON 3 m GROWING UP IN POVERTY PROJECT A SECTION 4 HGMES AND PARENNNG III This Section How well do families function? From whom do women gain social and economic support? How do mothers' parenting and early learning practices vary? Home Life: Supportive or Stressful? Almost all parents depend on others for social and economic support. This has never been more true. At century‘s end over two—thirds of all mothers with a preschool—age child were working at least part- time. up from just 15% in the 19505. So. we must call on kin members or friends to help with rides, to baby—sit our children. to lend a hand when were between jobs. For poor families—especially those headed by single mothers—these economic supports can be tenuous or absent. Simple interactions with household members. from boyfriends to boarders, can be marked by discord and stress. This section begins by sketching the social and emotional context of mothers’ daily lives. Most benefit from strong support networks and spend time with at least one other adult who aids in the child—rearing process. The majority of mothers enjoy the task of raising a child and feel quite efficacious 3.5 a parent. But many others—about one—fourth of all partici— pating mothers~——live alone with no other adults, report few social ties, and are disappointed either with their toddler’s characteristics or believe that M m breastfeeding her, and it's hard to get up in the morning, get the breast i'nilk ready, get the kids up, tier dressed. m: least by a vear old, i can put her on straight milk But l wish they could wait a veal”, until she is a year old.” Gloria they are not an effective parent. This latter group suffers from high levels ofisolation and often from severe depression. Decades of earlier research has detailed how maternal depression erodes the vitality of home settings and undercuts young children’s early development.58 One reason that employment for some low—income women may yield positive child effects is that it can nurture in the mother a sense of efficacy, new friendships, and a fresh feeling ofopportunity.39 This likely depends on the nature of the job and the accompanying demands associated with getting to work, including finding reliable child care. BERKELEY — YALE EII SECTION 4 Our specific aim in this section is to understand the mother‘s social context. her sources ofsupport, her feelings ofefficacy in raising a young child, and her positive parenting practices. These dynamics offer a foundation upon which getting a job may yield additional motivational benefits. But when these pillars are crumbling at home, one must question whether simply getting a minimum wage job or being “sanctioned” by the welfare office will alone alter a single mothers behavior. Jhe Houseeoldis Sedal Structure uh :>l‘t,‘ 30% o\° ’l 6% 00/0 . 20% 1 00/0 A. ’ .. 1 00/ ““‘T‘I“ i‘ 0‘ 1/539} 0% . Experimental Control group group n = 151 n = 138 Alcohol/drug problem in the household Connecticut mothers in the experimental group were significantly more likely to have someone in the household with an alcohol or drug problem, compared to control group. 1 7% Fights are common in We often criticize each Electricity/phone cut Alcohol/drug problem household other I California n = 412 Connecticut n = 288 in past year in household Florida n = 196 Mothers in Connecticut were less likely to report that fights in the household are common, that members often criticize each other, and that someone living in the household had an alcohol or drug problem, compared to mothers in California and Florida Mothers in Florida were significantly more likely to have had their electricity and/or phone service cut in the past year, compared to mothers in California and Connecticut. BERKELEY , YALE m we - ‘4“ Q! . :SS Eli Figure 4.6 Frequency 0f mother’s reading 120 the focal Child 1000/ E ° g 67% 50% 24% 00/0 California Connecticut n = 41 5 n = 289 Reading frequency differs significantly by state. median California and Florida mothers who reported having between 3 and 9 books For their child. \Vomen in the Connecticut experimental group reported slightly more books for their preschooler. compared to the control group. About 85% of all mothers across the three state samples said that they sing. tell stories. or play games with their young child “most days“ of the week. It‘s important to note that most women reported a great deal of interaction with their preschool—age child. although this contact does not always include reading or other actions intended to promote early learning and development. Educational Outings Between 3300 and 40% of the mothers indicated that they took their child to the library during the previous month, with the incidence slightly higher among California mothers. All mothers were less likely to have taken their child to a museum during the previous month. GROWING up IN POVERTY PROJECT A 41 % Most days 1-2 times per week I Rarely Florida n = 200 Television Viewing The incidence of television viewing among very young children is very high. liarlier research has shown negative child development effects stemming from watching TV, in part because it detracts from more stimulating activities between parent and child. *“ Figure 4.7 reports on the median number of‘hours young children reported watching television during weekdays. Two to three hours each weekday was consistently reported by mothers in all three states. \We also asked how many hours the TV was turned on during the weekend. In lilorida. the median mother reported six hours per day (not shown in figure). The question was asked a bit differently in Connecticut. but the levels of viewing are similar. HOMES AND PARENTING Figure 4.7 Hours Of tEIEVlSlOI'I VlEWEd bV focal Children by state i 3 l l a 'o 2 L a) O. K’ 3 O .C C .9 8 1 2 0 California Connecticut n = 414 n = 284 Summary: The Social Quality b__ of Life Inside Homes The homes oflow—income mothers are complex. These numbers provide a few pieces to this compli— cated puzzle. But they do offer a feel for the quality of social life and the extent to which these women engage in early learning activities. We have selected factors that earlier research has shown are related to positive development and children’s initial perfor- mance in school, such as early and frequent reading to one’s young child. Which findings are most important to keep in mind? Whether a mother entering welfare lives with at least one other adult may make a sizable differ— ence for her economic situation, levels of social support and mental health, and time available to be with her children. The sharply varying likelihoods between our state samples—67% live at least one other adult in California versus just 28% in Connecticut—may hold telling implications. And when one—fifth of all women Florida n = 200 in the Florida sample report living with an adult who suffers from alcohol or drug abuse, the demands on and lack of support felt by many women become even clearer. In sum, living with others does not equate to feeling more support, but it may be a necessary condition for gaining crucial aid from family or friends. The large fraction of women—about one— fourth—who appear to live in social isolation, only with their children, is cause for alarm. Our future work will examine whether this subgroup experiences other negative outcomes, including less work experience, higher maternal depres- sion, and less effective child—rearing practices. Between one—fourth and one—third of all women report a very low propensity to engage in activities that nurture development of their young children. In California, almost one—third of the mothers reported reading to their toddler twice a month or less; this share ofwomen was 27% in Florida. The better educated Connecticut mothers read more to their youngsters. From our earlier paper, BERKELEY - YALE SECTION 4 we know that the mothers with stronger work experience actually read more to their young— sters, compared to those who have spent more months on welfare. But this effect disappears after taking into account maternal education levels, as seen with the Connecticut mothers.‘H Higher school attainment leads to all sorts of positive outcomes for mother and child alike. Next we turn to how these women survive economi— cally and discuss risk factors that arise for young children when money is in short supply. Together, both social and economic supports contribute to a mother‘s capacity to provide a stable and nurturing environment for her child. an GROWING UP IN POVERTY PROJECT A SECTION 5 In This SECtiOII How do women get by economically? is What wage levels have women experienced recently? it How many women have trouble putting food on the table? Getting BVW'th Welfareor W9rk. . -:~: ’VK‘ST:5WEE:SYPWM$Y9§L\. '. . You have seen how these women vary in important ways. Some arrived at the front door of new welfare programs with ample work experience and adequate levels of schooling. Others have spent much of their adult lives drawing cash assistance. These prior experiences—and women‘s underlying levels of education, emotional resources, and social sup- ports—likely condition the extent to which new welfare rules actually change their daily lives. Policy thrusts interact with individual circumstances to shape the well-being of mother and child alike. In this section, we focus on the crucial area of employ— ment and income. We report on typical levels of their economic activity, and continue to show how these women vary widely. For the Connecticut sample we also can begin to assess welfare reform’s impact on women‘s propensity to move into jobs, and what effect this has on their earnings. MDRC’S accompanying report provides additional details for the larger sample of Connecticut families, beyond our subsample of mothers with preschool—age children.“ Working Women’s recent employment behavior varied substantially among the three project states. Figure FAMILY ECONOMY II was told that they're only going to pay for child care for 12 months. There's a one year limit. All three of my kids are in child care. That would take more than half my income to pay. I'm going to have to stay home with my kids because there's no way I can aFFord to pay child care." — Regina 5.1 displays the percentage of women who reported that they worked for pay for any period during the 12 months prior to our interview. Among the Florida women, 84% were employed at some time during the previous year, compared to just 46% for partici- pating women in both California and Connecticut. These differing levels of prior work experience can be explained, in part, by the fact that welfare reform in Florida was well underway by the summer of 1998 when our maternal interviews began. Thus we BERKELEY — YALE m ‘EéECTiON 3 Figure 5.1 Share of mothers working at any point during the prior 12 months '1 2 100% l . g i 2 m (D f; 50% c} 00/0 California Connecticut Florida n=415 n=291 n=2OO Between—state differences are statistically significant Connecticut data prior to random assignment. Figure 5.2 ' HOCUI'IV wages for working mothers $8 ‘ . $7.78 """" ' $7.24 $6.36 7 co $5.45 0 N 3 2 $4 5 0 I $0 California Connecticut Florida n=154 n=105 n=192 I mean median Hourly wages for mothers in Florida are significantly lower than for mothers in California and Connecticut. Average wages in Connecticut do not differ significantly by experimental and control groups. Wages were not adjusted for between-state differences in prices or purchasing power. a ceowuyc UP IN POVERTY PROJECT A would expect to be sampling a mix ofshort—term and long-term clients. Yet Connecticut’s jobs First program was 18 months into implementation at the time ofour interviews, and recent employment rates were relatively low. We will see below how the shares of women in the California and Connecticut samples who were long-term welfare recipients is higher, relative to Florida. Caution is warranted in comparing the Connecticut findings—some of which stem from the period prior to the mothers being randomly assigned to the experimental or control group—to the California and Florida samples where work experience data relate to the year prior to our interviews. \Vithin Connecticut, 56% of the women in the experimental—group were working when interviewed, compared to 4100 ofthose in the control group, a statistically significant advantage (p<.01, from Project interviews). Under the new rules ofthe Jobs First program, mothers in the experimental group faced more intense pressure to move from welfare to Figure 5.3 Difficulty buying enough food 80/0 68% California n = 41 1 Often 24% sometimes work than the control group. The latter lived under the old rules ofAFDC, facing less intense pressure from the system to find a job. However, four in 10 women in the control group, however, had success— fully found employment within 18 months after entry. wages While a large number of women in all three state samples were employed at some time during the past year, their jobs paid low wages. Figure 5.2 illustrates the hourly wage rates for women‘s most recent jobs. The median woman in Florida, who was 32 in 1998, earned $5.45 per hour, or a reported monthly income of $630. Average hourly wages were higher in California ($6.36) and Connecticut ($7.24). Even so, this resulted in median monthly earnings ofjust $700 and $799, respectively. Note the differing cost— of—living levels experienced by families among the Project cities (reported in Section 3). These low wage levels suggest that mothers earn little discretionary income, available for savings or Florida n = 200 I Never Differences between states are not significant, BERKELEY ~ YALE SECTION 5 for services that are essential to remaining employed, such as sufficient income to pay for child care. Hunger A mothers economic resources obviously hold implications for the well—being of her young child. For example, Figure 5.3 reports on mothers‘ percep— tions ofwhether they had enough money to buy food for themselves and their children. In respond— ing to the statement, “The food we bought (in the prior year) wasnt enough and we did 1ft have money to buy more,“ 24% ofthe California mothers responded that this fit their situation “some of the time,“ and another 8% said it was “often“ true. The proportions for Florida women were 22% some of the time and 6% often. To verify the accuracy ofthe first responses, we returned to the topic later, asking the women whether they could afford to buy “enough and the kinds of food (you) want.” In California, 45% of the mothers said yes, and 33% of the Florida mothers responded affirmatively. Figure 5.4 shows that levels of food rationing and hunger are almost three times greater among partici— pating families, compared to national averages. We compared the responses to two interview questions that were taken from periodic hunger surveys conducted by the U.S. Department of Agriculture (USDA). For instance, on the question, “The food we bought wasn’t enough and we didn’t have money to buy more,“ a total of3l% of the participating mothers responded “often“ or “sometimes,” compared to 1 100 among the general population nationwide. On another question, “I relied on low—cost food to feed my child(ren) because I was running out of money to buy food.“ 32% of our participating mothers said “often" or “sometimes,“ while nation— wide the figure is 14%. Appendix 2 includes further details and data sources. Figure 5.4 Hunger lEVElS compared to national norms 40°/o~ 30%— 20%* 1 00/0— 1 1 0/0 0/0 sometimes 0|” often, DFTOF 1 2 months 00/0 j Ran out of money for Food - Participating CUP families 1 40/0 Relied on cheap food for children National average Special thanks to Gary Bickel, USDA measurement expert, For reporting item scores from national norming survey. M. Nord, K. Jamison, and G. Bickel, "Prevalence of Food Insecurity and Hunger by State, 1996—1998." (Alexandria, VA: Office of Analysis, USDA, 1999). See Appendix 2 E] GROWING up IN POVERTY PROJECT A Figure 5.5 Employed co-resident adults / 100% - , , ~ 1/ ‘l 40% i ‘i l l 1 50% 0% California Connecticut n = 409 n = 295 Household Members Who Help to Support Children Beyond a mother‘s earned income, two additional sources of income may aid mother and child alike. First, other members ofthe household may bring home a paycheck. Second, the mother and co— residents of her household may be drawing addi— tional family—support benefits, including food stamps, housing subsidies, or cash from the Earned Income Tax Credit (EITC), a refundable income supplement for the working poor. \When the mother lived with an employed adult, we asked the mother whether this income helped to support her child. Figure 5.5 first shows that sizable percentages of women live in a household with at least one working adult, ranging from 25% among Connecticut women to 40% in California. \We then asked whether the co—resident(s) provided economic support for the focal child. The specific question was: “Does the income received by other members ofyour household help to support you and 'i M i Li." E C G hi C; M employed co-resident adun non-working co-resident adult no other I adult in the household Florida n = 200 your child?” Note that mothers could have included cash income from family—support programs, not only earned income from their co—residents’ jobs. Participation in Family Support Programs The mothers themselves could certainly draw benefits from other programs, beyond the earned income from their job. Figure 5.6, for example, shows that over 90% of participating mothers in California and Florida participated in the food St’JIT‘lp program. In contrast, only about 1 in 6 women received child support payments from fathers. And only a fraction drew income from Supplemental Security Income (SSI), typically for a disability that they or their child experience. This latter finding is partially attributable to the fact that many applicants for TANF cash assistance who have known disabilities are diverted out ofwelfare-to—work programs. BERKELEY ~ YALE E sit: i’i ii‘tl Figure 5.5 Participation in other family support programs 1000/0— ; 95% U3 0) >‘ 50%— 09 16% m I OO/Oj Food Stamps Child Support SSI I California n = 411 Florida n = 196 Participation rates for other family support programs do not differ significantly by state. Overall, keep in mind these high rates ofparticipa— tion in Medicaid and food stamps. In Section 8 we discuss comparatively low rates of participation in the newer child care subsidy program. Summary;Lixing.,9nwtfle.,.Edge . Two effects of welfare reform have prompted celebrations in some policy circles: client caseloads have fallen dramatically, and the share of‘single mothers who still enter the welfare system and then successfully find a job is higher, due in part to new welfare rules. Women participating in new welfare programs where true experiments are being run, are between 8% and 15% more likely to be employed. compared to women in control groups (depending on the state and the study)."“ Both events demon— strate that strong action by government can influ— ence the actions of low—income women. m GROWING up IN POVERTY PROJECT A A deeper worry is whether getting off welfare and into low—wage jobs will discernibly boost the mothers quality oflife. If net income rises to higher, sustainable levels, compared to being on welfare, then a mothers future may look brighter. Yet this assumes that the costs associated with staying employed, such as child care costs, do not reduce the mothers discretionary income. \We are beginning to see some women approach a cliff, over which they fall, when they hit the time limit on child care aid or their rising income makes them no longer eligible. The second issue is whether young children benefit from their mother being away from home and at work. That is, the mothers economic and social well—being could rise, due to employment and the social engagement that work may bring, but the environments in which children are placed may hold a detrimental effect on their child‘s early develop— ment and learning. lhis could lead to a net decrease in the faiiiilys quality of life, balancing gains for the mother against losses experienced by the child. in Section 8 we return to the question of whether young children are now spending their days in stimulating or stultifying settings. The point here is that mothers’ recent wage rates and added economic returns—in these initial months under new welfare rules—are modest at best. This is a key finding among the results presented in the section: i Just under halfof all women in California and Connecticut had worked for any amount of time in the 12 months prior to our interview. In Florida. this share was about 80%, given that these women were older and the job market was more favorable. The extent to which welfare reforms have boosted this employment rate for clients who remain in the system is an important question. The experimental results in Connecti- cut suggest a significant level of success: The women assigned to the experimental (Jobs First) group were significantly more likely to be em— ployed when interviewed (56%), compared to the control group (41%). I When women found a job it typically paid a very low wage, between $5.45 and $7.24 hourly on average among the three project states. These wage levels are consistent with other recent evaluations (Section 1). Overall, the jury is still out as to whether “work pays,” that is, whether going off cash assistance brings higher earnings, and whether these earnings can be sustained after work supports such as child care subsidies are no longer available to women who have moved off welfare. I Many mothers sometimes have trouble even affording enough food to feed their children. We saw how 24% of California and Florida mothers reported that they didn’t have enough food in their cupboards and they “didn’t have money to buy more” (during the previous 12 months). I At most only a third of the mothers lived with another adult who shared his or her income in ways that benefited the focal child (Florida). But FAMILY ECONOMY in the other two states less than one in five women lived with an adult who pooled income in this way. A significant share of women, across all three state samples, lives with at least one other adult. But this doesn’t necessarily bring income that directly benefits the child. Next we turn to the area of health. How do mothers view their physical and mental health and utilize health insurance and clinical services? How do motivation and perseverance, or debilitating depres— sion, contribute to employability and the quality of parenting that unfolds inside the home? BERKELEY — YALE SECTION S 33 GROWING UP IN POVERTY PROJECT A SECTION 6 III This SECtiOII What gaps exist in securing health services? m How do mothers assess their health and their child's? m How well are women coping emotionally? “ea'thVL'Ves’- The health status of women and children—includ- ing their physical and emotional vitality—is an area about which little is known. It’s a crucial domain for those interested in how the development of children may be influenced by welfare reform. Concerns have arisen. once again, about the hardest- to—employ welfare clients, those suffering from the greatest barriers to employment, ranging from substance abuse problems to deep feelings of alienation or clinical depression. Many of these factors that act to constrain the mother’s employ- ability also will slow her child’s cognitive and social development. We do know that access to some health services, such as enrollment in subsidized health plans, has fallen in the wake of welfare reform. One recent survey found that only 47% of former welfare clients were enrolled in Medicaid, even though many more remained eligible.*_ The proportion of eligible families using food stamps also declined since 1996 as welfare rules tightened. The recent detachment of allied health benefits from TANF cash assistance may play a role, as well as the general perception that access to all family supports has become more restricted. MATERNAL AND CHILD HEALTH HEALTH U need a root canal. There's a hole in this one here. Sometimes the pain is bad... I have to go to work with pain and taking aspirins. You have to get permission (with Medicaid) to go there and permission to do this. They still haven't worked on my teeth." —- Patricia Mental health is one particular factor that may powerfully limit the mother’s employability and her child’s early development. The companion Yale evaluation, led by our public health colleague Sarah Horwitz, reveals high rates of maternal depression among her sample of Connecticut mothers with older children.48 We supplement Professor Horwitz’s findings by reporting on depression among partici— pating mothers with preschool—age children, a rate of incidence which proved to be higher than among mothers with older children. This section first reports on the rate at which women are enrolled in Medicaid or private insurance programs. Then we turn to mothers’ perceptions of BERKELEY — YALE E SECTlON 6 their own physical health and that of their child, and assistance (one can now retain health benefits and assessed levels of. the women’s mental health. leave the TANF system) discourage Medicaid involvement. But in Connecticut the experimental— group mothers were 10% more likely to be enrolled Physical Health and in Medicaid, compared to the control group (p<.03). Access to Health Services This may be due to more extensive counseling C about allied benefits conducted by Jobs First case— . workers. Yet in Section 8 we will see how this does Gaps In Health Insurance . . . not carry over to effectively encouraging use of child We begin by looking at women‘s participation in care subsidies, Medicaid. the federal health insurance program for . . . I . . . I . . . . . . The a in Medicaid artic1)ation in Florida ma be g P P Y low—income families. Figure 6.1 reveals a high rate of . , . . . . . . . t 0 . . a partly due to higher rates of maternal employment, partiCipation in California ()5 /0), but Significantly . . . . . . relative to women in the other two states (Section 2). lower enrollments in Connecticut and Florida (both I . O , at 79%). Prior welfare experience is relatively high in In Florida, 10 /0 ofsampled women reported that . . . the were currentlv covered under an em lo er— the California sample, and may be related to the y ’ P y . . . . based insurance )lan. But since thev were re-enter— higher enrollment rate in Medicaid. Why the l ’ . . . . . ing welfare, it‘s likelv that thev had lost the job and enrollment rate lags behind in Connecticut is ' ’ . . . their plan would expire. interview data from Wave 2 unclear, given that this sample of women also had . . . will reveal stronger enrollment rates for Medicaid as cons1derable welfare experience. “ women in Florida get up to speed on allied benefits. One might argue that new welfare reform rules and . - . . . . . . - SliOhth higher rates of‘children were covered bV the de-linkage of Medicaid eligibility from cash b . ' . t‘ . ’ Medicaid, as reported by the mothers. Again, Figure 6.1" Mothers; participation in Medicaid 100% "o o i 2 79 /o l l '6 L 5 100% 8 Q’ 0 80% .C 50 /o 4..) o 60% E ‘5 40% O a 20% 0% Experimental Control n = 144 n = 151 0% In Connecticut, mothers in the experimental California Connecticut Florida group are Significantly more likely than these in n = 412 n = 295 n = 200 the control group to partitipate in Medicaid E GROWING UP IN POVERTY PROJECT A Figure 5.2 Children’s participation in Medicaid MATERNAL AND CHILD HEALTH 100% l l l g l S: a.) C 8 500/0 2 E U u. 0 § 00/0 California Connecticut Florida n=415 n=285 n=200 Children's Medicaid coverage differs significantly by state. Florida shows the lowest participation rate at 84% of all focal children, rising to 93% among participat— ing California children. Mothers' Physical Health We asked women to judge their own physical health and that of their preschool—age child. Overall, the average woman rated her health as between “good” and “very good.” This set of questions is commonly used in epidemiological studies and is highly predictive of other health outcomes, including mortality rates and the incidence of mental health problems.w Differences in reported levels of physical health are apparent among women in our three state samples (Figure 6.3). For example, women in Connecticut— who are younger and better educated on average— rated their physical health as stronger than did women in California and Florida. The median participant in California rated her health as “good” (an average score of3 on a 5—point scale). The median woman in each of the other two states indicated “very good” (an average score of 4, significantly higher at p<.001).SO Children's Physical Health We also asked women to rate the health of the focal child. In all three states, mothers rated the health of their child as being “excellent.” The mean value was highest in Connecticut. In the majority of cases the mothers viewed their child’s health as better than their own. On the other hand, many women find it difficult to secure community health services. This perception oflimited availability was greatest in Florida. Among these mothers, 23% reported that during the past year their young child was in need of medical care but did not receive it. Between 10% and 14% of all mothers on welfare reported that they were unable to find affordable dental care for their children. As a consequence, they simply didn’t take their youngster to the dentist. BERKELEY — YALE E] SECTION 6 Figure 5.3 Mothers' reports of their health status 5 2 i E g 0 California Connecticut Florida n=413 n=309 n=200 Mothers reported their health on a 5-point scale: ’I = poor, 5 = excellent Mental Health: Given the crucial importance of this domain, we Alienation and Depression assessed each woman’s emotional health and possible levels ofdepression with commonly used measures. A single mothers personal resources matter 3 lot in In California and Florida, we administered the CES- determining whether she can SUCCCSSFUH)’ iuggle D. a widely utilized set of interview questions which Child rearing and work. We sometimes forget this assess an individuals feelings ofalienation or simple fact as the debate focuses on the elements of anomie. It provides a thorough inventory of depres— the welfare system. not how differing women are sive symptoms. if present.“ responding to it. And you already have read about . ~ ~ ~ . Items are framed in positive and negative terms: the plentiful sources of stress—from alcoholic ~ ~ . . . . . . . “During the vast week I en'oved life." Or, “I housemates to feeling like one is raismg their child 3‘ l I . . . thought my life had been a failure? or “I felt alone—that can erode a mothers emotional strength. L * lonely.“ The CES-D contains a cut—off score. Those In turn. Inaternal dEPFCSSTOI] COITSlSt€ntly ledds [O a scoring above it are interpreted as exhibiting a variety ofnegatwé outcomes for children. from 1655 significant number of depressive symptoms that are secure attachment between mother and toddler. to related to other health and psychological events. highly constrained parent—child interactions that _ _ - A _ - . . - . . a, Figure 6.4 shows that 4800 of the California impede early cognitive and soc1al development. ~ ~ . - mothers and 32% ofthe Florida mothers exhibited The NICHD study of early child care recently found ’ . a SlOI‘ill’TC‘dllt incidence ofde wressive svm atoms. The that three year-olds With depressed mothers scored 3 l ' l . . . . . ~ . causes of this high rate will be the subject of Significantly lower on indicators of school readiness. L - . additional analvses. including stressors linked to verbal comprehenSion, and expresswe language. These - ~ - _ - . working. household situations, and child rearing. youngsters also displayed mOie behaVioral pioblems s s and were less cooperative with other children.” E GROWING UP IN POVERTY PROJECT $2 To set these levels in context, consider the mental health of the nation’s adults. The largest similar study that has been conducted, by researchers at the National Institute of Mental Health, found that about 20% of all adults suffer from these depressive symptoms, as assessed by the CES—D. The average score on the measure, across the state samples of women, equaled 16.9. This compares to a range of 7.5 to 9.3 among the general population, depending on age and ethnicity.S1 In Connecticut, we were able to include the CES-D questions and a second instrument, commonly known as the CIDI, to gauge the incidence of longer term and severe forms of depression. It also proved to be more effective in explaining women’s propensity to move from welfare to work and their wage rates, relative to the less discriminating CBS-D assessment.SS Figure 6.5 reveals that in Connecticut over 15% of all women displayed severe levels of clinical depres— sion. No significant difference on the CIDI was observed between the experimental and control Flgure64 iMdihér’siléfitai'ihé’éii‘h‘i lnticléiicé brag.) MATERNAL AND CHILD HEALTH groups. One can interpret this in a positive or negative light. The push on the experimental group to find a job, and their higher employment rate, has not increased average levels of depression. On the other hand, 1 in 6 of all women in the Connecticut sample suffer from clinical levels of depression. And the welfare-to-work push has not lessened this burden, which acts to limit employability and erode young children’s early development. A recent epidemiological study examined the overall rate of clinical depression across a nationally repre- sentative sample of men and women. Dr. Dan Blazer, at the Duke University medical center, found that the share of women (of ages matching our participants) displaying severe depression ranged from 3.5% for Anglos to 7.5% among Latinas. The rate of depression for young African—American women equaled 5.6% nationally. Thus, participating women displayed an incidence of depression that is twice to three times the national norm, depending on their age and ethnicity. 7:: 4751;“ Imaging u ressiveisi/mptoms (CES-D) 600/0 , ”W “"""’”““‘"7 m i E O 4-3 ‘ l 9- i E i > i m l g) 1 '5 300/0 U) 2 C). CD '0 .C a: 3 o\0 00/0 California Florida National norm n = 414 n = 1 97 Incidence does not differ significantly by state. Details on National norms appear in Appendix 2. BERKELEY — YALE E SECTION 6 summary: Troubling LEVGIS Of Mental Health One theory underlying welfare reform postulates that by rearranging economic incentives and moral obligations, single mothers will respond and move into the labor force. Yet this theory‘s foundations begin to crack once we recognize the fragile personal resources and shaky & k social supports held by a sizable share of single mothers. No example is clearer than the debilitating levels of depression displayed by 1 in 6 women in the Connecticut sample, and the common expres- sion ofless severe depression by a high number of women in California and Connecticut. To sum up our specific findings on maternal and child health: We have seen how levels ofsevere maternal depression in Connecticut—which compare mothers to norms set by an earlier national are two to three times the norming studV c . Figure 5.5 Mothers' mental health in Connecticut: 30% C .9 $ 8 Q. g 20% E U E U .C 4.) § 10% a" 0% Experimental Control n = 149 n = 1 59 Details on National norm appear in Appendix 2. m GROWING UP IN POVERTY PROJECT A national rate. Lower levels of depressive symptoms are widespread among sampled mothers in California and Florida, and over twice the rate observed in the general adult population. About 1 in 5 women in Connecticut and Florida are not enrolled in Medicaid. Connecticut’s welfare—to-work program has helped to boost participation rates. :—; Another piece of good news for Connecticut is that their participating mothers felt quite good about their physical health. In contrast, women in California and Florida—less well educated and older on average—~were not as upbeat about their health status. Next we turn to the question of what welfare reform means to these women: What elements ofwelfare- to—work programs drew the most participation? Are women grasping the messages and new rules of welfare? Is their knowledge oftime limits, work obligations, and sanctions accurate? Incidence of clinical depression (ClDl) National norm SECTION 7 III This SECtiOfl How do mothers differ in their work and welfare histories? Are they engaging core welfare-to- work services? Do they understand new welfare rules and their new obligations? Carrots and StiCkS The policy makers who crafted welfare reform—in \‘C'ashington and state capitals—believe that a combination of carrots and sticks are motivating single mothers to leave home and move into jobs. The policy engineers are banking heavily on the assumed influence of‘economic incentives to change women‘s daily behavior. In all three states, participating women were re- quired to attend a job club or independently search for work shortly after qualifying for cash assistance. The policy theory also assumes that moving women into any work setting, rather than training or further education, is the best way to integrate them into the labor force. The policy engineers are banking heavily on the assumed influence of economic incentives to change women’s behavior. 3" " , M J iiw t , EMT: "9’ " iii-raw tiltret‘e»i‘it tram” it ”thaw —l lot ”trite“ w battling tilt}; MW yeti: ti.“ Milt Mitt mm r iit‘li’r l i ll ti: 3735 i: or a job mm 5 3‘5' Wish ii new Whirl limit! on i la: HEM H This section examines the welfare histories of participating women, the extent to which they have recently engaged welfare-to-work services, and the knowledge that women hold about the new array of rules and sanctions. \Vomen’s knowledge of the new rules and whether they are engaging welfare—to—work services hold implications for how their daily schedules, and their child’s social settings, may change over time. Given that about 40% ofour entire sample had not yet selected a child care provider, it appears that many women were avoiding required work activities or were willing to take a cut in their welfare check.so A fraction were exempt from any work requirement during the Wave 1 data collection period. Others BERKELEY ~ YALE 6 SECTION 7 Figure 7.1 Months that mothers received cash assistance in past year i 12 I l 12.0 10 10.3 I— i m 1 m . > .5 8 . I— . Q. .E u, 6 .C u C O E 4 LL- 0 3t: 2 0 California I mean median had already left the rolls by the time we conducted our first interview. Each of‘these pathways may result in economic change for the mother and her household, a major question that we will study after \Wave 2 interviews have been completed. Welfare Histories Figure 7.1 reports the number of months, over the prior year. that women reported being on cash assistance. The median woman in the California sample received cash assistance for the entire year. compared to three months among Florida women. Part of this difference is the fact that existing clients in California were being re—enrolled into the new state welfare program prior to the end of 1998 legislated deadline. In contrast, Florida was in its second year ofimplementation. so a wider mix of‘short— and long—term clients were sampled (Section 2). Another factor is that black and Vietnamese—American women are disproportionately long—term clients. and comprise larger slices of the California sample. Participating women in Connecticut were asked about their prior experience on welfare at the time of random assignment (drawing from MDRC's background E GROWING UP iN POVERTY PROJECT A I 30 Florida information file). Fully 43% of all Connecticut women had been on welfare for one or more years prior to their current episode (Figure 7.2).? Of the entire MDRC sample. 58% were first-time applicants for cash assistance. Overall, the Connecticut sample falls in between the California and Florida samples in terms of prior time receiving cash assistance. Welfare—to—Work Activities Next we focus on participating Connecticut mothers, since they had 18 months to avail themselves of core activities. New entrants to welfare programs branch rather quickly into one of‘three groups: those who find a job right away and go off cash assistance; those who vigorously engage welfare—to—work activities; and those who fail to engage these services and do not seek work. Figure 7.3 reports on the percentage of women who reported that they participated for any length of time in job clubs. a selflinitiated job search. or classroom training.“ Overall iust loot) of‘the Connecticut women reported that they had partici- pated in a formal job club program. However. a significantly greater share of experimental group members had attended job clubs (23%), compared to the control group (10%, significantly different at Flgure 72 Prlor, WEIfare m,V0|Vement p<.003). The overall participation rate is worrisome in among women m COHHECtICUt that the job club inten'ention—focusing on building confidence, resumes and knowledge of the job mar— No prior welfare ket—has been a cornerstone of the effort to reorga— More than 5 involvement (38%) nize welfare services. The companion MDRC report indicates that of‘the entire Connecticut experimen— tal group‘ 6400 reported that they participated in at least one employmerit—related support program. Many women pursued classroom training, primarily at the postsecondary level‘ including community colleges‘ trade schools‘ and F81, classes. Adding up BGtWGGH 1 and all these options, we found that 490/0 of the experi— years (1 9%) 5 years (280/0) _ . . Less than 1 mental group engaged in classroom trainmg, versus year (13%) 3200 otthe control group members. This difference is mainly explained by the higher share of experimental Reported duration on cash assistance is for periods prior to random assignment. Figure 7.3 Welfare to Work activities among women in CONNECthUt % women participating 50%— 40%— 30%— 20%— 1O%# 1 0% 6% 0% m l | ' l 1 Job club Individual or self— GED or other On-the-job training directed job search classroom training I Experimental n = 149 Control n = 159 The experimental group was significantly more likely to participate in a job club and conduct an individual job search than were members of the control group. The two groups did not differ significantly for enrollment in classroom or on—the—job training. BERKELEY — YALE 67 group members who enrolled in college-level courses (17%). relative to the control group (9%, /)<.()4). It is encouraging that the former group Found time to attend postsecondary training. despite the pressure to work. Maternal education is one or the most consistent determinants of positive parenting practices and raises the likelihood that a mother will select higher quality child care. Knowledge of New Welfare Rules \Ve asked women several questions about the new welfare programs that they had entered. The aim was to assess whether they understood the new obligations and time-limits tied to cash assistance. Carrots and sticks work only it they are compre— hended. Figure 7.4 summarizes the extent to which women understood the new rules. The control group in Connecticut was excluded from this analysis. since its members are not subject to the new welfare rules. Just over 30% of all women did not understand that they could draw cash assistance for only a limited number of months under the new welfare rules. The most knowledgeable group was in Florida where just 19% didn‘t grasp the time-limited nature of cash aid. We also asked how many months they could receive cash aid over their lifetime. In California, just 52% (of those who knew there was a time limit) gave the correct answer (60 months). Another 28% said 24 months which is the number of(011511711706 months that clients can draw cash assistance. Connecticut women in the experi- mental group were very well informed: 83% under— stood the ll-month lifetime limit.w We also asked each woman, “What happens to people that hit their time limits?“ just under 50% believed that “nearly everyone will have their benefits canceled.“ Another 52% felt that “only some people“ would lose their cash assistance. Figure 7.4 Knowledge of new welfare rules: Is there a time limit on welfare? 100% ,,,,, 2%. _ 1 50/0 80/0 50% 0% California Connecticut n = 410 n = 130 (experimental group only) 16% don't l know no - V85 Florida n = 197 Mothers in Connecticut are significantly more likely to know that thei e is a time limit on welfare, compared to mothers in California and Florida. [:3 GROWlNG up IN POVERTY PROJECT A Summary: Sporadic Engagement The Wave 2 interviews with mothers will yield more extensive information on the extent to which they are meeting the best intentions of policy makers and program designers. But at this early stage, drawing in large part from Connecticut’s first 18 months of experience, these patterns are evident: fist} W/omen’s engagement in core services is uneven at best. Less than a quarter of the Connecticut experi— mental group attended a job club for any length of time during the 18-month period. Remember that we are tracking only women with a pre- school-age child. The accompanying MDRC report details involvement in core activities for the entire experimental group. But for our subset, il‘program designers see job clubs as pivotal to the welfare—to—work effort, more needs to be done to boost participation. One bright piece of news is the big/J share of women pursuing postseeona’ary education. Some advocates have expressed concern that the push on clients to move quickly into jobs is discouraging further training and investing in the long-term human capital of women. But the Connecticut findings show that the experimen— tal group actually participated more in classroom programs at the college level than members of the control group. This is particularly encourag— ing, given the consistent relationship between women‘s school attainment, positive parenting practices, and stronger child development. W/ornen’s lenawiedge of new welfare rules is fair [0 p007”. Over two—thirds of all women across the three states failed to understand that cash aid is now limited to a finite number of months. And among those who are aware of time limits, over half of this subgroup doesn’t understand what the time limit is. Although they may in time as they become familiar with the new system, this WELFARE TO WORK lack of information is worrisome. It seems directly related to poor women’s decreasing use of Medicaid, food stamps, and child care subsidies. Next we turn to a major area of concern: child care. Some see it as yet another support service to help women get into jobs. This is true. But increasingly young children are being raised by new adults as their mothers go to work and the youngsters enter new child care settings, from a quality preschool to the aunt's apartment downstairs. BERKELEY — YALE EH SECTION 7 EE GROWING up IN POVERTY PROJECT A SECTION 8 $33 3 g fill. » i «r r In This SECtiOII iii Where do young children now spend their days? Are child care subsidies off-setting the cost of care? 32 What's the quality of these new child care settings? Children Enter a New Frontier Prior to 1996, single mothers with an infant or toddler at home were rarely pushed hard by their caseworkers to leave home and find a job. Indeed the original philosophy ofstate programs that first helped wartime widows, and then poor women, was that they should stay home and out of the workplace (\(l to properly raise their children. But under the reforms of the past decade, almost all single mothers on welfare are being stiffIy nudged to look for a job. In turn, their infants and toddlers are entering a new frontier. Many critics have worried that young children are being placed in unsafe or low-quality child care settings as their mothers move from welfare to work. Proponents counter that the role model inherent in a working mother, as well as rising income levels, will enrich the lives ofchildren over time. Hopes are high in this camp that welfare reform holds current effects for mothers and brighter futures for their children. To bOOSt the odds that these high hopes will come to pass, policy makers have dramatically increased spending for child care and preschool programs. The 1996 reforms brought rising federal support and CHILD CARE II he was two... she had to stay in the bedroom by herself. I noticed she was getting bruises on places, like just below her butt. Matter of fact, she did get turned in, and i think she lost her license." m Graciela sizable state child care initiatives in urban states, like Illinois, New York, and California. The federal government now spends $1 1 billion annually in child care and early education programs, including Head Start, block grants to the states, and $1 billion in preschool funding Via the elementary and secondary education act. States put up another ()1 $4 billion annually. A major aim of our Project is to examine the types and quality of child care settings that women are finding as they move from welfare to work. In this section we report on the kinds of care that mothers BERKELEY — YALE SECTION 8 are selecting, who pays for this care, and we explore whether the quality of these settings advances or hinders child development. During our initial interview we asked women whether they were currently using a child care provider for at least 10 hours per week. We then endeavored to contact each woman a second time to learn more about her child care selection as she presumably engaged in welfare—to-work activities. Connecticut women were interviewed 18 months into the program, whereas women in California and Florida had been participating between 2 and 6 months at the time of the child care interview. Mothers' Propensity to User Child Care: :‘Jt’a'WMU' 1., 3:..ng .- ..~._ts.~,..a.~z,,,a.r _ he , Overall, 62% of the participating mothers were using a child care provider for at least 10 hours a week (569 of the full sample of 948 women). The actual rate may be a bit higher, since we lost contact with 17% of all women between the initial interview and the follow-up. The proportion using child care ranged from 64% of mothers in the Florida sample, to 62% in California, and in Connecticut, 49% among women in the control group and 60% of the experimental group. One important finding is that many women already had secured a child care provider prior to entering the new state welfare program. Since many women had a long history of being on welfare, they had benefited in terms of finding a subsidized child care or preschool program. Other mothers who had been working likely qualified for Head Start or state— funded preschool programs.“ The results that follow draw from those mothers who had selected a child care provider for at least 10 hours a week. Ev] GROWING up IN POVERTY PROJECT A Types of Child Care Selected by Mothers Figure 8.1 reports on cross—state differences in the types ofchild care that mothers are selecting. Re— member that the median focal child was two—and one—halfyears when we first interviewed the mother. Just under 40% were at least three—years-old. Child care organizations and informal caregivers can be boiled down to three types: E Center—based programs, including preschools and licensed centers, typically serving children between 2—4 years of age. 3315 Family Child (are homes (FCCH programs) most commonly run by a woman who obtains a license from the state to serve multiple children. Individual lair/,7 and kin caregivers, including a relative, friend, or paid baby—sitter. These individuals may care for more than one child without a license, operating legally or illegally, depending on state regulations. Among California mothers using child care, 29% placed their youngster in a center-based program. In contrast, just 13% of Connecticut mothers selected a center. The propensity to select a center was greatest in Florida: fully 70% chose this option. We see dramatic differences among the three state samples in the types of care selected (Figure 8.1). Among the California mothers using child care, 29% placed their youngster in a center-based program, 17% chose a family child care home (FCCH), and 54% relied on kith and kin.“ In contrast, just 13% of Connecticut mothers selected center-based care, 10% were using a FCCH. and 77% selected kith or kin as their caregiver. The Figure 8.1 TVDE Of Child care providers bV state 100% 93 ('5 U H— 0 (D Q. 5 E 500/0 C 2 E E U o\° 00/0 California Connecticut n = 262 n = 168 propensity to select a center is greatest in Florida where fully 700/0 chose this option; just 5% used FCCHs, and 25% were relying upon kith or kin. Note that these between—state disparities essentially match the supply conditions detailed in Section 3. One might argue that the Connecticut parents prefer less formal kinds of child care, given their ethnic background or favored ways of raising chil— dren. But the close mapping against the simple availability of child care organizations, that is, centers and FCCHs, indicates that supply condi— tions are contributing to selection “choices” of parents, not solely differences in apriori preferences. Young children spend a lot of time each week in their child care settings. Figure 8.2 displays the average hours that these children are with a care provider other than the mother. In all three states, youngsters spent 39 or 40 hours each week in child care. Given that many women are working just part-time, we can see that work schedules and child care schedules are not always tightly “if" “C MD “Ell“: Unregulated kith & kin care Family child care Homes Center or preschool Florida n = 140 Distribution of types of child care providers differs significantly by state. coupled. Like middle-class parents, mothers work— ing part-time often invest in full—day child care. Covering the Cost of Child Care Welfare reformers, as mentioned before, have strong faith in economic incentives and monetary supports. The argument has been that by providing stronger subsidies for child care this barrier to employment will be lowered. But let’s look at the share of mothers who actually draw their child care voucher or find a subsidized slot in a licensed organization (center or FCCH). Figure 8.3 reports the percentages of mothers who reported full or partial subsidies for their child care provider. Again, this question was asked 6 months into women’s new welfare experience in California and Florida, and in Connecticut, 18 months after entry. Connecticut stands out with a low take-up rate for any kind of child care subsidy, equaling just under 13% when combining experimental and control BERKELEY — YALE m SECTION 8 Figure 8.2 Hours Children spend in Child care 40—§ x a) (D B L (D Q. E’ 3 20— O .C C N a) E 07 California Connecticut Florida n=268 n=153 n=140 Average hours spent in child care settings does not differ significantly by state. ,5 m Figure 8.3 PercentageiOf mothers using Child care Subsidies between 6 and 18-monthsiinto welfare programs 50% ‘i l l i > .‘9 U) .Q 8 U) 250/0 .2 U) :5 o? 0% California Connecticut Florida n=225 n=153 n=118 Child care subsidy utilization rates differ significantly by state for Connecticut. GROWING UP IN POVERTY PROJECT A groups. Take-up rates in both California and Florida are higher, 48% and 50%, respectively. This includes any source of public subsidy, including entry to a subsidized center or preschool, receipt of a local scholarship for one’s child, or participation in the welfare—linked voucher system. More work needs to be done to figure out what Florida and California are doing right and why Connecticut is doing so poorly on this front. This low rate in Connecticut is likely linked to the scarce supply of center—based care in poor neighborhoods. We discovered that subsidy take—up is highly corre- lated with the selection of center-based care. It appears to be a more discrete avenue for financial assistance in the minds of caseworkers and clients alike.‘H Word is not getting out that vouchers can be used for kith and kin care. Subsidy take-up rates of almost 1 in every 2 clients, revealed in San Francisco, Santa Clara county, and Tampa, are high relative to earlier studies and when ()5 compared to administrative records. Earlier we CHILD CARE cited Los Angeles county’s recent disclosure that only 1 in every 5 new clients draws their child care voucher. Part of the difference is explained by the fact that a fair number of mothers have found Head Start centers or state funded subsidy streams that operate outside the TANF system, sources of aid that welfare records do not pick up. The flip—side is that some women are paying for child care out of their own pockets, even though they qualify for cash assistance and two years of child care support. Figure 8.4 shows the share of the state samples for which private spending on care was reported, and among this subset of mothers, their monthly payments (in the most recent month for which out-of—pocket spending occurred). We asked a set of questions to verify that women were paying out-of—pocket, if so, how much, and to what extent do small payments represent mandatory co—payments (for Connecticut and Florida where they are required). For instance, 55% of all partici— pating mothers in Florida pay some cash for their Figure 8.4 0ut—of—pocket spending on child care’ui'mo‘tners not using subsidies $300 i U) E 2 $200 a) Q. U) u (D x U 0 Q- 3 $100 {.3 :3 0 $0 California Connecticut n = 36 n = 48 Florida n = 64 Among those who pay out-of-pocket for child care, mothers in Florida pay significantly less than do mothers in California or Connecticut. BERKELEY — YALE SECTION 8 child care provider. The median payment is about $62 per month, indicating that many women are facing co-payments. In California. just 16% of the participating mothers pay out—of—pocket. but those that do, appear to be disconnected from any subsidy stream. Their me— dian private spending equaled $283 per month. In Connecticut, 31% pay cash for their child care and the median monthly payment was $292. This suggests that almost one—third of all clients have insufficient knowledge of available subsidies, or that caseworkers have been unable to sign-up these clients for child care subsidies. In Connecticut, 31% of all mothers paid cash for their child care and the median monthly payment was $292. Almost one-third of all clients have insufficient knowledge of subsidies .1 :1: grams... ,..»—nx;:~‘i1\r fulu‘s‘xsm'pnkuszsz ”House”; r . a We also found that the bulk of private spending is for home-based care providers, either a licensed FCCH or kith and kin caregivers. Among those who paid out—of-pocket. cash payments equaled $277 monthly to home—based providers and $1 14 to center-based programs. This again shows that the subsidy structure—including the way that casework— ers and mothers think about aid—is heavily associ— ated with centers. This. despite the federal governments attempt to make vouchers more fungible. to be used by kith and kin caregivers as easily as using them for center—based programs. Ha GROWING UP IN POVERTY PROJECT A Child Care Quality What is the character and quality of this varied array of child care providers that mothers and young children face in low—income communities? We asked each of the 569 mothers using care whether we could visit and observe their provider and briefly assess their child. Of these, we gained access to 352 provider settings or 61%. Women whose children were attending child care centers were most responsive to the visits; those with kith and kin caregivers were the least comfortable. This partially explains why the participation rate in Connecticut was the lowest among our three states, since a majority of these mothers were using kith or kin. When we could not gain access to the child care setting. we asked the mother if we could come to their home to assess the child. Many were agreeable to this option. Observing Multiple Dimensions of Quality Like all American families with young children, out participating women chose from among the three contrasting types ofchild care. In assessing the quality ofthese very different settings, we knew that some attributes would be more difficult to measure than others. For example, it is difficult to compare the care provided by a loving grandmother to that provided by a top-notch preschool program. On the other hand. child development specialists employ assessment tools that can measure common features of diverse child care settings. For instance. one can reliably assess whether the caregiver reads with the child; how much time the youngster watches television; the types oflanguage interaction that occur between provider and child; and how the child responds to materials and concrete activities. During our visits we assessed a variety of attributes exhibited by the person caring for each child and we observed the quality ofsocial interaction between adults and children in the setting. We also report on quality indicators separately for the three types of child care, at times combining family child care homes (ECCHS) with home—based kith and kin providers. The key quality indicators include: Packer or provider attributes. We know from earlier research that key attributes of the provider and child care setting affect a child‘s rate of development and early learning. Those factors that are most consistently linked empirically to positive childhood devel- opment include the caregivers education level, professional commitment to the field, and her earnings. Together, these attributes create more stable environments for young children.“ During a half—hour interview of each provider or center teacher, we explored other attributes as well: her beliefs about child rearing, training in child development, experience caring for pre— school-age children, and her relationship with the participating mother. list: Facilities, learning tasks, and rich language. These attributes of provider settings are com— monly assessed by child care researchers with two measurement tools: the Early Childhood Environment Rating Scale (ECERS) and the Family Day Care Rating Scale (EDCRS).“S We selected 15 individual scales from each instrument covering four basic domains: space and basic furnishings, personal care routines, language—related materials, and discipline strategies exercised by the provider. Provider’s sensitivity, warm afiiect, discipline met/Joel, ana’ propensity to explain. This assessment tool, known as the Arnett Scale, focuses on the provider’s behavior and how she interacts with youngsters in the child care setting. In analyzing the scales results we found that it identified three attributes: (1) the provider’s apparent enjoyment in working with children; (2) her ability to interact in a participa— tory, non—authoritarian manner; and (3) her skill CHILD CARE in explaining things to children, as, for example, a sequence of activities related to a project, or the reasons why a child’s misbehavior is unacceptable. I Discrete social interactions and language between child and caregiver. To better understand the range of social interac— tions between the child and provider, types of activities, and the child’s emotional state in the setting, we employed the Child Care Observa— tion System (C-COS). This tool was developed by Mathematica Policy Research to help evaluate the quality of Early Head Start programs. With the C-COS we observed 40 timed “snapshots” of action, focusing on the child or caregiver, spread over the 3-hour visit. Observers recorded the following behaviors: & The frequency of language between child and provider, including who elicited the talk. The use of materials or activities by the child; the child’s affect and emotional response (laughing, smiling, crying, upset). 3% The frequency and duration of watching television or videos. a; An overall rating of the provider’s engagement and affect in responding to the child. Local Child Care Markets: Variation in Supply and Quality A pivotal question is whether mothers and children face different child care markets within their neigh— borhoods, as introduced in Section 3. There we reported on how the supply of center—based programs and FCCHs varies dramatically among the five Project cities. For example, Tampa’s comparatively large number of center—based slots may be due to the presence of many for—profit organizations that enter the field with relative ease, due to the state’s light regulatory requirements. Tampa and San Francisco stood out in terms of their higher per capita avail- ability of child slots. BERKELEY — YALE m SECTION 8 Figure 8.5 Child care quality: Percentage Of providers With more than a high SChOOi education 60% 30% 0% California Connecticut n = 176 n = 74 A closely linked question is whether cities vary in the quality of care made available to welfare families in low—income neighborhoods. Let‘s turn to how quality levels may vary across the state samples and among the three types of child care. Between-State Differences in Quality Attributes ofrlri/a’ care providers. Figure 8.5 displays average education levels for center-based teachers or for caregivers in FCCHs and kith or kin arrange— ments. Providers selected by participating women in Connecticut were considerably less educated than in the other two states. In Connecticut, just 34% of selected child care providers had more than a high school education. In California and Florida, 58% and 44%, respectively. had pursued some form of postsecondary education (p<.001). The mix of the three types ofproviders varies among the three states. Since Connecticut mothers were less likely to select a center—based program— and we will see that center teachers are better a: GROWlNG UP IN POVERTY PROJECT 1 Florida n = 1 20 Average education levels are significantly greater among California providers. educated than home—based providers—it is under- standable that the average education level is lowest for providers in Connecticut. Table 8.] details additional between-state differences in the attributes of child care providers. Center teachers in California are significantly older (41 years on average) compared to teachers in Connecticut (36 years) and Florida (34 years), and they are the most experienced. Among the three states, Florida home-based providers are the youngest. The ethnic characteristics ofproviders chosen by participating mothers across the three states also differ substantially. The share of center—based teachers who are African—American was 54% in Florida, but 17% in Connecticut. In California. 17% of the center teachers were Anglo, compared to 58% of teachers in Connecticut. Fully 41% of the home— based providers in California were Latina. compared to just 10% in Florida. ngc’ 11211! sum/l r/Ji/d rare settings. Certain features of child care organizations have been empirically I; , I} 5.4» ,4, ‘7‘: 312:7? an I t. (.3 s— In I; Table 8.1 Additional differences in Child care quality across state samples (means and selected medians in parentheses) F-value1 California Connecticut Florida CHARACTERISTICS OF CHILD CARE PROVIDERS Age (years) Center teachers 41 56 54 6.85*** Home-based providers 46 45 59 555* Years of experience taking care of young children Center teachers 16 14 11 6.09** Home-based providers 21 28 19 1.04 Ethnicity of providers Center teachers % Asian 22 — 1 1218*” % Black 56 17 54 4.75** % Latina 21 17 11 1.20 % White/Anglo 17 58 51 4.92** Home—based providers % Asian 14 5 — 4.77** % Black 20 45 45 7.74*** % Latina 41 12 10 12.70*** % White/Anglo 19 57 54 597* OBSERVED INDICATORS OF QUALITY Number of children in the setting Centers 15.5 15.8 12.1 4.76** Home-based providers 5.9 2.5 5.2 476* C-COS interaction measures2 Provider responds to focal child talk 4.6 (1) 10.9 (8) 5.6 (1) 16.28*** Provider requests focal child to talk 6.9 (5) 8.6 (6) 4.6 (5) 9.80*** Provider reading to focal child 0.9 (O) 2.1 (1) 1.5 (O) 4.69** Focal child interacting w/other children 12.4 (10) 12.5 (9) 15.0 (15) 2.55 Focal child interacting w/materials 25.1 (27) 29.5 (51) 25.1 (26) 6.21 ** Focal child watching video or TV 5.8 (O) 5.5 (1) 5.2 (O) 1.88 Focal child wandering, unoccupied 2.5 (O) 1.6 (1) 9.6 (9) 81.41 *** NUMBERS OF PROVIDERS (N OF CASES) Centers 65 12 100 Family child care homes 57 51 10 Individual providers 65 5 15 ACE OF FOCAL CHILD (MONTHS) 28 25 29 Home—based providers include family child care homes and kith or kin caregivers. These two types did not differ significantly overall. ' Statistically significant differences are assessed by analysis of variance (ANOVA). *p<.05 **p<.O1 ***p<.OO1. i The Child Care Observation System (C-COS) includes 40 possible snapshots for each behavior or action observed. BERKELEY ~ YALE SECTlON 8 associated with stronger levels ofchild development. Young children tend to learn more—both in terms of cognitive skills and social development—when they are in educational settings with fewer children or where the ratio ofchildren to well-educated adults is low.“8 In Florida, for example, the ratio ofchildren per adult caregiver equaled 6.6: 1, compared to Connecticut where it was just 26:1 and California where it was 3.221. Again, these differences, in part, rellect the differing mix of center—based versus home—based providers among the state samples. That is, this stalling ratio will likely be higher in the state samples where more children are attending centers, compared to states with families that rely on smaller kin settings. So, in Figure 8.6 we report stalling ratios by type of care. \Ve see that Florida centers display signifi— cantly higher childzteacher ratios, suggesting lower quality, compared especially to the centers in California selected by our families. Figure 8.7 reports on the average number of chil- dren in the setting. Sometimes this variable can predict rates of early learning, independently of the stalling ratio. Group size in centers is highest in Connecticut (not Florida). FCCHs in both California and Florida serve more children, on average, than in Connecticut. Many children are attending center classrooms that have too many children, according to professional standards. Figure 8.8 displays the distribution of group size and childzstaff ratios, combining data for all three states. Each set of three bars show the 25th percentile, the median, and the 75th percentile when ranking center classrooms from the smallest to the largest. The median group size nationally equaled 1 1 children in the classroom. But one— fourth of all centers contained 17 children or more. This is too big for a classroom of two— to four—year— olds. Similarly, one—fourth of all centers serving our families had childrstaff ratios of 9:1 or higher. This is not a good situation for young toddlers. Figure 8.6 Child care quality: Ratio of children per adult provider Number of children per adult ‘T 3_ 2* 2.2 1 _ 07 California Connecticut Florida n=186 n=68 n=125 I Center Family child care home Unregulated kith & kin care Among centers, Florida's average child: staff ratio is significantly greater than in the other two states. En] cROWiNc UP IN POVERTY PROJECT A Figure 8.7 Child care quality: Number of children in provider settings 16 14 12 10 2.6 California Connecticut Florida n=177 n=67 n=121 I Center Family child care home Unregulated kith & kin care Among centers, Florida‘s average number of children is significantly lower than the other two states. Figure 8.8 Group size and staff ratios in child care settings vary widely 1 8 — 1 7 1 5 — 1 2 — ,n’iiifi‘i: 11 9— . 9 6“ <éflL$ 6 34_ 43£45 /T7? 3 cf;....~;;',;;.5 3 2 a ll 0* ‘ l l | Centers Family child care Centers Family child care n = 158 homes n = 1 58 homes n = 116 n = 1 1 6 CHILD GROUP SIZE CHILDREN PER ADULT I Smallest quarter of settings Median setting Largest quarter of settings Smallest settings are those with fewest children, serving 25/, of CUP children. Largest settings have the most children, serving 25/ of CUP children. All three state samples combined. BERKELFY»~YALE SECTION 8 Facilities, organizational structure, and learning materials. Next we turn to the observational mea— sures of child care quality introduced above: the ECERS scale for center-based programs and the FDCRS for home-based providers. These measures largely assess the adequacy and safety of facilities, the extent to which the center or home setting is well organized, and the availability of spaces and learning materials that advance children’s early learning. Quality ofcbiki care centers. Figure 8.9 reports average ECERS scores for the 15 subscales that we scored during our observations in center-based programs. Each subscale ranges from a score of 1 (poor) to 7 (excellent). An average score of3 means that the provider setting is of mediocre quality; an average score of 5 indicates good care, according to the child development experts who developed and use the ECERS instrument.” Average quality differs widely among the states. ECERS scores for child care centers in California fell just above the “good” rating at 5.2, compared to L L Florida where the mean score was only 3.3 or of “minimal” quality. Centers in Connecticut scored even lower, averaging just 2.3 on the ECERS. The Connecticut results must be interpreted cautiously. They are based on the 12 centers in the New Haven and Manchester areas selected by participating families. Again, we are constrained by the reality that a very small proportion of all mothers in Connecticut find center—based programs. We can also focus on the distribution in quality among centers, ranging from poor to excellent. For instance, 42% of all the centers we assessed (71:165) scored below 3 on the ECERS, that is, these centers would be displaying poor quality in the eyes of early care and education professionals. In the Florida centers, 58% were rated as poor; whereas just 9% were assessed as poor in California. Across the three state samples of centers, 21% (just over 1 in 5) were assessed as providing good or excellent care and early education. Average center quality also differed between cities within California. Centers serving participating Figureiaa Child care quality: Average ECERS scores for centers 7 Average ECERS score A Connecticut n = 59 n = 12 Average score in California is significantly higher than in the other two states. California EH GROWING UP lN POVERTY PROJECT A Florida n = 94 CHILD CARE Figure 8.10 Child care quality: Average FDCRS scores for all home-based providers California children showed high quality in Santa Clara County with a mean ECERS score of 5.8. Quality was relatively high in San Francisco, although falling below the “good” level with an average of 4.6. Quality of home—based eloiia’ care providers. FDCRS scores for home—based providers are illustrated in Figure 8.10. Here we observed quite low levels of quality across all three states, ranging from a low of 2.5 in Connecticut to 3.0 in Florida (not statistically different). FCCHs and kith or kin settings came in at quite similar levels on the FDCRS indicators, so they are combined for simplicity. Pooling all home—based providers across the three state samples, we observed that fully 71% scored an average of 3 points or below across the quality indicators (712200 FCCHs and kith or kin caregivers). This indicates that 7 in 10 home—based settings offer poor quality, based on the F DCRS benchmarks of quality. Only 13% of all home—based settings exhibited good or excellent quality. Connecticut n = 118 n = 59 Florida n = 26 Average FDCRS scores, consistently low, do not differ significantly by state. Qualities of social interaction. The Arnett Scale taps into the child care providers apparent enjoyment in being with children, her ability to interact in a participatory, non—authoritarian manner, and the extent to which she explains to children the se- quence of activities presented, or reasons for misbe- havior. We were able to use 23 items on the Arnett instrument, each ranging from a score of 1 (low) to 4 (high). This measure can be used both in centers and home—based settings. Figure 8.1 1 summarizes Arnett scores among the three state samples, combining center—based and home—based settings. The character of social interac— tion is quite similar between California and Con— necticut settings, despite the Wide gap we saw in terms of the organizational features and materials available to the children. In both states the average Arnett score was 3.0 out ofa possible 4 points. But again, we see that settings in Tampa are lagging behind, with an average Arnett score of 2.6. Social ana’ linguistic oe/oaviors of children arial caregivers. Our 40 snapshot recordings of social BERKELEY — YALE EB SECTION 8 Figure 8.11 Child care quality: Average Arnett social interaction scores /———.—rv——~»——~———-~ 3 i K 1 3.0 l 2 1 0 California n = 138 n = 56 interactions using the C—COS instrument further illuminated life inside these child care settings. Table 8.1 reported on several of these social behaviors by state. We see, for example, that the number of snapshots where we observed the child care provider (P) responding to the focal child (FC) ranged from 4.6 in California to 10.9 snapshots in Connecticut (out of the 40). The incidence of reading was very low, across all three states. Connecticut providers were engaged in reading activities in 2.1 snapshots on average, compared to 0.9 in California. Interactions with other children were more commonly observed, ranging from 15 snapshots in Florida to 12.3 in Connecticut. Watching television occurred much more frequently than reading in all three states. Q [flexibility ofazrrgizm's. Is it irrational for women to rely on kith and kin? Well, not if‘ flexibility is seen as being important. Figure 8.12 reports on one facet of flexibility: whether the caregiver can take the focal EH GROWING up IN POVERTY PROJECT A COHHGCfiCUt Florida n = 110 Arnett scores are significantly lower in Florida, compared to California and Connecticut. child early or keep the child later than usual. Note that mothers who rely [655 on centers report having more flexible caregivers. IS Welfare Reform Pushing Children intp .Subetaflflfird guild care? h‘<4".é;.1‘~«_- >r.w®'1r( man The short answer is, yes. Earlier national studies, using the same quality measures, make it possible to compare the quality of our participating caregivers to quality levels observed in a wider range of communities. The findings just presented show that under the welfare—to—work mandate. many mothers are placing their young children, for the first time, in mediocre to poor quality child care settings. But this second method of comparing the quality of caregivers asks whether L L the quality of care available to welfare families meets quality norms set in other child care markets, including those that serve middle-class parents. This approach doesn’t assume that professional standards are the only useful point of comparison. Figure 8.13 compares our observed ECERS and FDCRS scores against the most sound national studies conducted earlier, each observed child care settings around the county, including sites in our Project states, and each study used trained observers who met a certain criterion of reliability in their quality ratings. We compare ECERS scores for centers against two earlier studies: the Berkeley—Yale GUP lil/idizjy Sltbsrszy, conducted in 1997—98, and f/JC’ Cost, Qua/fly, and C/Ji/r/ Outcomes Study, pub— lished in 1995. Details on these studies appear in Appendix 2. The center comparisons (first two clusters of vertical bars) show that the quality of participating centers in California is actually a bit higher than the wider sample ofcenters randomly selected in the two earlier studies. But in Connecticut, the average ECERS score for centers falls far below the average quality levels observed in the earlier studies, involv— ing a wider range ofcommunities. The comparative situation is worse when looking at the assessed quality of FCCHs. Quality levels for the FCCHs selected by participating mothers in both California and Connecticut are well below the quality levels observed in the two earlier studies. A third national study of home-based child care, conducted by Ellen Galinsky and colleagues pro— vides a third comparative study. Our participating FCCHs also rank below the average level of quality that they observed. Figure 8.12 FlEXibilitVI Caregiver can take the Child earlv 0|“ late if HEEGEG ’l 00% 950/0 - 86% 64% 50% 0% California n = 1 73 I Center Centers are significantly less flexible. Connecticut n = 60 n = 109 Family child care home 92% 78% Florida Unregulated kith and kin care BERKELEY — YALE {Ta ’3" SECTION 8 Quality Differences among the Three Types of Care Provider characteristics. Table 8.2 reports on the same indicators of quality but looks at average levels among the three types of child care: center-based programs, FCCHs, and kith or kin. The first row reports that average education levels for all types of‘child care providers were unimpres— sive. For centers, 65% of teachers serving participat— ing children had more than a high school diploma, with the remaining third reporting no schooling past high school. Among FCCH providers, just 51% had pursued any form of postsecondary schooling, and among individual providers, the figure was only 26%. Among the three types of care, children in centers were spending their days with somewhat better—educated adults. Among all providers, those at centers were the youngest—37 years on average—but they also possessed considerable experience in working with young children, 13 years on average. The oldest, most experienced providers were individuals in home settings, with a mean age of 47 years. They reported having 22 years ofexperience working with young children, which may include time they spent in their role as mother, aunt, or grandparent. The ethnicity of providers varied substantially by provider type. Fully 47% of all center-based teachers were African-American, while just 15% were Latina. For individual providers the situation was reversed: 43% were Latina and 21% were African American. This corresponds to the ethnic composition of our city samples and their provider markets. Provider romrrzitment andproflssiono/ orientation. We asked each caregiver two different sets of questions related to her commitment to child care work. First, we asked questions, like whether she “frequently Feels like quitting“ or “feels stuck taking care of other people‘s children.” Positive questions also were included in this scale. Providers overall expressed Figure 9.13 Are weifare thildren exposed to lower qLIalitv child care? I 6— 5.2 5— 4.5 4.1 ‘ 4.3 4-4 8 I g 4— ‘ i 3 , LE) 3_ 2.8 , ._ 2.3 O U) s 2— U LU 1 — I 07 I l I | 'i i l l i I CA centers CT centers CA FCCHs CT FCCHs National FCCHs I 1998-99 cup 1997-98 GUP comparison 1 995 C00 comparison 1 994 FWI comparison Quality scores are from the ECERS (centers) or FDCRS (family child-care homes). The 1995 study is the Cost, Quality, and Child Outcome (COO) project. The 1994 FWI Is the national investigation of home-based care conducted by the Families and Work Institute. Details appear in Appendix 2. E13 GROWlNC up IN POVERTY PROJECT A Table 8.2 Differences in quality among types Of Child care (means and selected medians in parentheses) Centers Family child Iiitli and Maine1 care homes kill Characteristics Of Child Care Providers Education level: percentage with more than high school diploma 65 51 26 1768*“ Age (years) 37 43 47 1969*” Years taking care of young 13 19 22 18.23*** Ethnicity of teachers/providers % Asian 8 23 5 7.55*** / Black 47 27 21 10.72*** 96 Latina 15 26 43 12.41*** % White/Anglo 25 2O 24 0.33 Professional orientation Intentionality 1 3.5 37 3.6 1.72 lntentionality 2 2.8 2.3 2.4 10.78*** Desire for inservice training 3.0 2.3 2.4 7.7 *** Authority oriented child—rearing beliefs 3.5 3.5 4.0 338* Provider’s agreement with parent (index) 2.8 3.2 3.4 15.62*** Months child with current provider 3.6 (2) 10.8 (5) 9.0 (2) 20.79*** Observed Indicators Of Quality Number of children in setting 13.1 (12) 4.8 (4) 2.6 (2) 112.63*** Ratio of children—to—adults 6.8 (6) 2.5 (2) 2.0 (1) 7020*” ECEPS scale for centers, average score (1-7 possible points); 4.3 — — — FDCRS scale for home—based providers (1—7 possible points)3 — 3.0 2.7 2.58 Social interaction scale (Arnett)E Factor 1 (main items) 2.9 3.0 2.9 1.00 Factor 2 (explanations) 2.2 2.0 1.9 301* C-COS interaction measures out of 40 snapshots3 Provider responds to focal child talk 4.2 (1) 7.0 (4) 7.1 (3) 6.04** Provider requests focal child to talk 4.2 (3) 7.6 (6) 8.6 (8) 20.36*** Provider reading to focal child 1.2 (0) 1.2 (0) 0.7 (0) 0.97 Focal child interacting w/other children 17.7 (17) 10.1 (7) 7.8 (4) 34.7 *** Focal child interacting w/materials 25.5 (27) 27.0 (29) 25.3 (26) 1.09 Focal child watching video or TV 1.6 (0) 5.0 (1) 6.9 (3) 8.56*** Focal child wandering, unoccupied 7.2 (5) 2.6 (1) 2.5 (0) 27.85*** Cases 158 64 72 Age of focal child (months) 31 29 26 ' Statistically Significant differences are assessed by analysis of " Significant at p<.07. x" ,(‘.:“l ()7, ”. ”’. . , , . . . variance A OVA D< OS D< 01 D< 001 . Paw scale scores for a 4-p0int scale on 23 well distributed items. ECEFS score reported is an average score based on 15 quality items Factor 1 includes 20 items, and Factor 2 includes 3 items, as detailed ass eses ed Each item score can range from 1 to 7 points in the text. C010 mbines “£60860 family Child care homes and klth Of km FDCPS ‘ The Child-Care Observation System (C—COS) includes 40 possible co rse re eported are averages based on 15 quality items assessed snap-shots for each behavior observed. Ea ch item score can range from 1 to 7 points, BERKELEY — YALE SECTION 8 strong agreement with statements related to the enjoyment or desirability of caring for young chil— dren. They also disagreed with statements that indicated child care was something that they did reluctantly. That is, most providers felt positive about their role as caregiver. We call this lntentionality 1 in Table 8.2. We found wider variability when we asked whether providing child care services was (a) “a stepping stone to the work (you) really want to do,” (b) “just something you do to help out the mother,” or (c) “temporary employment.“ As one might predict, teachers in child care centers feel strongly that what they are doing is not merely a temporary job or just something to get by on. Instead, their desire to be a paid child care provider is quite strong compared to the motivations of FCCH or kith and kin providers (lntentionality 2 in Table 8.2.) Finally, we asked providers ifthey would like addi— tional inservice training. Center—based teachers were much more interested in taking advantage ofoppor- tunities for training than borne—based providers. Provider beliefs: and relationships wit/J parents. We talked with providers about their beliefs regarding child rearing and relationships with our participat— ing mothers. The only significant finding was that kith and kin providers express more authoritarian beliefs about obedience and child socialization. The difference was not large, however, among the three types of caregivers. Organizational features. Table 8.2 also reports on group size and staffing ratios by type of provider (combining data from all three states). Center classrooms contain 13.] children on average, com- pared to 4.8 children in FCCH settings, and 2.6 youngsters with kith and kin. Interestingly, the median number ofchildren present in the latter two settings are not too far apart: a median of4 children in FCCHs and 2 children with kith and kin caregivers. Despite the licensing requirements which separate FCCHs and individual providers GROWING up IN POVERTY PROJECT A with more than one child, the social organization of the two types does not look all that different in low— income communities. Materials, facilities, and learning activities. Table 8.2 also reports the average ECERS scores for all centers in the three—state sample, which equaled 4.3 or a level between “minimal” and “good,” according to professional standards. For FDCRS scores, we see that FCCHS scored a bit higher, compared to kith or kin caregivers. Yet this difference is only marginally significant (p<.07 level). These two settings, in sum, are remarkably similar on this measure of quality. Provider sensitivity and discipline practices. The data revealed little difference on the Arnett scale among the three types of providers. However, there were differences in how the provider explains activities, task routines, or reasons why a child’s behavior is inappropriate. Center teachers scored significantly higher on this index (2.2); kith and kin providers scored the lowest (l.9,p<.05). Social and linguistic lie/laviors of child and caregiver. The C—COS assessments revealed more fine—grained differences among the three provider types. The first C—COS item in Table 8.2 concerns the average number ofsnapshots where the provider responded to the focal child‘s talk (out of40 possible times). This occurred in 7.0 and 7.1 snapshots, among FCCH and individual providers, respectively. In contrast, center teachers responded to the child’s talk just 4.2 times. This is not surprising, given the larger number ofchildren in center settings. However, it may hold implications for youngsters language and social development. We see a similar pattern for how often the provider requested the child to talk. The tendency of providers to participate in reading activities was very low, averaging less than 2 snapshots during the observation period. On the other hand. children watched television and videos far more often in FCCHs (‘3 snapshots) and kith or kin settings (6.9), compared to centers (1.6). The tendency of- child care providers to participate in reading activities was very low, averaging less than 2 of 40 snapshots. Children attending centers interacted with other children considerably more often (17.7 snapshots), compared to FCCI-ls (10.1) or individual providers (7.8). This is due to the fact that there are fewer other children in home—based settings. At the same time. children in centers were more likely to be wandering around, disengaged from any identifiable activity (7.2 snapshots), compared to both types of home—based settings (2.6 and 2.5 for FCCHs and kith or kin). Summary: A Child Care Infrastructure - (With Crumbling Foundations Welfare reform is built upon a central policy Filler: Women receiving cash assistance should find and hold down a job. This necessarily requires that their preschool—age child be placed in a child care setting. Federal and state governments now spend billions of dollars each year to help lower the child care barrier, expanding supply and subsidizing the cost. How are women responding to these new child care initiatives? We observe that only a fiaction oft/Jese single mothers utilize their child care suositly. This ranges from just 13% among participating women in Connecticut to 50% among those in Florida. Thus true access to financial aid falls well below the hopes of policy makers, and we see stark inequality in access to aid across the states. Some will argue that many parents “choose” to avoid the subsidy system, preferring to set—up CHILD CARE informal arrangements. But do individual—level preferences really vary this dramatically among women in different states? Local implementation and institutional constraints are more likely suspects in explaining these divergent take—up rates. Even the fact that 1 in 2 women utilize their child care subsidies in California and Florida is not impressive when put up against much higher enrollment rates for Medicaid and food stamps. We see enormous variability in t/oe types ofcliiltl care that mothers select across t/se state samples. This may be a function of the break down in subsidy flows or stem from the uneven supply conditions that we saw in Section 3. Among participating mothers in Florida, 70% placed their young child in a center—based program, compared to just 13% in Connecticut. A small part of this difference may be explained by the fact that children in the Connecticut sample, on average, are younger. But our earlier paper demonstrates that where families live explains much of this variation in type of care selected, after taking into account maternal and child attributes. Supply conditions and the efficacy with which caseworkers connect mothers to centers appear to be influential factors. We are not concluding that in all cases center- based programs are of higher quality than home— based providers. But our findings do suggest that centers, on average, offer more stimulating and educationally rich settings. The quality of center—based care varies enormously across participating states and cities. In California, 9% of all centers were assessed as providing poor quality. But among the more abundant number of centers in Florida, 58% were rated as poor. Higher supply alone does not ensure that these settings for young children are nurturing and stimulating places to be. Most women do not choose center—based programs. They select home-based settings, BERKELEY — YALE EH SECTION 8 either family child care homes or kith and kin arrangements. Overall, these settings are of lower quality than center—based programs, among those selected by participating mothers. Fully 71% were assessed as poor settings for young children. There are glimmers of positive ele— ments, for example, children are talking more with adults in home-based settings, relative to centers. But the television is on far more, fewer books and learning materials are present, and these homes offer less organized and sometimes less safe environments for young children. Our colleagues at MDRC have raised the important question of the magnitude with which the welfare— to—work push per re spurs this migration of young children into low quality child care. Perhaps, they argue, women facing welfare-to—work requirements do not select child care oflower quality, relative to other poor women. The difference in the shares of Connecticut women who had selected a child care provider. between the experimental and control groups, was only 7% (44% versus 37%, respec- tively). This stems directly from the fact that the employment effect of welfare—to-work in Connecticut was quite modest. One can not logi— cally claim that welfare—to-work policies hold a strong effect on maternal employment rates, but do not move children into new child care settings. In addition. welfare reformers over the past decade have pushed hard to implement a particular se— quence of behavior on the part ofsingle mothers: They should find a job, then find child care to ensure that they keep their job. The migration of young children into new child care arrangements is a }l(’(‘(’.\‘.\‘117:)/316/), recognized by policy makers, if their mothers are to leave home and work. l’oor women who do not face the welfare—to-work mandate may choose to select lousy child care, but they are not under a government Ill/Illfllu'l‘t’ to do so. Finally, if most low—income women—be they welfare poor or working poor—can only find bad— to—mediocre child care options in their neighbor- hoods, then is there not a fundamental public Ed GROWING UP IN POVERTY PROJECT A interest in expanding and improving the quality of these options? And even if one does not worry about young children’s new settings, there is a utilitarian interest in expanding the availability of quality child care: their mothers are more likely to hold down a job and stay off welfare if they can find child care which they trust. Our report contributes to the emerging evidence that shows how welfare—to-work efforts are leading to some positive results for some women. These child care findings, however, illuminate how the destinations of their young children are highly variable across locales and, in general, reflect settings that are not well organized, nurturing, or cognitively stimulating. The obvious conclusion is that govern— ment, so intent on moving mothers from welfare to work, has yet to show the same efficacy in expanding child care and improving the quality of these settings—places where hundreds of thousands of additional children are now being raised each day. Finally, we turn to how these youngsters are develop— ing, both in their early language and cognitive growth, as well as their social development. We essentially offer baseline data for California and Florida. Initial experimental and control differ- ences can be assessed with the Connecticut sample. And for participating children in all three states, we also offer comparisons with wider national samples ofchildren. SECTION 9 III TIIIS SECtIOII a How well are children growing and learning? Are social behavior problems emerging? a Does the welfare-to-work push help or hinder children's development? The-!3°tt0'11'-'“e _W tawu: . ‘.w>1&éa{v'vii;. Manx _ .,!‘;?1: 9:3»; 2-5:; After taking into account womens personal resources, their households. and child care support, we come to one bottom—line question: Does the push on mothers to move out from welfare to work advance or hinder their children‘s early learning and development? \We already have described the quality of child care into which youngsters are being placed. A majority of children are moving into settings of mediocre or poor quality. We also have detailed how many home settings are not as nurturing as they could be: a large share of mothers read infrequently to their children; over one-third of the women feel alone as a parent; many mothers suffer from depression and feel only modest efficacy in raising their young child. And the basic stability of households is often shaken by economic insecurity and by co—residents who suffer from myriad problems. In short, policy makers and local program designers must think carefully about two sets of forces—the quality of child care and the collateral influence of parenting practices and the mother’s own personal resources. We already know that these factors heavily EARLY LEARNING ll hey learn their colors and their ABCs. They get to interact with other children. So they're getting good associations at school, too. And they teach them discipline, obedience. They teach a lot of things. —— Marta influence children’s early cognitive and social devel— opment. We know less about how welfare-to—work pressures will embolden or diminish mothers’ parenting practices or their ability to find quality child care. In this light, the mechanisms through which employment alone will lighten these heavy burdens felt by many mothers remain unclear, unless income is raised markedly or parenting practices are enriched significantly. This section does not definitively answer the bottom line question: Does welfare reform help or hinder young children’s early learning? Part of the analytic problem is to disentangle it from these other collat— eral factors. Connecticut’s experimental design helps in advancing causal understandings. But more analysis is required before we can shine a light on BERKELEY ~ YALE En SECTION 9 the human—scale mechanisms through which wel— fare—to—work processes influence parenting practices and child care decision—making. This section largely focuses on baseline levels of language and social development for participating children in California and Florida. We are currently tracking youngsters‘ developmental trajectories into the Wave 2 data collection. A subsequent report will examine the relationship between participating women‘s employment histories with the growth trajectories exhibited by their young children. We report below on how children compare to national norms of development exhibited by wider samples of youngsters. It is no surprise that most children in welfare families lag behind middle-class norms when it comes to early language develop- ment. The research community has long known that children in welfare poor and working poor families begin kindergarten with suppressed language profi— ciency and often lower levels of social development. exhibiting early forms of aggression and less capacity to concentrate on cognitive tasks.“ For toddlers in California and Florida we show that this gap appears as early as 24 months. . a, ”Jana-,xsxswx ., Whether and how maternal employment is going to rectify this early inequality in child development is a question that should be moved to the front burner. m "minnhgn .~ *mam'z.-rnr5x=;:2v. .mpgra» .mw‘wvw .4. (3- ,.. .. aw?»ia».e<\‘»wnJ‘l"-\l‘~vr>i m1.“- ”may“ « m » \Whether and how maternal employment is going to rectify this early inequality in child development is a question that should be moved to the front burner. Even these baseline findings suggest that we must separate the high hopes surrounding welfare reforms promise for children from the reality of their daily lives. In Connecticut. we can take the next step to assess whether children in the experimental and control m GROWING up lN POVERTY PROJECT A groups display differing levels ofdevelopment. This analysis, in short, reveals no significant differences. It can be seen as good news: A slightly larger share of mothers have moved into jobs with no discernible harm to their children. But the flip-side is that we found no evidence to back the high hopes of policy makers that welfare reform will advance the well— being of children:I We are looking at the early months ofwelfare experience, felt by mothers and young children. Perhaps child—level benefits will accrue over a longer stretch of time. Measuring Child Development . wand .5?le learning Early Language and Cognitive Proficiency One challenge in assessing children’s development —at age two or three—is to find measures that can reliably detect differences and be used in a large— scale field study. Few validated measures of early language growth are available for children under two years of age. We decided to include young toddlers of 12—74 months in the Project. given worries among some policy makers and local practioners over the adequacy ofchild care and other commu— nity resources for this young group. One set of measures—used with young children and predictive oflater development and early school performance—is the MacArthur Communicative Development Inventory (hereafter. the MacArthur). Both reliability and validity studies have been conducted over the past decade, using broad popula- tions of middle—class and low—income families? Another advantage in using the MacArthur is that it can be adapted for a large scale field study, since the parent is asked about the child‘s different forms of communicative proficiency. including word recogni- tion and production. sentence complexity. and gesturing. Spanish and Vietnamese versions ofthe parent reports and direct child assessments were developed.—3 Then. to verify the reliability of parent reports we directly assessed childrens level of word recognition and production. We found that mothers of the youngest group, children of 12—23 months, over—estimated their youngsters‘ word comprehension. Social Development Two standard measures of young children‘s social growth and early behavioral problems were used in California and Florida. The Emotionality, Activity, Sociability, and Impulsity Scale (EASI) was em— ployed in assessing young toddlers, of 12—23 months: the Child Behavior Checklist (CBC or Achenbach) was utilized for the older preschoolers, 01‘24—42 months. These measures tap into the mothers view of her child‘s social proficiencies, emotionality, and tendency to act impulsively. They ask the mother both about positive and negative forms of social or expressive behaviors (e.g., crying, tantrums, friendly, talkative with strangers). Given time constraints and priorities in Connecticut, the social development measures were not included. EarIVLanQuaee Development ... Maternal Reports We begin with children’s variable levels of word comprehension and production, as reported by the mother, on a standard set of20 words (Figure 9.1). This graph reports the average number ofwords that mothers believed their toddlers (12—23 months) understood, or their preschoolers (24—42 months) understood and produced. This scale ranges from 0 to 20. Figure 9.1 Mothers' reports of children’s word recognition (MacArthur scores) 16 14 14.5 12 1O MacArthur score (scale 0 - 20) 00 California n = 118 toddlers n = 270 preschoolers I Toddlers 12-23 months 1 4.0 Connecticut Florida n = 142 toddlers n = 147 preschoolers n = 51 toddlers n = 1 28 preschoolers PFESChOOIGFS 24-42 months Toddler scores are significantly higher in Connecticut, compared to California and Florida. Preschooler scores do not differ significantly by state. BERKELEY — YALE if SECTION 9 Figure 9.2 Mothers' reports Of children’s sentence complexity for preschoolers, 24-42 months Average score (scale 0 - 10) (N 2 1 0 California n = 261 Mothers‘ perceptions of their children’s proficiency, for both the toddler and the preschooler cohorts, are higher for the Connecticut sample, relative to maternal reports for California and Florida. For example, the mothers of Connecticut preschoolers reported that their youngsters understood and used just under 16 of the 20 possible words, compared to 14 words among the mothers of Florida preschoolers. The between—state differences for preschoolers are not statistically significant while the observed differences for toddlers are significant. It‘s important to recognize that the Connecticut advantage may be due to these mothers‘ higher level of school attainment, on average, compared to the education levels of participating women in California and Florida. In addition, lower language proficiency in California is likely related to the higher share oflanguage—minority children who are grow- ing up speaking Spanish or Vietnamese at home." m GROWING UP IN POVERTY PROJECT A Connecticut Florida n = 129 Children's sentence complexity scores are significantly lower in California, compared to Connecticut and Florida. No significant differences were observed between the experimental and control groups in Connecticut. This is good news in the sense that children with mothers who are facing the welfare—to-work push 18 months into the experiment, have not been disadvantaged in this one measure of early language development. Another scale within the MacArthur battery pertains to the complexity of phrases or sentences used by preschoolers (age 24-42 months), as reported by the mother. For example, mothers are asked whether their child says, “Where did mommy go?” or “Where mommy go?“ The age at which greater complexity oforal sentence construction appears is correlated with school readiness and early school performance. Here too, we see that Connecticut children appear to be doing better (Figure 9.2). The widest gap appears between Connecticut and California young— sters, the latter sample again including more language- minority households. This scale ranges from 0—10. Connecticut preschoolers scored just over 6 out of 10 possible points, with California children averag- ing just under 5 items correct. Again, remember that Connecticut mothers, on average, had more school— ing that women in the other two state samples. Direct Assessment of Children's Word Recognition We adapted the word recognition and production sections of the MacArthur, building a picture version for 16 or 14 words (toddler and preschooler forms, respectively). The child was asked to identify the picture of each word where other objects appeared along with the correct picture:a This adaptation worked well for the toddlers (12—25 months—old), yielding a fairly normal distribution of scores, especially when the higher scoring Connecticut children were separated out. But the preschooler version resulted in a skewed distribution. This picture version proved to be too easy for the older group. There were 5 words that less than two-thirds EARLY LEARNING of the children correctly identified. This reduced set of items is more normally distributed and will be used in future analyses. The direct child assessments revealed lower levels of proficiency, relative to their mothers’ own assess- ment, when we compared recognition of particular words (detailed in Appendix 2). The direct assessments also verified another consis— tent finding: Connecticut children appear to be developing these basic elements oflanguage profi- ciency at a higher rate, compared to toddlers and preschoolers in California and Florida (Figure 9.3). These differences must be interpreted carefully for the toddlers, since the number directly assessed was less than the number for which mother reports are available. The between—state differences for preschoolers are more robust, given larger subsamples of this older group. Once again, keep in mind that Connecticut mothers are better educated and a smaller proportion are non—English speaker, especially compared to the California sample. - , .ea “fine-L315» 4' Figure 95.3”ijiriett~ child asiséssmeritswo‘f language déiéio'p'ment C(pictureyersion ofillllacArthur) 80%— 600/0— 40%— 200/0_ T 21% ll % of pictures correctly identified 00/0 1 Toddlers ’I 2 - 23 months I none - 25% 54% 7 I 26% - 50% Preschoolers 24 - 42 months 51 % - 75% 76% - 1 00% Among both toddlers and preschoolers, children in Connecticut scored significantly higher, In Connecticut, children did not differ significantly by experimental and control groups. BERKELEY — YALE EB SECTION 9 Figure 9.4 Participating GUP toddlers' direct language assessment scores, compared t0 national norms 100 80 {”84 ' 762 ' 63 'O L O B a) U 3 'O E Q. \ (D N ‘3 o 60 U 8 2 4.) U at) 40 8 o 32 E g 20 2 '0 'U 0 "’ 0 °\° Book Blocks Doll I GUP toddlers 32 g76 57 Hat Shoe Teddy Bear National toddler sample Toddlers participating in the GUP Project were 22 months when directly assessed, on average The national norm sample was 24 months of age, on average. Appendix 2 contains details The figure reports the average number of words correctly identified by children, divided by state and age. The average California toddler, for instance, recognized almost 9 words out or 16, compared to 14 among Connecticut toddlers. The pattern is similar for the preschoolers, although the Connecti- cut advantage is smaller, relative to California and Florida. Preschoolers in Connecticut averaged 13 out or 14 words correct, versus about 1 1 words among California and Florida preschoolers. Note that this measure proved to be too easy for children overall, relative to the ideal distribution, limiting our ability to discriminate among preschoolers. Are Children in Welfare Families Lagging Behind? The short answer is yes—compared to normal levels ol‘language development experienced by wider populations or young children. We were not expect— ing participating children, just 6 to 18 months into m GROWING UP IN POVERTY PROJECT A the welfare reform experiment, to be performing on par with middle—class children. In addition, from the Connecticut experimental data, we observe no decrement, nor any discernible advantage, among children whose mothers are racing the welfare—to- work push, relative to the control group. More work is required, including Wave 2 data, to better understand how maternal employment may advance or hinder early learning and development. And possible effects will likely be conditioned by the quality of‘child care that mothers are able to find in their neighborhoods. Yet we can compare how toddlers performed on the direct assessment or word recognition in California and Florida, relative to national norms. Combining these two state samples yielded 108 toddlers who were subject to less selection bias. That is, we were able to directly assess a large share otchildren in these two states, whether they attended a child care center or home—based setting. But in Connecticut, where the overwhelming majority attended home- based settings, participation in the direct child assessments lagged behind California and Florida. An analysis of possible sample selection bias for the child assessments appears in Appendix 3. Figure 9.4 displays levels of the directly assessed recognition of words (nouns) appearing for California and Florida children, compared to the percentage of children who correctly produce selected words at 24 months of age, based on a national study that established normative performance levels. Production of words, of course, follows recognition. And note that our sampled toddlers were 22 months of age when the direct assessments were conducted, on average, compared to the 24-month benchmark set by the MacArthur designers. Among the toddlers, 32% correctly pointed to the picture of a book, compared to 84% of a wider cross-section of toddlers nationwide. We see or instance, among our participating toddlers that 32% pointed to the picture of a book, correctly picking it out from among three different pictures. Among a wider cross—section of toddlers—not only children from poor families—84% can produce and use the word book at 24 months of age. The two— month difference in the average age of our sample would not explain this gap. On the other hand, you see that words such as dog and shoe are known equally well among our sampled toddlers, compared to national norms. These findings and comparisons are tentative. Nor can we attribute these developmental delays for toddlers to welfare reform per 56. If the gaps hold up with Wave 2 data, based on a wider array of child EARLY LEARNING development gauges at age four, then stronger claims can be advanced. We will assess the discrete effects stemming from maternal employment in the context of the mother’s personal resources (school attain- ment, mental health, social support), deep—seated parenting practices, and the quality of the youngsters child care setting. We don’t yet know whether maternal employment exerts an added influence on the child’s early learning, or whether such an effect is swamped by these other factors. Findings on Social Development and Behavioral Problems Mothers were asked about a variety of attributes regarding their children’s social skills and emotional well—being, including behavioral problems. The two standard instruments, sketched above, were used: the EASI for toddlers and the CBC for preschoolers. These instruments include questions such as, “(child) cries easily,” “reacts intensely with anger,” or “has sudden changes of mood or feelings.” Many of these items, both negative and positive, have been empirically related to cognitive development and early school performance?“ Figure 9.5 reports average EASI scores for California and Florida toddlers combined, compared to na— tional norms. No significant differences were ob— served between the two state samples. In general, we did not detect any differences between our partici— pating children and national norms on the two major scales: emotional stability and sociability. The incidence of positive and negative behaviors by participating toddlers were in line with normative levels determined in earlier national studies with wider child populations. The preschooler cohort, of 24—42 months did show a higher incidence of behavioral problems, as reported by participating mothers on the CBC, compared to norms derived from wider child popu— lations. The complete CBC asks parents a variety of questions to assess whether the young child displays problems related to being withdrawn, aggressive, BERKELEY - YALE EH SECTION 9 anxious or impulsive, and destructive or angry. We used 59 of the original 100 items that comprise the CBC interview questions, so not all domains were covered. Figure 9.6 shows how our )artici mtin v )reschoolers b E: l exhibited more aggressive and anxious behaviors, on average, than normative levels. The interview questions for this subscale ask each mother if a set of 16 attributes are not true, sometimes true, or ver y true about their focal child. The attributes include “(child) hits others,“ “has temper tantrums,” “easily gets jealous,“ and “has sudden changes in moods or feelings.“ When these individual items were com- bined into subscale scores, our participating children placed in the 70th percentile, relative to normal scores, indicating significantlv more aggressive and Q. J k & anxious behaviors. Getting to the Bottom Line Question Will welfare reform, over time, yield positive ben— efits for children’s early learning and development? Through a stronger family economy, more robust role models, and better child care, policy makers argue that welfare to work will boost the long term well—being ofchildren. We would be surprised to see any discernible gains in child development and learning this early in the implementation process. However, the absence ofany significant harm, as seen in the Connecticut experimental data, is good news. The fact that our participating toddlers, at 22 months, already display sharp developmental delays is not surprising, but it is sobering. This leads to the question of how welfare-to—work reforms will narrow these gaps, long present among children who are growing up in poverty. As we follow these Figure 9.5 Children’s level 0f social development and behavior problems, 12-23 months 20 ’15 EASI indices _\ O Emotional stability I cup toddlers Sociability National norm for toddlers Higher emotionality scores indicate that the child less frequently cries, fusses, or gets upset. Higher sociability scores indicate that the child is more open to strangers, prefers playing with others, rather than alone, and makes friends easily. Children's social and emotional development (EASI) do not differ significantly by state, m GROWING UP lN POVERTY PROJECT A iMQLYLEARRHNG Figure 9.5 Aggressive behavior DV participating DI‘GSCHOOIEFS, 24-42 months (CBC subscale) 12 10 8.7 Subscale, 0-50 points 01 BOVS - Participating GUP children 8.7 Girls Normative level Participating CUP children displayed aggressive and anxious behaviors at significantly higher levels, compared to normative samples of children. Includes only preschoolers from California and Florida. mothers and children over time, we will keep looking for positive gains in children‘s well-being. \‘C'e also will seek to identify the multiple mecha— nisms by which maternal employment might boost early development and learning, and how this process is mediated by the character ofparenting practices and the quality of child care settings. Some will be satisfied with the pattern seen thus far in Connecticut: welfare reform is working if the rolls shrink and children are simply not harmed after their mothers leave home for work. Yet proponents of welfare-to—work, prior to and following the 1996 legislation, promised more. This included more positive maternal role models, higher income, and expanded child care programs, all of which were supposed to benefit children. But how long will it take before children feel these benefits, and via what specific mechanisms? Will women moving off welfare find better child care? Will they enrich their parenting practices after working in low—wage jobs.> Will the move to safer, better nei hborhoods.> Y g The other way to ask the question is whether going to work may yield positive benefits for young children, but only under certain conditions. For instance, better educated women already display stronger parenting practices, like reading to their toddlers and preschoolers more consistentlyfifi' Or, participating women in San Jose are more likely to find a child care center of reasonable quality, com— pared to mothers in Tampa or New Haven. Allied policies can build these supportive local conditions under which welfare reform’s effects on children could become positive. Yet this is a broader policy strategy, one that moves beyond the claim that requiring work or setting time limits on cash assistance, alone, will create more promising futures for children—youngsters who continue to grow up in poverty. BERKELEY — YALE EH SECTBON 9 {IE GROWING UP IN POVERTY PROJECT A ENDNOTES ‘ Comparisons to national norms depend upon the age and ethnicity of women. See Section 6 for details. 3 This comparison to welfare mothers with school—age children is made possible by a companion study, led by Yale public health professor Sarah Horwitz. This analysis shows that maternal depression is a stronger predictor of whether women are working and moving off the welfare rolls, than simply whether they are participating in the Jobs First program. See: S. McCue Horwitz and B. Kerker, Initial Results from the Interim Survey for W/omen and Children 37 months to 10 Years. (New Haven: Yale University, Department of Epidemiology and Public Health, 2000). Sections 4 and 6 detail earlier work on how parenting practices and maternal depression influence child development. *1 Nor did we find discernible deficits in the development of older toddler and preschool-age children, 24—42 months of age. ; The overall evaluation of Connecticut’s welfare reform program, 18 months into implementation, appears in: D. Bloom, L. Melton, C. Michalopoulos, S. Scrivener, and J. Walter, jolts First: Implementation and Early Impacts of Connecticut} W/elfare Rcfitrm Initiative. (New York: Manpower Demonstration Research Corporation, 2000). [\ Quotations appear in: National Desk, “Text of President Clinton‘s Announcement on Welfare Legislation, New York Times, August 1 (1996):A24. National Desk, “The Welfare Bill: The Republicans View,” New York Times, August 1 (1996): 25. Mr. Lieberman’s comments appeared a week earlier: J. Lieberman, “Welfare as We Know It,” New York Times, July 25 (1996):A23. Council of Economic Advisors, “Explaining the Decline in Welfare Receipt, 1993-1996.“ (Washington D.C.: The White House, 1997). An update issued in 1999 attributes 26 to 36 percent of caseload decline since 1996 to economic growth, a shrinking share relative to the earlier period, given a signifi— cant slowdown in job growth. Council of Economic Advisors, “The Effects of Welfare Policy and the Economic Expansion on Welfare Caseloads: An Update.” (Washington D.C.: The White House, 1999). Rapid growth in the earned income tax credit (EITC) during the post-1992 period also may have contributed to the decline in poor families’ propensity to enter the welfare system. For review: W Primus, L. Rawlings, K. Latin, and K. Porter, “The Initial Impacts ofWelfare Reform on the Incomes of Single—Mother Families.” (Wash— ington D.C.: Center on Budget and Policy Priorities, 1999). I S. Brauner and P. Loprest, “Where Are They Now? What Studies of People Who Left Welfare Tell Us.” (Washington D.C.: Urban Institute, 1999) no. A—32. To date, many of these state or county—level studies are based on small samples with high levels of attrition of families who cannot be found ENDNOTES after leaving the rolls. This may lead to inflated estimates of positive outcomes, since the most difficult family cases are harder to track over time. 9 P. Loprest, “How Families that Left Welfare Are Doing: A National Picture.” (Washington D.C.: Urban Institute, 1999) no. B— 1. ‘0 Primus et al. (1999)- I] P. Loprest. (1999). Under the present benefits schedule, a family with a wage earner working at the minimum wage can reach the federal poverty level if they file for the EITC and receive food stamps. But without these benefits, this family remains at about 60% of the poverty level, depending upon their earned income. ‘3 A. Collins and L. Kreader, Patterns and Growth ofChild Care Voucher Use by Families Connected to Cash Assistance in Illinois and Maryland. (New York: National Center for Children in Poverty, Columbia University, 1999). "I Data reported to PACE at Berkeley by Los Angeles County, Department of Public Social Services, research and data division, for summer 1999. 1“ Government Accounting Office, GAO, “Education and Care: Early Childhood Programs and Services for Low—Income Families.” (Washingon, D.C., GAO/HEHS-OO-l 1, 1999). H Thanks are due Cheryl Black, state Department of Finance in Sacramento, for assembling these historical numbers. ”’ California data are from: Policy Analysis for California Education, “Child Care Indicators, 1998, Volume 2,” (Berkeley: University of California, 1999). A review of the recent expansion ofstate preschool programs appears in: K. Schulman, H. Blank, and D. Ewen, Seeds ofSuccess: State Prekindergarten Initiatives, [998—1999. (Washington D.C.: Children’s Defense Fund, 1999). 1" Reviewed in: M. Burchinal, “Child Care Experiences and Developmental Outcomes,” in S. Helburn, ed., “The Silent Crisis in US Child Care,” Annals ofthe American Academy of Political and Social Science, vol. 563 (1999):73-97. NICHD Child Care Research Network, “Poverty and Patterns of Child Care,” in G.Duncan and J.Brooks-Gunn, eds., Consequences ofGrowing Up Poor (New York: Russell Sage, 1997). 1“ Of the 948 women interviewed we failed to contact 17% to complete the child care follow-up questions. This is due to high rates of mobility, cut—off phone numbers, and uneasiness among a small share ofwomen in discussing child care. 1" S. McLanahan and G. Sandefur, Growing Up with a Single Parent. (Cambridge, Mass.: Harvard University Press, 1994). J. Singer, B. Fuller, M. Keiley, and A. Wolf, “Early Child Care Selection: Variation by Geographic Location, Maternal Characteristics, and Family Structure,” Developmental Psychology, 34 (1998):1129-1144. Also see papers in: G. Duncan and J. Brooks—Gunn, eds., Consequences ofGrowing Up Poor. (New York: Russell Sage, 1997). BERKELEY — YALE [El] ENDNOTES 30 B. Fuller, S. Kagan, et al., “Variation in Poor Children’s Home and Child Care Settings: Does Maternal Employment Matter?" (Washington D.C.: Joint Center for Poverty Research conference, September 1999). 3‘ For a review of these patterns, see: B. Fuller, S. Holloway, and X. Liang, “Family Selection of Child Care Centers: The Influence of Household Support, Ethnicity, and Parental Practices," Child Development, 67 (1996):3320-3337. 33 Again we see between—city differences in the maternal samples: just 26% of the San Francisco mothers, more heavily African—American in composition, had been married, com- pared to 37% among women in Santa Clara County, more heavily Latina in their composition. 3" Details on marital patterns appear in Fuller, Kagan, et al. (1999). 3" State welfare programs now have their own particular names: Florida‘s is called WAGES, Connecticut's is Jobs First, and California‘s is called, CalWORKs, with the R and K signifying “responsibility to kids.“ 5 R. Sampson, J. Morenoff, and F. Earls, “Beyond Social Capital: Spatial Dynamics of Collective Efficacy for Chil— dren,“ American Sociological Review, 64 (1999):633—660. 3“ For reviews of this recent work on neighborhood variabil— ity, representing a return to the Chicago school of urban ecology, see: F. Furstenberg, Jr., T. Cook, J. Eccles, G. Elder. Jr., and A. Sameroff, Alanaging to [Wake It: Urban Families anaI Adolescent Success. (Chicago: University ofChicago Press, 1999). P. Jargowsky. Poverty and Plaee: Glzettox, Barrios, and tlreAmerican City. (New York: Russell Sage, 1997). S. Mayer, W’lrat [Honey Cant Buy: Family Income and Children} Life Clranres. (Cambridge, Mass: Harvard University Press, 1997). 3— R. Gordon and 1.. Chase—Lansdale, “\Vomens Participa— tion in Market \Work and the Availability of Child Care in the United States.“ (Chicago: Sloan Working Families Center, University of Chicago, 1999). 3‘“ B. Fuller, Y. Choong, C. Coonerty, and F. Kipnis, “An Unfair Head Start: Unequal Child Child Availability Across California.“ (Berkeley: PACE. University of California. 1997). 3“ Percentage of children living in households with more than OHC pC‘I‘SOI] PCI‘ 1‘00111. “‘ For a family of four in 1996. Fry/eral Reg/stun: 61 (1996) March 111212—1214. It‘s important to note that after federal and state income—support benefits are factored in, child poverty rates fall to 900 in Connecticut, 16% in California, and 1400 in Florida. ‘1 Includes spending from state and federal sources (which move through state capitals). Preliminary spending data are courtesy of Gina Adams at the Urban Institute. For family poverty data, see: J. Knitter and S. Page, Ala/7 anr/ li'ael’: State lnitiatirtesfin' lining (Jill/(1,127] anaI Film/lies. (New York: National Center for Children in Poverty, Columbia University, 1996). [BE GROWING up IN POVERTY PROJECT A “3 Against the national cost-of-living index, set at 100 by the National Chamber of Commerce, San Francisco’s index value equaled 184 in 1999. That is, the cost ofliving for the average family (not for a poor family) in San Francisco was 84% higher than the national average. The lowest cost city is Tampa, with an index value of 105, that is, 5% above the national average. Source: Chamber of Commerce Research Association, www.accra.org. 5" Data are from the Current Population Survey and the Bureau of Labor Statistics for the second halfof1998, http:/l stats.bls.gov:80/datahome.htm. 3“ US. Bureau of the Census , Current Population Survey data. (1999) March. Website in note above. P A related analytic approach is to examine the economic and demographic features of all block—groups in the zip codes in which low—income parents reside. This may help in detecting variable job opportunities, public spaces, and school quality. When we conducted this analysis we found an even wider range of economic levels across communities. 3“ Participating families are somewhat less concentrated in California and Florida where 52% and 51%, respectively, lived in the one—quarter of census tracts with the highest concentrations ofstudy participants. For review of this evidence at family and community levels, see Holloway and Fuller (1999). Also: S. Hofferth, K. Shauman, and R. Henke, (,liziiyzr‘tt'i'i,vtir‘s o/‘Clzildren's Early Care and Education Programs: Data/ion! the 1995 National Home/told Education Surrey. (Washington D.C.: US. Department of Education, 1998). 3‘“ For reviews ofthe evidence: D. (ielfand and D. Teri, “The Effects of Maternal Depression on Children," Clinical [Eyre/trilogy Review, 10 (1990):}29—533. V. McLoyd, T. Jayaratne, R. Ceballo, and J. Borquez, “Unemployment and \X/ork Interruption among African American Single Mothers: Effects on Parenting and Adolescent Socioemotional Func- tioning,“ C/Jilrl Deeelo/nnent, 65 (19‘)4):3(i2—389. S. Hofferth, J. Smith, V. McLoyd, and J. Finkelstein, “Achieve— ment and Behavior among Children of \‘C’elfare Recipients. \Velfare Leavers, and Low Income Single Mothers.“ (Ann Arbor: Institute for Social Research, University of Michigan, 1999 manuscript). 1“ For review of research on the social and psychological benefits ofemployment, for low—income women, see chapter 1 in: l.. Hoffman and I.. Youngblade. Aim/Jeri at Work: Effects on Children} \l'e/l-lzeing. (Cambridge, U.I\'.: Cambridge University Press. 1999). Also, for a theoretical overview: M. I\Ioorehouse, “linking Maternal Employment Patterns to I\Iother—Child Activities and Cliildi‘ens Social Competence.“ Developmental llfyr‘lrology, 2T (1991):l‘)3—3()3. ‘*” These items are adapted from Abidin and Bi‘unners (1993) parenting alliance inventory which has been utilized in several studies to assess how the mothers relationship with another adult, who spends considerable time with the child, can influence early development. See: R. Abidin and J. Brunner, “Development ofa Parenting Alliance Inventory,n journal of Clinical Child Psychology, 24 (l995):31—40. *1 Hoffman and Youngblade (1999). *3 For reviews: R. Hess and S. Holloway, “Family and School as Educational Institutions,“ in R. Parke, ed., Review ofChild Development Research, vol. 7 (Chicago: University of Chicago Press, 1984). NICHD Early Child Care Network, “Relations Between Family Predictors and Child Outcomes: Are They Weaker for Children in Child Care?” Developmental Psychol- ogy, 34 (1998):1 1 19—1 128. Holloway and Fuller (1999). *3 P. Barton and R. Coley, Americas Smallest School: The Family. (Princeton: Educational Testing Service, 1992). N. Glenn, “Television Watching, Newspaper Reading, and Cohort Differences in Verbal Ability,“ Sociology ofEducation, 67 (1994):.216-250. “ Fuller, Kagan, and Caspary (1999). ‘5 The MDRC evaluation of Connecticut’s Jobs First program followed a larger number ofwomen and families via administrative records to study employment and wage effects, as detailed in Appendix 1. ‘“‘ Several states have now published surveys of welfare leavers, reporting on the percent of leavers who are employed following exit. These studies often suffer from high levels of exit selection bias. That is, the most difficult to employ women are more difficult to track over time. Brauner and Loprest (1999). “1— Loprest (1999). *“ Horwitz and Kerker (2000). *l) E. Idler and R. Angel, “Self-Rated Health and Mortality in the NHANES—I Epidemiologic Follow—Up Study,” American journal ofPuhlic Health, 80 (1990):446—452. G. Kaplan and T. Camacho, “Perceived Health and Mortality: A Nine-Year Follow—Up of the Human Population Laboratory Cohort,” Americanjournal offpidemiology, 1 17 (1983):.292—304. i" These findings may help to explain why the Medicaid enrollment rate is lower among Connecticut mothers. They see themselves as healthy and therefore do not perceive a need to pursue health insurance options. <1 For reviews of how maternal depression undercuts young children’s early development, see Hoffman and Youngblade (1999) and D. Gelfand and D. Teri, “The Effects of Maternal Depression on Children,” Clinical Psychology Review, 10 (1990):329—353. ‘3 This write-up is available as a working paper from the NICHD Early Child Care Consortium at www.nichd.nih.gov/publications/pubs/earlychildcare. as The Center for Epidemiologic Studies Depression ENDNOTES Inventory (CES-D). I. McDowell and C. Newell, Measuring Health: A Guide to Rating Scales and Questionnaires, second edition (New York: Oxford University Press, 1996). L. Radloff, “The CES—D Scale: A Self—Report Depression Scale for Research in the General Population,” Applications of Psychological Measurement, vol. 1 (1977):385—401. i“ L. Radloff and B. Locke, “The Community Mental Health Assessment Survey and the CES—D Scale.” In M. Weissman, J. Myers, and C. Ross, eds., Community Surveys of Psychiatric Disorders (New Brunswick, NJ: Rutgers University Press, 1986). See Appendix 2 for details on the comparison with national norms. “ Horwitz and Kerker (2000). The Composite International Diagnostic Interview (CIDI is described in: D. Blazer, R. Kessler, K. McGonagle, and M. Swartz, “The Prevalence and Distribution of Major Depression in the National Commu— nity Sample: The National Comorbidity Survey,” American journal ofPsychiatry, 151 (1994):979—986. S“ See Section 2. This 40% estimate is the difference between our original sample ofwomen and the number who had selected a child care provider, minus the 17% attrition experienced over the course of Wave 1 data collection. T On most measures the random assignment ofincoming clients to the experimental or control group appeared to have been done properly by MDRC and welfare office staff. But on this welfare history measure, the subset of the entire experimental group that met our sampling criteria (single mothers with a preschool—age child) had less experience on welfare than the subset of the control group. Future modeling studies will assess whether this difference exerted a significant effect on outcomes. $8 Classroom training included ESL classes, job—specific vocational education, community college or proprietary school course work. a" This is a bit slippery, since many women who have hit their 21—month limit have been given waivers, under the condition that they continue in some kind of “work activity.” See Appendix 1 for MDRC’s research on these clients. 6“ On paper, earlier “workfare” experiments did encourage mothers to find child care and go to work. But a very small fraction with preschool-age children were called into the welfare office to face this pressure. L. Gordon, Pitied hut Not Entitled: Single Mothers and the History ofW/elfizre. (New York: Free Press, 1994). 6' Government Accounting Office, GAO, “Low—income Childhood Care,” GAO/HEHS-OO-ll (Washington D.C., 1999). Schulman, Blank, and Ewen (1999). ()3 Details for California and Florida appear in Fuller, Kagan, et al. (1999). (’5 Figures may not add to 100% due to rounding. BERKELEY - YALE [E ENDNOTES ““ Fuller, Kagan, et al. (1999). (\S Marcia Meyers and her colleagues have studied the rate and determinants of child care subsidy use for different family samples. See, for example, M. Meyers and T. Heintze, “The Performance of the Child Care Subsidy System: Target Efficiency, Coverage Adequacy, and Equity.“ (New York: Columbia University, School ofSocial Work and Public Affairs, 1998, manuscript). For related data on New York state: R. Hernandez, “Millions in State Child Care Funds Going Unspent in New York,“ New York Times (1999):A29, October 25. ““ Burchinal (1999); Cost, Quality, and Child Outcomes Study Team (1995). “— T. Harms, R. Clifford, and D. Cryer, Published ECERS and FDCRS scales. (New York: Teachers College Press, 1997). “S For review, see: D. Cryer, "Defining and Assessing Early Childhood Program Quality,“ in S. Helburn, ed., “The Silent Crisis in US. Child Care,“ Annals oft/n) American Academy of Political and Social Science, vol. 563 (1999):39—55. “‘l Cryer (1999). V“ L. Chase-Lansdale. R. Gordon, J. Brooks-Gunn, and P. Klebanov, “Neighborhoods and Family Influences on the Intellectual Development and Behavioral Competence of Preschool and Early School—Age Children.“ In J. Brooks— Gunn and L. Aber, eds., Neighborhood Poverty: Context and Conseqaencesfor Children. (New York: Russell Sage. 1998). C. Jencks and M. Phillips, The Black-117nm Acliiez'emenr Gap. (\Washington D.C.: Brookings, 1998). 9‘ For example. the HHS—funded child impact studies include five states with random assignment designs, one of which involves the Connecticut family sample. ‘3 Details ofthe CDl and measurement studies are reviewed in: L. Fenson et al., “ "ariability in Early Communicative Development." illonogiaplrs oft/Je Socieryfor Research in (.‘lri/ri Development, serial no. 242. vol. 39 (Chicago: University of Chicago Press, 1994). 71 The standard MacArthur scales are available for infants. age 8—16 months. and toddlers, 16-30 months of age. Based on earlier studies with the MacArthur we assumed that children would be somewhat delayed in their language and communications skills. So. we used the toddler form for our young cohort. those aged 12—23 months. and the preschooler form with our older group. age 24—42 months of age. 9‘ We expected that the Connecticut children‘s means might be lower due to sample selection bias. Remember that we initially tried to conduct the direct child assessment at the child care setting. leading to a higher completion rate among those children in child care and particularly those attending center—based programs. In Connecticut, a larger proportion of child assessments were done in the home. since fewer Em GROWING UP IN POVERTY PROJECT FA children were attending centers. Yet still we see higher direct assessment scores for Connecticut toddlers and preschoolers. '1 The authors of the MacArthur also are developing a picture version of their scales. For review: E. Ring, Assessment of Language Comprehension and Production: C/Jila’ Performance versus Parent judgment. (San Diego: San Diego State Univer- sity, 1999, unpublished thesis.) 7’ The EASI instrument is detailed in: A. Buss and R. Plomin. A Emperment Theory of Personality Development. (New York: Wiley 8C Sons, 1975). The CBC, authored by Thomas Achenbach, is detailed in his “Manual for the Child Behavior Checklist / 2—3 and 1992 Profile,” available from the University of Vermont’s department of psychiatry. Normative samples and scores are discussed in: T. Achenbach, C. Edelbrookm and C. Howell, “Empirically based assess- ment of the behavioral/emotional problems of2—3 year—old children,” journal of Almormal C/Jild Psychology, 15 (l987):629—630. _- Our earlier research paper details how the parenting practices of. and child care quality selected by, were both stronger among women with more extensive work experience. But any discrete effect of working disappears after taking into account maternal education levels, child’s age, and other basic demographic factors. This suggests that personal resources and neighborhood conditions precede employability. not necessarily that maternal employment yields positive effects for kids perse. This question requires additional analysis. See: Fuller, Kagan, and Caspary et al. (1999). 1“ In a related study MDRC examined the post—welfare experiences of a group of families whose welfare cases were closed when they reached Jobs Firsts 21—month time limit on cash assistance receipt. Funded by the State of Connecticut. the Post—Time Limit ’l‘racking Study, focused on six areas: Bridgeport, Hartford. Manchester. New Haven. Norwich, and \Waterbury. 9" The second survey includes a special module ofquestions focused on the well—being of respondents children; these questions are asked of all respondents who have at least one child between 3 and 12 years old at the point the interview takes place. 3“ A total of 6.1 13 people were randomly assigned. However. four categories of people are excluded from the analysis: 6‘9 cases that included no adult recipient at the point of random assignment: 38F two—parent cases; 240 cases that were randomly assigned in error: and 8 cases for which no social security number was available. S1 The Jobs First research sample does not include the entire welfare population in Manchester and New Haven. To control the workload for staff, only half ofthose who applied for benefits between January and July 1996 went through the random assignment process: the others were enrolled into Jobs First. but are not part of the study (this selection was made randomly by caseworkers). Several thousand applicants and recipients who had been previously randomly assigned for an earlier study of Connecticut’s prior welfare reform were not assigned again; they too were placed directly in Jobs First. 33 This figure includes 254 people who were initially identified as having a child under 18 months old at the point of random assignment, and 59 people who did not have a child under 18 months old at random assignment, but who nonetheless had a child between 18 and 36 months old when interviewed. The second group includes, for example, people who gave birth within 18 months after their random assignment date. 35 For reviews of the impact of these dimensions of quality on early development, see: M. Burchinal, “Child Care Experiences and Developmental Outcomes,” in S. Helburn, ed., “The Silent Crisis in U.S. Child Care,” Annals oft/1e American Academy of Political and Social Science,” vol. 563 (1999):73—97. 3* Cost, Quality, and Child Outcomes Study Team, Cost, Quality, and Child Outcomes in C/Jild Care Centers. (Denver: Economics Department, University of Colorado, 1995). 3; Researchers working on the unprecedented National Day Care Study in the mid-19705 did observe structural facets of quality, including child group size and childzstaff ratios, for a large national sample of centers. But measures of social organization and human process dimensions of quality were not assessed until the Child Care Staffing Study was com- pleted in 1989. M. Whitebook, C. Howes, and D. Phillips, W/lro Cares? C/Jild Care Packers and the Quality of Care in America. (Oakland, Calif: National Child Care StaHing Project, 1989). The earlier federal study: C. Coelen, R Glantz, and D. Calore, Day Care Centers in the United States: A National Profile, 1976—1977. (Cambridge, Mass.: Abt Associates. 1978.) N‘ E. Galinsky, C. Howes, S. Kontos, and M. Shinn, “The Study of Children in Family Child Care and Relative Care.” (New York: Families and Work Institute, 1994). The study suffered from selection bias: 947 parents were reached by phone and deemed eligible to participate, using a home—based care provider, but just 145 providers allowed researchers to observe their settings. An additional 81 providers were then selected independently of the first sampling process. V M. Whitebook, L. Sakai, and C. Howes, NAEYCAccredi— ration as a Strategy fi2r Improving Child Care Quality. (Wash— ington D.C.: National Center for the Early Childhood Work Force, 1997). "* Administration for Children and Families, “Access to Child Care for Low—Income Working Families.” (Washington D.C.: US. Department of Health and Human Services, 1999). 8" The most complete discussion of the MacArthur measures appears in: Fenson er a1. (1994). This volume also includes normative item scores for the infant and toddler forms. ENDNOTES “0 R. Arriaga, L. Fenson, T. Cronan, and S. Pethick, “Scores on the MacArthur Communicative Development Inventory of Children from Low and Middle—Income Families,” Applied Sociolingistics, 19 (1998):209—223. "1 For reviews see: Gelfand and Teri (1990), Hoffman and Youngblade (1999). "3 Blazer et a1. (1994)‘ ("I L. Radloff and B. Locke, “The Community Mental Health Assessment Survey and the CES—D Scale.” In M. Weissman, J. Myers, and C. Ross, eds., Community Surveys of Psychiatric Disorders (New Brunswick, NJ: Rutgers University Press, 1986). “4 M. Weissman et al., “Assessing Depressive Symptoms in Five Psychiatric Populations: A Validation Study.” American journal opridemiology, 106 (1997):203—214. 9‘ Nord, Jamison, and Bickel (1999). BERKELEY — YALE [IE ENDNOTES m GROWING up IN POVERTY PROJECT A APPENDIX 1 Evaluating Connecticut welfare reform: MDRC ancl Yale-Berkeley studies This appendix reviews three linked studies of Connecticut’s welfare reform initiative, Jobs First. Connecticut is one of three states included in the Growing Up in Poverty (CUP) Project. The Jobs First evaluation is being conducted as a random assignment experiment, allowing us to discern the discrete effects ofwelfare reform, and it illuminates issues relevant to California and Florida. The three related studies for Connecticut include: I The overall evaluation of Jobs First, including assess- ments of adults and children at 18 and 36 months following random assignment to the new program or a control group. This effort is directed by the Manpower Development Research Corporation (MDRC) in New York. Tracking of preschool—age children and their families, focusing on the parent’s quality of family life, parenting practices. social and economic support of child rearing, and the quality of child care selected by mothers as they face welfare-to-work pressures. The focal children were between 12-42 months of age when they entered the study. 3 Tracking older children and their families, focusing on how the mental health of mothers and other factors influence her employability and medium—term earnings. These children were between 3 and 9 years—old when they entered the study. The Overall Jobs First Evaluation The evaluation was originally required as a condition of the federal waivers that allowed Connecticut to implement Jobs First. In 1997 Connecticut received enhanced federal funding from DHHS to support continuation of the evaluation. The state later received a second federal grant to expand the study to examine Jobs First‘s impact on the well—being of children. The evaluation began in 1996 and is scheduled to end in late 2001. It focuses on two of the state’s welfare offices, Manchester and New Haven.“ It includes three major components: Impact analysis. This provides estimates of the changes motivated by Jobs First in clients’ employment rates and earnings, rates and amounts of welfare receipt, family income, the extent of welfare dependency, child well-being, and other outcomes, compared to the welfare system (AFDC) that preceded it. Implementation analysis. This component examines how Jobs First is operated by staff in the research sites. It assesses whether Jobs First’s policies have translated into concrete changes in the day—to—day operations of the welfare system and identifies obstacles that have been encountered. This information is necessary in order to understand the impact result, and may help DSS identify ways to improve the program’s performance. APPENDICES Benefit-cost analysis. This analysis uses data from the impact study, along with fiscal data, to compare the financial benefits and costs generated by Jobs First for both taxpayers and eligible families. Research Design and Data Sources The evaluation uses a random assignment research design to assess the program’s impacts. During 1996 and early 1997, several thousand welfare applicants and recipients in Manchester and New Haven were assigned, at random, to one of two groups: the jobs First group, whose members are subject to the welfare reform policies, and the Aid to Families with Dependent Children (AFDC) group, whose members are subject to the prior welfare rules. Because people were assigned to the groups through a random process, any differences that emerge between the two groups over time— for example, in employment rates or income—can be attributed to Jobs First. The evaluation will eventually follow members of the two groups for up to four years. The evaluation uses a wide variety of data sources to assess Jobs First’s implementation, effects, and costs: Baseline data. Virtually all sample members completed (via a brief interview) a one—page form describing their demo— graphic and socio-economic characteristics at the point they were randomly assigned. Administrative records. The State of Connecticut has provided MDRC with data on individuals’ monthly cash assistance and Food Stamp payments and quarterly earnings in jobs covered by the state’s Unemployment Insurance (U1) system. Follow-up surveys. M DRC has subcontracted with Roper Starch Worldwide (RSW) to conduct two surveys of the experimental and control groups. The surveys capture information that cannot be obtained from the administrative records (e.g., job characteristics, household income, and participation in work-related activities). The Interim Client Survey, completed in 1998, was adminis— tered about 18 months after each respondent’s date of random assignment. A total of 772 people were interviewed, including the preschool and school-age child and family samples co-flnanced by the Yale-Berkeley study teams. The second survey, administered about 36 months after each person’s random assignment date, will include more than 2,000 respondents. It began in mid-1999, and is scheduled to be completed in mid-2000.09 Early Evaluation Reports MDRC has produced several reports as part of the Jobs First Evaluation and the Post—Time Limit Tracking Study. These write—ups are available from Dan Bloom at MDRC in New York. The project’s final report—looking at program partici- pants 36 months after entry—is scheduled for completion in late 2001. A companion report will examine the program’s impact on the well-being of children. BERKELEY - YALE APPENDICES Collaborative Studies with TWo Yale-Berkeley Research Teams The State of Connecticut and MDRC came together with Yale University researchers in 1997 to explore how jobs First and collateral child care programs may affect the well—being of participating parents and children. They were joined by scholars from the University of California, Berkeley. The second study, led by Yale‘s Sharon Lynn Kagan and Berkeley’s Bruce Fuller, focuses on single mothers with preschool—age children. The third study, led by Yale’s Sarah Horwitz, assesses families with older children. The Yale— Berkeley collaboration is funded by the Connecticut legisla- ture and foundations and federal agencies that support the three—state GUP Project. Research Design and Family Samples All three studies being released in February 2000, including the present one, use data from the Interim Client Survey. Figure A.1 illustrates how the samples for the overall survey were selected. The upper—most box represents the full research sample for the Jobs First evaluation. It includes 4,803 single—parent cases randomly assigned to the Jobs First and AFDC groups between January 1996 and early 1997.80 Individuals were randomly assigned when they catne to the DSS office to apply for welfare or to have their eligibility for benefits reviewed.“ As the figure shows, a subset of the full research sample was selected for the Interim Client Survey. That survey included three separate sections: Core interview questions. All respondents answered a set of questions focusing on participation in employment—reIated activities since random assignment, characteristics of all jobs held since random assienment, household income in the b , month prior to the mtervtew, and other issues. MDRL led the analysis of these questions. Iblmg (Iii/d modules. This set of questions was administered to all respondents with a child between 18 and 36 months old at the point ofthe interview. It focused on child care and child well—being. The GUI) I’roiect funded this module and is leading its analysis. Older (Iii/(1' modules. This set ofquestions was administered to all respondents who had a child between 3 and 9 years old (but no child between 18 and 36 months old). Its content was similar to that ofthe young child module. The Connecti— cut legislature funded this module and Professor Horwitz is leading its analysis. Figure A] shows that, in order to maximize the number of respondents for the young child module, the sample that was selected for the Interim Client Survey first included all females in the full sample who were randomly assigned between September 1996 and January 1997 (the end ofthe sample intake period) and who had at least one child under 18 months old at the point of random assignment; these m GROWING UP 1N POVERTY PROJECT +1 children would have been between 18 and 36 months old at the point of the interview (71:286). Second, the sample was comprised of a random subset of all other sample members assigned during the period (11:678). Total Family Sample—18 months After Entry In sum, the entire Connecticut sample included 964 people selected for the fielded survey sample. RSW located and interviewed 772 (80 percent) of these people. A total of 293 of the respondents completed the young child module for the three-state GUP analysis.“ A total of 288 of the respondents completed the older child module for the Yale public health study. For greater comparability with California and Florida, the CUP analysis includes 18 cases where the focal child was 36-42 months ofage (total n=31 I). The remaining 191 respondents completed only the core module. MDRC’S analysis focuses on all 772 respondents. However, since individuals with young children were over sampled, MDRC’S analysis used a weighting process to ensure that such cases do not disproportionately affect the results. Figure /\.I Connecticut's Jobs Iiirst I’rogram Key Samples L'sed in Connecticut Welfare Reform Studies FuII Research Sample for the Jobs First Evaluation F Singleparenl taxes randomly assigned 1/96 » 20)— 4,803 Fielded Sample for the Interim Client Survey Samples lIM'kI m ['1‘ .‘\II single mothers randomly .migncd 9/90 » 1/9? t . . . I l 1” random \ubset ol .1“ other sample members assigned during period. o M ‘|l\l 964 liralnation Interim Client Survey Respondents > \.II L'Uliiplctcd IIk‘ core module "'72 I Respondents who Responldenctls \Eho , . . . com ete t 6 51:33, 3:523:23“ _____ , Yoggrlgjgdatguk o..t..c*’h..dM..t... and Florida 288 29} A l A Sample for the Sample for the (ironing up in \ale pttl‘lit health I‘m 1‘] I\‘ Project \III\I\ [PPENDIX 2 Comparing GUP Children, Mothers, and Child Care Quality to National Norms Child Care Quality How does the child care selected by participating mothers compare to that available to middle—class or low-income families assessed in earlier studies? A handful of national investigations, conducted in various states over the past decade, have revealed quite uneven quality. This raises the issue of whether the welfare-to-work push felt by TANF mothers places their children at greater risk, compared to other families who already confront local child care markets that offer highly variable quality. The empirical answer may vary between center-based programs and home— based child care (licensed homes and kith or kin settings) Two dimensions of child care quality predict early cognitive and social growth in young children at age 3 or 4: the education level of the adult(s) who cares for the youngster and observed quality of the materials and social organization of the day—care environment.“ The latter set of variables is typically measured with the Early Childhood Environment Rating Scale (ECERS) for children in centers and the Family Day Care Rating Scale (FDCRS) for youngsters in home— based settings, as described in Section 8. Summary of Project Findings We reported in Section 8 that the mean ECERS score, averaging across all centers situated in the five GUP cities, equaled 4.3 on the 7—point scale. Only 21% of the centers displayed a good or excellent level of quality (with average items scores at 5 or above). Mean differences among cities were stark. Centers in California exhibited the highest observed quality overall, averaging 5.2 points on the ECERS, compared to just 2.3 in Connecticut and 3.3 in Florida. Variability among cities was evident in California: San Jose centers showed the highest quality on average, 5.8 on the ECERS, compared to 4.6 among San Francisco centers. The average F DCRS score across all five cities was even lower, equaling just 3.0 for licensed child care homes and 2.7 for un— regulated kith and kin caregivers. This translates into 13% of all home-based providers falling into the good or excellent range. We found that 65% of all center based teachers had completed some postsecondary education. Attainment levels for women working in home-based settings were considerably lower: just 51% oflicensed home providers and 26% of kith and kin caregivers had pursued any type of schooling after high school. Quality Compared: GUP Families’ Child Care Compared to Wider Populations How do these tandem indicators of quality compare to earlier national studies? APPENDICES I In 1997 and 1998 our Project team visited about 45 centers and 45 family child care homes in both California and Connecticut (total useable n=178 providers). We call this the GUP 1997—98 Validation Study, since we were assessing the validity of phone interview items, against the conventional observational measures. We first stratified all zip-code areas based on their median household income for San Francisco Bay Area communities and the New Haven—Hartford region. From these income strata we randomly selected centers and licensed or registered home—based providers. This allowed us to take stock of the quality ofproviders serving a wide range of families in these two states. We also developed phone interview items that proved to be moderately to highly correlated with observa- tional measures of quality. For this diverse array of centers we found that the average ECERS score equaled 3.9 (across all quality indicators, ranging between 1—7). The mean FDCRS score equaled 2.8 (on a similar 7—point scale). We found that just 65% of all center teachers and 37% of family child care directors had completed some postsecondary training. See inside the back cover for information on this research paper. 3 In 1993 researchers began to observe the quality of 100 child care centers in each of four states, working under the Cost, Quality, and C/Jild Outcome Study (CQO).X4 These centers were randomly selected, serving a diverse and representative range of families within each state. The mean ECERS score was 4.0. Just 14% ofall centers were ofgood or excellent quality (averaging 5 or above). Centers in California were of higher quality overall, averaging 4.5 points; Connecticut centers scored 4.3 on average. Nonprofit centers scored a half-point higher on average, compared to for—profit organizations, a significant gap in quality, and one that we see in the GUP quality assessments for Tampa’s for— profit centers. The CQO team also found that 39% of surveyed teachers in nonprofit centers had completed some postsecondary training, compared to 33% in for-profit organizations. I In 1988 the first extensive observational study of center quality began, drawing on a stratified sample of center classrooms in five cities.8s Scholars working on the National Child Care Stufling Study first stratified communities by the household income level of census tracts in selected cities— Atlanta, Boston, Detroit, Phoenix, and Seattle—then sampled centers across low, middle, and high-income tracts. A total of 227 centers participated across the five cities. The average ECERS score equaled 3.6 for this national sample of centers. Importantly, accredited centers—more frequently found in middle—class and affluent communities—scored considerably higher, a 5.0 on average. For-profit centers scored just 3.3 on the ECERS, compared to 4.3 for non- church nonprofits. Across all center teachers, 66% had completed some college or more schooling. I The only observational study of home-based child care with multiple sites, prior to the present report, was published BERKELEY — YALE [EE APPENDICES in 1994 by the Families and Work Institute of New York. It included quality measures for licensed child care homes and kith or kin caregivers.80 Families came from a wide range of backgrounds with 47% earning more than $40,000 annually. Twenty—nine percent had completed only some college; another 33% had received bachelors degrees. These families were drawn from popula— tions residing in Los Angeles. Dallas—Fort Worth, and Charlotte. North Carolina. The average FDCRS score equaled 3.4 (on the 7-point scale). A total of‘ 56% of~ all home-based care providers had com— pleted some postsecondary education. I One California study provides additional comparative data, focusing on accredited and non-accredited centers. A diverse range of centers was randomly sampled in Palo Alto, San Jose, and Santa Cruz in 1994 by Marcy Whitebook and her colleagues?" Families served by these 102 centers were diverse across lines of‘ family income and ethnicity. The average ECERS score for accredited centers equaled 5.2, compared to 4.2 for nonaccredited centers. Among center teachers, 51% had completed some postsecondary education. Did GUP Participants Find Child Care Settings of Lower Quality? The short answer is, yes. Compared to the quality of‘child care selected by wider, more middle—class samples of parents, CUP mothers found centers and less formal caregivers that displayed lower levels of‘ quality. The widest gap is for young children placed in home—based settings, either licensed homes or kith and kin providers. The average FDCRS score for CUP caregivers equaled 2.8, compared to 3.4 in the wider national sample studied by the Families and Work Institute. Authors of‘ the 1994 study also found that 5600 of‘these caregivers had completed some postsecondary education, compared to just 26% of‘CUP kith and kin providers, the bulk of‘home—based caregivers. The quality oficenter—based programs found by participating mothers is more comparable with the quality of care found by more middle—class populations. But inequalities faced by TANF parents remain wide in Connecticut and Florida. The average ECFRS score among CUP centers equaled 4.3. But this level is boosted substantially by relatively high quality in California. The mean FCFRS scores in Connecticut was just 2.3 and in Florida, 3.3. These levels are far below the national averages found in the CQO study and the 1994 accreditation study. The Connecticut centers accessed by CUP mothers also displayed much lower quality relative to the range of centers studied in CQO. Another way to look at this quality gap is to compare the average of‘4.3 on the liCliRS, observed for CUP centers, relative to the 4.9 and 5.0 observed for accredited centers, typically concentrated in affluent suburban areas. In concrete terms, this means that centers selected by CUP mothers were less likely to have careful supervision of‘children and a wide [IE GROWING UP lN POVERTY PROJECT A selection of books and materials for pre-literacy skills, compared to typical centers situated in middle—class communities. One piece of good news is that centers accessed by CUP families did have teachers ofcomparable school attainment, relative to the more middle-class range ofcenters studied in the early national investigations. Sixty—five percent ofCUP teachers had some college experience beyond high school, compared to 36% in CQO study and 66% among teachers in the 1988 staffing study. Utilization of Child Care Subsidies Data are scarce on the share ofTANF or working—poor families who utilize their child care subsidy (often eligible for two years of‘support). The Department of Health and Human Services published a report in 1999, showing very low rates of‘take—up among all families eligible for federal block—grant subsidies, including welfare-poor and working- poor families.”8 However, this report understated take—up rates for states with sizable subsidy programs or state—funded preschool initiatives. Nor were Head Start enrollments included. It does appear that subsidy utilization rates vary widely by state and among counties. For example, in parallel work we found that only about one in five new entrants to Los Angeles County‘s welfare program, who enter the main welfare—to—work stream. take up their child care voucher or find a subsidized slot in a center-based program. Such administrative data under count other forms of subsidized care, for example, women who find a space at a state—funded preschool. Yet this 20 percent nominal rate is far below the 50% utilization rate reported by CUP mothers in San Francisco. Levels of Child Development Language and Communicative Proficiencies To assess early cognitive development we used segments of the MacArthur Communicative Development Inventories (“the MacArthur“ or the CD1). as discussed in Section 9. This set of. measures focuses on early communicative skills of infants and toddlers, age 8—30 months ofage.“ These assessment tools have several advantages: (1) parents are asked to report on their child‘s communicative proficiencies, including gestures, word recognition and production, and complexity of‘speech, allowing inclusion in an interview, (2) certain measures are amenable to direct verification with the child, and (3) national norms are available by which the CUP children can be compared. We knew that we were taking one measurement risk with our older cohort of‘children. age 24—42 months of‘ age: the measures could he too easy for them. given that the toddler forms of‘the MacArthur are calibrated for kids age 16-30 months of age. l\’leasurement studies have shown that children from low-income families perform less well on most scales, moving the distribution toward the lower end of perfornrance.”U Yet, after collecting maternal report and direct assessment data for our older cohort. a ceiling effect was observed, that is, the measures overall were not effective in discriminating levels of development for these older toddlers and preschoolers. The distribution for some individual items is fine. Future work will use these items. The direct assessment of the younger CUP children, age 12— 23 months of age, in California and Florida worked well. We did discover that mothers over estimated the word recogni— tion and production proficiencies of their children. We have set aside for the moment the Connecticut assessment data, given possible sample selection bias below. Children} proficiency [eve/5 against national norms. Table A2.1 reports. for all words directly assessed with CUP children, the percentage that recognized each word (according to their mother). the actual percentage who identified a picture of the word, and the national norm. Note that the final 6 words were used only on the direct assessment. For this younger cohort. the children were 18 months of age, on average, at the time of the maternal interview. Child assessments were conducted 4 months later, on average, at 22 months. The norming studies are set at the percentage of children. at 24 months, who can recognize each word. This corresponds to the columns in the table. It‘s important to note that the national norms at 24 months are for recogni- tion and production of the word. Our direct assessment involved simply having the child point-out a picture of the word, not orally produce it. Table A2.1 GUP Toddlers' Word Recognition Levels Against National Norms (Younoer cohort only, aoe 12—23 months of a ’e) D . ‘O E) \V'ord CUP mother CUP child report assessment at 18 months at 22 months National norm at 24 months ("0 recognize word) (00 identify picture) (00 produce word) Car 7400 44% 52% Dog 8000 55% 52% Hat 5700 32% 76% Juice 84% 43% 52% Kitty/ meow 86% 48% 52% Lamp 33% 22% 29% Radio 56% 1 1% 39% S hoe 8 2 % 5 5 % 5 7% Television 76% 50% 65% Tooth 58% 25% 67% Book na 320/0 840/0 Blocks na 21% 62% Doll na 33% 63% Puzzle na 69% 45% Table na 31% 51% Teddy bear na 43% 57% Number of 190 108 1,308 assessments Median age 18 mos. 22 mos. 24 mos. APPENDICES Social Development and Behavioral Problems Toddlerr. We included two social—emotional attributes of toddlers, drawing from the broader range of subscales contained within the EASI instrument. These included emotionality, such as, “(child) cries easily”, “is easily fright— ened”, “reacts intensely when upset.” And we included items regarding the child’s sociability, such as, “(child) tends to be shy”, “makes friends easily”, and “prefers playing with others rather than being alone.” For each attribute, the parent is asked to indicate whether this “is not at all like my child (scale score, 1), to “is very much like my child” (scale score, 5). Differences on the two subscales that we employed, between CUP toddlers and national norms, are reported in Section 9. Preschoolers. We included partial or complete items for six of the seven CBC subscales, excluding somatic problems displayed by the child. Our 400 preschoolers in California and Florida (with complete data) showed significantly higher scores on four of the six subscales: aggressive and anxious behavior (as detailed in Section 9), impulsive and destructive behavior, and other problems (especially being restless, can’t sit still, and sulks a lot). We did not observe differences in sleeping problems or acting depressed or down. Scores for wanting attention from the mother were higher among participating children, relative to normative levels. Mental Health, Maternal Depression Composite International Diagnostic Interview (CIDI) The mothers level of motivation or depression is a major concern since this factor is a consistent and strong predictor of early child development."1 The CIDI, as described in Section 6, is a clinical measure of severe depression. When a specified level is reached, psychiatric specialists consider the individual to be suffering major depression (DSM—III) in ways that are quite debilitating. As you saw in Figure 6.5, Participating mothers in Connecti— cut displayed a prevalence rate ofjust over 15% on the CIDI items. How does this compare with national samples of adults, looking across the general population? The most recent epidemiological study using the CIDI was published in 1994, based on a large national probability sample (71:8,098). Dan Blazer and his colleagues found a 4.9% rate of occurrence among the general population. Rates varied by the age and ethnicity of women in the sample. For example, just 3.5% of all Anglo women, age 25—34 years of age, showed signs of severe depression in the month prior to the interview. This compared to 5.6% among black women and 7.5% among Latina women in the same age range?2 Center for Epidemiological Studies Depression Inventory (CBS-D) The CES-D has been administered to large samples by its author, Lenore Sawyer Radloff at the National Institute of Mental Health (NIMH), and by other scholars. One of the largest norming studies was conducted with 3,845 adults who BERKELEY — YALE [III APPENDICES resided in Maryland and Missouri. Dr. Radloff and her colleagues found that one in live adults surpassed the threshold level of displaying symptoms of depression.‘H This compares to 48% of the GUP mothers who crossed the cut- off score of 16. over twice the level of depression observed in the norming study. Another way to compare our participating mothers to other populations is to look at average scores on the CBS—D. The CUP women scored 16.9 on average. This compares to 7.5 to 9.3 for the general population, depending on the study.” Hunger Our interview items on hunger and food rationing were taken from standard surveys periodically run by the US. Department of Agriculture (USDA). This allows us to compare our participating families to national surveys and normative levels established by the USDA. Table All compares how our participating mothers responded to three standard questions. compared to national samples studied by the USDA in 1998 (illustrated in Section 5).0i Table A2.2 GUP Hunger and Food Rationing ltems, Compared to USDA National Norms (1n the prior 12 months) APPENDlX 3 Do Families with Children Directly Assessed Differ? on responding. often or sometimes true (1UP mothers USDA national norm Interview question The food that I bought 31% 1 1% inst didnt last. and we didn‘t have enough money [0 gCI 1110 l'C. 1 relied on a few kinds 3300 1400 of low-cost food to feed my children. Did you or other adults 13% (300 in the household cut the size of your meals or skip meals because there wasn‘t enough food.> (yes) HE GROWING UP IN POVERTY PROJECT A We preferer to directly assess children’s language develop— ment at the youngsters child care setting. This enabled us to make one field visit to conduct the child care quality and the child assessment on the same day. But ifthe a child was not attending a child care setting, or the mother was not comfort- able in having us visit the provider's settings, then we sched— uled a home visit to conduct the MacArthur child assessment. Predictably, the participation rate for children attending a center-based program was highest, equaling 67% of this subset. For children in licensed family child care homes, we obtained direct child assessments for 55%. But for youngsters being cared for by kith or kin, just 37% were directly assessed. Connecticut is problematic, at least for the toddler subgroup, since very few were attending center-based programs. \We have yet to detect any other source of systematic selection bias in which subset of toddlers and preschoolers were directly assessed. There are no significant differences in participation rates by state, ethnicity, or membership in experimental or control group for Connecticut. APPENDICES APPENDIX 4 Figure 3.2 Children in poverty—San Francisco COUI’itV 7 Percent of Children in Poverty Block Group By US Census 3 30-10 1 [:111-25 ' -26-50 !51-100 WW No Data 0 Locations of Participating Families D County Lines \ Highways Source: 1998 of the Census,1990 PACE. 1998; Claritas, US Bureau SAN FRANCISCO COUN Y i 7 _ reated by Greenlnfo Network www.greenlnfo.org Brisbane Ma I SAN MATEO COUNTY I Daly City BERKELEY — YALE [IE Figure 3.3 Children in poverty—Santa Clara County \ 7/ anta lara \ I a E Percent of Children in Poverty By US Census Block Group [30—10 E1145 -26-50 -51-100 l:] NoData amilies D County Lines /;V Highways Source: PACE, 1998; Claritas, 1998 US Bureau of the Census Figure 3.4 Children in DOVEFtV—NEW Haven COUHtV Percent of Children in Poverty By US C sssss Block Group (:| 0 - 10 “T“‘i-fi. 11 - 25 - 26 50 - 51 100 E No D t 0 Lo t f Part p t 9 Families A DCounty LIne /\/ Highw y 0 1 2 3 Mb N 1 Source: Map created byGreenlnfo Network PACE, 1998; Claritas, 1998 www.greeninfo_org US Bureau of the Census 1990 Figure 5.5 Children in poverty—Hillsborough COUHtV, Florida cent of Children in Poverty By US Census Block Group i 1 Per i :0—10 :111-25 -26-50 -51-100 0 Locations of Participating Families D County Lines 4%\»/ Highways Source: PACE, 1998; Claritas, 1998; US7Bureau7of the Census, 199077 1 0 177772 374 5Miles Map Creatd by Greenlnfo Network www.greeninfo.org Initial pl‘OiECt papers I Growing Up in Poverty Project, "1999 Progress Report.” (Berkeley: University of California and Yale University, 1999). I S. Holloway and B. Fuller, ”Families and Child Care: Divergent Viewpoints.” In S. Helburn, ed., "The Silent Crisis in US. Child Care," Annals of the American Academy of Political and Social Science, 563 (1999):98-115. I B. Fuller, S.L. Kagan, J. McCarthy, 0. Caspary, D. Lubotsky, and L. Cascue, "Who Selects Formal Child Care? The Role of Subsidies as Low— Income Mothers Negotiate Welfare Reform." (Berkeley: University of California and Yale University, 1999). Paper presented at the Society for Research in Child Development, Albuquerque. I B. Fuller, S.L. Kagan, and C. Caspary, "Variation in Poor Children’s Home and Child Care Settings: Does Maternal Employment Matter?” (Berkeley: University of California and Yale University, 1999). Paper for the Joint Center for Poverty Research Washington, DC. conference. I S. Holloway, S. L. Kagan, B. Fuller, L. Tsou, and J. Carroll, "Measuring Child Care Quality with a Telephone Survey.” (Berkeley: University of California and Yale University, 2000). Papers are available for $15 from the project center: Graduate School of Education 3653 Tolman Hall, University of California Berkeley, CA 94720. Please call 510—642-7225 or on the Web: http:\\pace.berkeley.edu EYLIBRARIES llllllllllllllllllllllllllllllllllllllllllllllllllllllllllll T0 learn more Cl“ Administration for Families and Children, US. Department of Health and Human Services wwwacfdhhsgov Center for Law and Social Policy www.clasporg Child Trends lnc. www.childtrends Heritage Foundation wwwl'ieritageorg Hudson Institute www.hudson.org/wpc/ Joint Center for Poverty Research University of Chicago and Northwestern University www.jcpr.org Manpower Demonstration Research Corporation www.mdrcorg National Center for Children in Poverty wwwresearchforumorg Urban institute wwwnewfedera/ism.urbanorg